HWBOT 'Important! Opinions regarding HWBoints algorithm revision'

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Oct 30, 2009

I'm all ears for your opinion ! :)

I'll be updating the starting post with the answers to certain questions

------------------------------------------------------------------------------

Thor941 said: I think it won't change the fact that "big" clockers with huge HW will still be in the top rankings.

Of course, the global ranking will always be ruled by those with cooling skills, tweaking skills and the expensive hardware to use them on.

K404 said: This idea makes slightly older, cheap cards more viable for global points, which while it isnt a bad thing... it will massively increase the competition in these categories..... LN2 will be necessary for hardware points for the cards, which will push more people out ...

We are actually trying to improve the hardware points as well. This does take quite a bit of work in balancing, though.

Linuxfan2000 said: How would you rank the dual card GPUs?

The initial idea would be to have two rankings:

1) Single GPU
2) Multi GPU

Multi-GPU would then indeed be dominated by those who can afford multiple cards. However, it seems that in the current generation of hardware even 3D benchmarks are CPU-limited. 4 cards won't be more powerful than 3 per definition.

It's quite difficult to have a solution that works out perfect for all situations. Creating more global categories also means that people have to buy more hardware to compete at the top. Also, it puts more stress on our server.

ether.real said: Instead of the proposed change, I would much rather see a push to using Vantage X instead of Vantage P in hwbot league. Remove the CPU bottleneck and let GPU power be the main determinating factor in getting boints in the GPU tests.

The fact that dual GPUs are now in the 'single'-category is part of the problem of CPU-bound 3D tests. The more powerful you make the 3D part of your system (VGA), the more powerful you need to make your 2D part (CPU) to see scaling from adding VGA OC.

ether.real said: So, when do we classify a Yorkfield as a "multiple" Wolfdale? Cause that makes just as much sense.

There are many differences involved in this, the most apparent possibly the fact that there are only two benchmarks that really scale with multi-core design being Wprime 32M and Wprime 1024M whereas nearly every benchmark scales with multi-GPU design. We may have been postponing splitting up the socket for Wprime a bit too much, though.

Another very big difference is that within the GPU there are in fact multiple processors. We don't split them up in 192, 216 or 512 categories, though.

ether.real said: What about dual GPU cards that don't have a single GPU equivalent? How would those get classified?

Initial suggestion described it as two seperate categories:

-Single GPU
-Multi GPU (dual, triple, quadriple)

In the latter in other words.

TiN said: There is the same ranks with CPU sockets, not cores?

As for Wprime: noted on the agenda.

ether.real said: All sports and competitive leagues are "Pay to Play". You are not going to be able to equalize the dollar influence unless you yourself are willing to buy top end hardware for everyone.

I agree with you. However, I also believe that if there's an easy way to reduce the dollar influence we should at least consider it. It may not change that much for the so-called top dogs, but if it changes something for the less fortunate enthousiasts I'm already happy.

Thor941 said: Isn't it possible to make a simulation on the top 100 scorers or I do'nt know how many ...

Next meeting is on monday and it's on the agenda.

komadyret said: This should also be followed by a increase in the hardwareboints ceiling from 300 to lets say 500 points?

Good suggestion. We have already discussed this possibility in the staff section as well and it came to our attention that increasing the hardware points is in fact a very short-term solution as:

- Hardware categories are added by the day
- For one hardware category, there are 7 (CPU) or 6/12 (GPU) new rankings
- For each newly added benchmark there are countless new hardware points available

Whether or not we increase the hardware points cap for the individual league, there will always be a point where the cap forms an issue. Raising the cap will work just fine for a while, but at a certain point we will return to the same situation as we are in today. Raising it again will work fine again ánd bring us back to the same situation. Eventually, we will have a hardware cap that is higher than the amount of points you can win in global categories. The problem is:

The Overclockers League is supposed to be filled with those who show extreme skill using extreme cooling on extreme hardware. The top results, in other words.

This is the main reason why we set the hardware cap in the first place. If we increase the cap so much that it doesn't have it effect anymore, why bother having it in the first place.

komadyret said: Making a third gpu-global-class will seriously shift the weight of the points in the gpu-direction. I'm not sure it's a good thing to make the cpu's so much less important in the big points ...

Again, excellent remark. Creating single, dual and more GPU categories is in itself no problem, but it will most definitly shift the weight of the points to GPU benchmarking. In particular: multi-gpu benchmarking. In other words, instead of making it a more balanced game, we would be making it a more expensive game.

Dualist said: As you are surjesting you either have to buy one card or four, so much for making it cheaper and more acessable.

The other solution would be exactly the same situation, but then forcing even more to buy four. This is how it works when you buy ONE videocard:

- Current system: compete in single-card category, so you have to buy a very expensive dual-gpu card. You participate in 50% of the GPU global categories.
- Suggested system: compete in single-gpu category, so you can choose between a couple of less expensive single-gpu cards. More expensive will be better for higher scores, but less expensive doesn't mean no points instantly. The gap between top-dollar and middle-dollar is less big. You participate in 50% of the GPU global categories.
- Multi-category system: compete in single-gpu category, so you can choose between a couple of less expensive single-gpu cards. More expensive will be better for higher scores, but less expensive doesn't mean no points instantly. The gap between top-dollar and middle-dollar is less big. You compete in 25% of the global categories.

In other words, if you want to increase your ranking in the Overclockers League (or actually: any league) you will have to invest in a second multi-gpu solution as well. However, if you have a multi-category system (dual, triple, quadriple) you need a solution for each category. This would be even MORE expensive than the one suggested in the newspost.

bazx said: is there a technical reason why we cannot have multiple GPU ranks?

No, we can rank results based on number of sockets, number of cores and number of cards.

ether.real said: Splitting up the Single/Dual GPUs is all brass on the titanic until the CPU problem gets fixed.

There's very little HWBot can do about the CPU problem. This is a characteristic of the benchmarks used; Vantage Xtreme is one we could add, but it won't change the cpu problem in all other benchmarks.

Changing to a single GPU system does in fact solve the cpu problem a bit for at least half of the 3D rankings.

ether.real said: How is this any different than someone else coming with a 5870 that crushes you at stock speeds?

It's already a difference in money. Also, it's a new generation; not really a fair comparison. We can't ban innovation.

drnip said: I say why have a ceiling? You don't have one on global points. That might be a little bit of influence on people breaking back out all the old hardware they have which in return equals more participation.

We have a cap because the Overclockers League was designed to rank overclockers on extreme overclocking skill. This means:

- Extreme cooling (LN2)
- Extreme hardware (High-end systems)
- Extreme results (Best scores in the benchmark)

Without a cap, this league would be dominated by people who just bench very much and although I'm not at all saying that this doesn't require skill, it's not what the Overclockers League is designed for. Note that the amount of available hardware points increase spectacularly as hardware categories increase.

Comment from ToxicWk (Hellas Overclocking Team) - link
Time: Oct 30, 2009

As you said it is going to give really nice boost to newbie ocers with no dual GPU cards.Also it will give them a chance to compete with more weathly guys who buy a 295 just to test it .I think we must differ them by core:D

Comment from Polpi_91 (HwProject.net Italy OC Team) - link
Time: Oct 30, 2009

i'm agree this opinion

Comment from Thor941 (JMax-Hardware) - link
Time: Oct 30, 2009

I think it won't change the fact that "big" clockers with huge HW will still be in the top rankings.

Comment from richba5tard (hwbot.org benching team) - link
Time: Oct 30, 2009

You make it sound like that's a bad thing. :)

Comment from ToxicWk (Hellas Overclocking Team) - link
Time: Oct 30, 2009

@thor941:I never tried to compete Hipro ,Andre or Vince but some others that don't beat me by their skillz , crazy OCes or tweaks but some 295 and 4870x2 being stock only because I play with a decent 4850.Always talking for global ranking

Comment from Thor941 (JMax-Hardware) - link
Time: Oct 30, 2009

richba5tard said: You make it sound like that's a bad thing. :)

Not at all ;)

That's just the way it is (to me)

I think that our o/c "world" is like that, unless will have to ban all the sponsored people as well, but it is just impossible :)

I like the way points are calculated now, and the new way won't change things that much in my opinion so I'd say that it sounds good to me as well.

Comment from K404 - link
Time: Oct 30, 2009

This isnt my final answer but...

This idea makes slightly older, cheap cards more viable for global points, which while it isnt a bad thing... it will massively increase the competition in these categories..... LN2 will be necessary for hardware points for the cards, which will push more people out.

Benching wont really become any cheaper- LN2 or no points. LN2 needs specific lifestyle choices to already be in place before it can become a viable option.

At least with the dual-core cards, the bigger names can fight over those for the top spots and the less-sponsored guys with less money and less cooling can still do well in hardware categories

I am aware this answer is selfish... I dont think its my final answer

Comment from canna1988 (PcBrain OC Team) - link
Time: Oct 30, 2009

Hello,

I agree with this new possible rules, because is not good that some people with a GTX295 with liquid cooling got better result than a GTX285 maybe under LN2... so for me is possible change

Also my team think same about this.

Sorry for my english.

Alessio

Comment from Linuxfan (XtremeSystems) - link
Time: Oct 30, 2009

Sounds like it might be a good change.
One question though, how would you rank the dual card GPUs? Will there be three catagories with single card, single card dual GPU, and SLI/CF cards? I don't know how that would work out, but if we just changed the 4870X2 and GTX 295 to the dual cards ranking, they wouldn't be as good as for example 2x 4870X2 or 2x GTX 295, so there wouldn't be much of a point to bench them, right?
I get the idea though, single GPU competition for global boints sounds good! :)

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

I disagree with the change. Keep the 4870X2 in its own category as a single card. What is dominating the hwboints is not the dual-GPU cards vs. single GPU, it is that GPU intensive tests are almost solely determined by CPU clocks. It is less disturbing to me that a stock dual GPU card gets more boints than an extreme OC'd single GPU card. It is more disturbing to me that an extreme OC'd CPU with a stock GPU will get more boints than a mildly OC'd CPU with an extreme OC'd GPU.

Instead of the proposed change, I would much rather see a push to using Vantage X instead of Vantage P in hwbot league. Remove the CPU bottleneck and let GPU power be the main determinating factor in getting boints in the GPU tests.

Comment from [R&B]Beleno (HwProject.net Italy OC Team) - link
Time: Oct 30, 2009

I think is a good idea, more competition and more fun

Comment from ReverendMaynard (OC Alliance) - link
Time: Oct 30, 2009

ether.real said: I disagree with the change. Keep the 4870X2 in its own category as a single card. What is dominating the hwboints is not the dual-GPU cards vs. single GPU, it is that GPU intensive tests are almost solely determined by CPU clocks. It is less disturbing to me that a stock dual GPU card gets more boints than an extreme OC'd single GPU card. It is more disturbing to me that an extreme OC'd CPU with a stock GPU will get more boints than a mildly OC'd CPU with an extreme OC'd GPU.

Instead of the proposed change, I would much rather see a push to using Vantage X instead of Vantage P in hwbot league. Remove the CPU bottleneck and let GPU power be the main determinating factor in getting boints in the GPU tests.

I agree 100%.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Oct 30, 2009

K404 said: This idea makes slightly older, cheap cards more viable for global points, which while it isnt a bad thing... it will massively increase the competition in these categories..... LN2 will be necessary for hardware points for the cards, which will push more people out ...

We are actually trying to improve the hardware points as well. This does take quite a bit of work in balancing, though.

Linuxfan2000 said: Sounds like it might be a good change.
One question though, how would you rank the dual card GPUs? Will there be three catagories with single card, single card dual GPU, and SLI/CF cards? I don't know how that would work out, but if we just changed the 4870X2 and GTX 295 to the dual cards ranking, they wouldn't be as good as for example 2x 4870X2 or 2x GTX 295, so there wouldn't be much of a point to bench them, right?
I get the idea though, single GPU competition for global boints sounds good! :)

The initial idea would be to have two rankings:

1) Single GPU
2) Multi GPU

Multi-GPU would then indeed be dominated by those who can afford multiple cards. However, it seems that in the current generation of hardware even 3D benchmarks are CPU-limited. 4 cards won't be more powerful than 3 per definition.

It's quite difficult to have a solution that works out perfect for all situations. Creating more global categories also means that people have to buy more hardware to compete at the top. Also, it puts more stress on our server.

ether.real said: Instead of the proposed change, I would much rather see a push to using Vantage X instead of Vantage P in hwbot league. Remove the CPU bottleneck and let GPU power be the main determinating factor in getting boints in the GPU tests.

The fact that dual GPUs are now in the 'single'-category is part of the problem of CPU-bound 3D tests. The more powerful you make the 3D part of your system (VGA), the more powerful you need to make your 2D part (CPU) to see scaling from adding VGA OC.

Comment from K404 - link
Time: Oct 30, 2009

Massman said: We are actually trying to improve the hardware points as well. This does take quite a bit of work in balancing, though.

What ya got in mind? :)

random idea... not thought out and possibly inappropriate for this thread...

what if a score was removed from a hardware category if it had global points?

Comment from Mafio (HwProject.net Italy OC Team) - link
Time: Oct 30, 2009

i agree, this is a good idea

Comment from aperacer (AwardFabrik) - link
Time: Oct 30, 2009

I think you should look on the numbers of GPUs on a card and not on single or dual card

Comment from giukey79 (HwProject.net Italy OC Team) - link
Time: Oct 30, 2009

I agree with the initiative [img] http://forum.hwproject.net/images/smilies/sisi.gif[/img]

Comment from 7ornado - Antani OC (HwProject.net Italy OC Team) - link
Time: Oct 30, 2009

i agree with the possible change. in effect, gtx295 and 4870x2 are almost like a sli/crossfire of other less power vga...so they shouldn't count as single vga
just imo

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

So, when do we classify a Yorkfield as a "multiple" Wolfdale? Cause that makes just as much sense.

What about dual GPU cards that don't have a single GPU equivalent? How would those get classified?

Comment from Thor941 (JMax-Hardware) - link
Time: Oct 30, 2009

7ornado said: i agree with the possible change. in effect, gtx295 and 4870x2 are almost like a sli/crossfire of other less power vga...so they shouldn't count as single vga
just imo

I think it is more difficult to clock a CF of 4870 than a 4870x2 for example, so they should not be in the same category ...

Comment from Splave (OC Alliance) - link
Time: Oct 30, 2009

Im against the change.
-I already think to be on top you need ln2 and the top dogs dont care if its dual or single they will still do them both

-Its always going to be the big money guys on top. and I dont think money should be a driving point as to make the change. In all reality as a hobby computers are pretty cheap compared to other things

-If you did this, how about a global points change for cpu's? single core, dual core, quad etc getting their own global categories?

-Please dont take all my gtx-295 points rofl ;)

Comment from Delex - GiaXi (Hardware Station OC Team Italy) - link
Time: Oct 30, 2009

hi,
that's a really good news!! I think the same like you and other memebers of this great community! At the moment every dual GPU card (stock cooled, too) beats a single card Ln2 cooled, so I mean that the way to go is to group single gpu as single card and dual/multiGPU as multi gpu.

+1 for HWBot!!

Have a nice day!

Alex

Comment from El Gappo (Bench Tec UK) - link
Time: Oct 30, 2009

Splave said:
-If you did this, how about a global points change for cpu's? single core, dual core, quad etc getting their own global categories?

not sure if your serious about that but it is definitely a great idea imo.AMD users would actually get some global points :D but it does throw up the issue of 6 and 12 core categories along with god knows how many in the future.

I'm kind of split on this issue. I never thought the dual gpu cards belonged in that category and the competition it will bring will be great although the high cost of the cooling needed defeats what they are trying to achieve so I'm going to keep my trap shut and not vote at all.

Comment from r1ch (Bench Tec UK) - link
Time: Oct 30, 2009

Splave said: Im against the change.

-Please dont take all my gtx-295 points rofl ;)

Please think about the subject, and what's best for the community - not what you stand to gain/lose from it.

Please don't drag CPU's into the discussion either - this is about GPUs.

Please read ALL of massman's news post and make a considered decision.

I'm for this change. :)

Comment from PUNX223 (OC Alliance) - link
Time: Oct 30, 2009

i like this idea as it can give the guys who cannot afford the higher end systems can still be well ranked for the skill they have with the hardware they have.

I understand taht a guy with LN2 with the same platform will still wipe the floor with a guy on the same platform on air. But if its split possibly even down to different platforms it could help with giving people without the sponsorships or big budgets to compete

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

r1ch said:
Please don't drag CPU's into the discussion either - this is about GPUs.

There should be consistency between the CPU rankings and the GPU rankings.

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

PUNX223 said: i like this idea as it can give the guys who cannot afford the higher end systems can still be well ranked for the skill they have with the hardware they have.

I understand taht a guy with LN2 with the same platform will still wipe the floor with a guy on the same platform on air. But if its split possibly even down to different platforms it could help with giving people without the sponsorships or big budgets to compete

All its going to change is the sponsored LN2 guys with deep pockets are going to dominate both the single *and* dual GPU rankings.

Comment from RawZ (Bench Tec UK) - link
Time: Oct 30, 2009

Do you know, last night I was actually browsing through the scores and seeing what my next choice of CPU's would be and where I could gain more points as my budget is rather limited. Then it suddenly dawned on me that, compared to other benchers.. i have little funds to afford top hardware or better cooling to gain quite a bit more points for me and the team. It's at that point I said to myself.. if only I had the money.. I could be considered a great bencher racking in tonnes of points for the team. There are differences from regular benchers with lots of points and little points, but the main I feel is money. To have lots of points (other than skill and experience), you need the dosh to compete at the top level which kind of defeats the point in spending god knows how many hours each and every day benching if all you can get is 1 point if that. I know that if I had extra cash lying around, I would have a tonne more points under my name and for the team.

However, that said, if I did have a lot of points and money right now, I'd be saying the opposite and pretty much say, tough **** that's life. Any where you want to make it, whatever in life, you need money and sadly in relation to benching, it ain't cheap. If you can afford to, then great, congrats on the many points ahead of you. If you can't afford the best hardware, then I'm afraid your limited.

Comment from DrNip (OC Alliance) - link
Time: Oct 30, 2009

+1 on the change.

Comment from TiN (XtremeLabs.org) - link
Time: Oct 30, 2009

+1 for change, and to go more : there is the same ranks with CPU sockets, not cores :)

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

RawZ said: Do you know, last night I was actually browsing through the scores and seeing what my next choice of CPU's would be and where I could gain more points as my budget is rather limited. Then it suddenly dawned on me that, compared to other benchers.. i have little funds to afford top hardware or better cooling to gain quite a bit more points for me and the team. It's at that point I said to myself.. if only I had the money.. I could be considered a great bencher racking in tonnes of points for the team. There are differences from regular benchers with lots of points and little points, but the main I feel is money. To have lots of points (other than skill and experience), you need the dosh to compete at the top level which kind of defeats the point in spending god knows how many hours each and every day benching if all you can get is 1 point if that. I know that if I had extra cash lying around, I would have a tonne more points under my name and for the team.

However, that said, if I did have a lot of points and money right now, I'd be saying the opposite and pretty much say, tough **** that's life. Any where you want to make it, whatever in life, you need money and sadly in relation to benching, it ain't cheap. If you can afford to, then great, congrats on the many points ahead of you. If you can't afford the best hardware, then I'm afraid your limited.

All sports and competitive leagues are "Pay to Play". You are not going to be able to equalize the dollar influence unless you yourself are willing to buy top end hardware for everyone. The Global vs. Hardware categories does an excellent job of bringing boint opportunities to everyone, its a great system. But to get hundreds of boints and high global ranking is going to take time, money, and talent just like everything else.

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

TiN said: +1 for change, and to go more : there is the same ranks with CPU sockets, not cores :)

The current CPU layout is exactly like the current GPU layout. Sockets = Slots. MCM CPUs are counted as single CPUs just like X2s are counted like single GPUs.

Comment from DrNip (OC Alliance) - link
Time: Oct 30, 2009

Sometimes change is good and sometimes its not. If it doesn't work then go back to what you know.

Comment from Kal-EL (OC Alliance) - link
Time: Oct 30, 2009

A few quick random thought from work:

What would happen to single pcb "Dual GPU" submissions?

I think they'd become extinct from hwbot, IF, tri-card and quad-card setups trump a single "dual gpu" card at the upper range.

In essence you'd kill off any sort of single "Dual GPU" submission.

The money gap will always be there.

Faster cards cost more
slower cards cost less

This idea might shorten that margin but it doesn't resolve it.

I think the biggest thing is PARTICIPATION. Seeing boints awarded for any benchmark is like crack. One hit and you're an addict. A few more and you're a junky. Don't we all want everyone to be junkies?

To take a page from the scumbag, dirtbag, lowest beings on the planet, horrid, miserable, soul less, street drug dealers........... give people a hit, get them hooked, fuel the fever.

The only way to do this is by finding a way to bring boints to those with even mediocre scores that are above the average of a stock run system.

How? I'm not sure. Maybe creating a base line score average for each generation leap. Taking base line scores from the top tear cards from x,y and z vendors/manufacturers , run them at stock clocks and average them scores together, then award boints upward from there?

How do you create an addict? Give them something to be addicted to. Eventually they'll get better, aquire more knowledge, find inventive "Home Brew" solutions and graduate to the upper levels. Meanwhile they enjoy competition from the get go.

Comment from Lt.JD (Overclock.net) - link
Time: Oct 30, 2009

ether.real said: All sports and competitive leagues are "Pay to Play". You are not going to be able to equalize the dollar influence unless you yourself are willing to buy top end hardware for everyone. The Global vs. Hardware categories does an excellent job of bringing boint opportunities to everyone, its a great system. But to get hundreds of boints and high global ranking is going to take time, money, and talent just like everything else.

I disagree... isn't OCing a "sport" and in the sports I play it doesn't matter if your family makes six figures or four the best players are on the field... prepared for the flame.

Comment from RawZ (Bench Tec UK) - link
Time: Oct 30, 2009

ether.real said: All sports and competitive leagues are "Pay to Play". You are not going to be able to equalize the dollar influence unless you yourself are willing to buy top end hardware for everyone. The Global vs. Hardware categories does an excellent job of bringing boint opportunities to everyone, its a great system. But to get hundreds of boints and high global ranking is going to take time, money, and talent just like everything else.

I'm not knocking the system, it's a great system. I'm just pointing out that, like anywhere, you need the money to progress further.

drnip said: Sometimes change is good and sometimes its not. If it doesn't work then go back to what you know.

Sound like an idea. Trial basis perhaps.

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

Lt.JD said: I disagree... isn't OCing a "sport" and in the sports I play it doesn't matter if your family makes six figures or four the best players are on the field... prepared for the flame.

Guess who the best players tend to be? The ones that can afford the nicer equipment, the personalized trainers, the customized diets and workout regimens.

Comment from Thor941 (JMax-Hardware) - link
Time: Oct 30, 2009

Isn't it possible to make a simulation on the top 100 scorers or I do'nt know how many ...

Comment from Kal-EL (OC Alliance) - link
Time: Oct 30, 2009

The code work required for running sims on each scenario seems a bit too much to be considered.

I think this is why the brainstorm must occur before dedicating those kind of resources to a sim.

Comment from rbuass (Team Adrenaline) - link
Time: Oct 30, 2009

My opinion is that errors have much more serious than that of GPUs.
Many people make points with notebook hardware and equipment that do not need or overclocking.
Many people do items of equipment that no competitors (the person struggles to earn points overclocked a CPU in LN2 and for example, takes place 5 or 6 and get the same score of a person who overclock a laptop that only has it) .
What should happen is that the points of hardware should be studied so that they have value.
For example, overclock a GTX 275 and the person to reach 1200 Mhz and it gets points even if you have your 300 points of hardware.
I believe it is better to hit the points of hardware to modify the overall points (which in my opinion should separate the number of PCI ex used).
Sorry the broken english...i hope you understand me

Comment from Lt.JD (Overclock.net) - link
Time: Oct 30, 2009

ether.real said: Guess who the best players tend to be? The ones that can afford the nicer equipment, the personalized trainers, the customized diets and workout regimens.

I wont lie... yes it happens to be that way most of the time but I know people who dont have those amenities and still play

Comment from komadyret (411 Overkill) - link
Time: Oct 30, 2009

The change is rather welcome. IMHO the categories should have been divided in GPUs from the start, not "cards".
The 295 wasn't even that, it was two cards, crammed into one pci-e slot. And it is really just a question of time until some manufacturer tries the trick again and slaps two of the future equivalent of a single-pcb 295 or x2 against each other and there is a 4-core "single-card" on a PCI-e3.0 slot. Goodbye single-card benching.

I have tried to think this through over a cup of coffee, and I think there are two (preferably 3) ways to try fixing this issue:

1) Divide on gpu, not card
1.1) (preferably) set up categories as "single", "dual" and "multi" gpu, with global points for all 3, and maybe some "universal" points for the top scores independent of gpu# to make the sponsored and wealthy clockers happy

2) Drop gpu classes all together. Award global 3D points as one (like it is for CPUs now, # of cpus or cores aren't divided), and increase the points awarded for hardware in the existing single and multi (gpu, not card, as I still think they should be divided) classes.
This should also be followed by a increase in the hardwareboints ceiling from 300 to lets say 500 points? Or set in place a limit on gpu separate from cpu hardware points?
This would equal the playing field for CPU vs GPU points in the totals, and maybe down the road a similar dividing can be done for CPUs, splitting the single-socket from the dual/multi socket for hardware points.

There are of course problems to consider in all this. Making a third gpu-global-class will seriously shift the weight of the points in the gpu-direction. I'm not sure it's a good thing to make the cpu's so much less important in the big points...

Lots to think about, to ensure the system works as intended. I'm confident the man with the understanding of the algorithm will come to the correct decision, not based on my wishes and fantasies.

Sorry for the long post. I just got a little carried away over my coffee.

I'm sure of one thing tho; I support the initiative to get the dual-gpu cards WAY OUT of the single card (gpu) class. I find it "disturbing" that they ended up there in the first place :)

Comment from bazx (Bench Tec UK) - link
Time: Oct 30, 2009

my views on this should be well known to anyone that uses these forums

I am in favor of a single GPU ranking system

reading the news post it was not clear to me if the x2 would get there own ranking with the same global points system as the proposed single GPU system

so would there be a:

Single GPU rank

and a

x2 rank

both with global points

Comment from Dualist (Bench Tec UK) - link
Time: Oct 30, 2009

komadyret said: set up categories as "single", "dual" and "multi" gpu, with global points for all 3, and maybe some "universal" points for the top scores independent of gpu# to make the sponsored and wealthy clockers happy

I'm all for this, we need three catagories not two.! Otherwise manufactures will not produce dual gpus for games (some of us like to game as well as bench).

As you are surjesting you either have to buy one card or four, so much for making it cheaper and more acessable.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Oct 30, 2009

K404 said: What ya got in mind? :)

Little tuning of the Sigmoids: increase the upper popularity limits. Something like that

K404 said: what if a score was removed from a hardware category if it had global points?

You probably mean 'person'. Otherwise one could just submit 50002 and 50001, the first taking global, the latter taking hardware.

ether.real said: So, when do we classify a Yorkfield as a "multiple" Wolfdale? Cause that makes just as much sense.

There are many differences involved in this, the most apparent possibly the fact that there are only two benchmarks that really scale with multi-core design being Wprime 32M and Wprime 1024M whereas nearly every benchmark scales with multi-GPU design. We may have been postponing splitting up the socket for Wprime a bit too much, though.

Another very big difference is that within the GPU there are in fact multiple processors. We don't split them up in 192, 216 or 512 categories, though.

ether.real said: What about dual GPU cards that don't have a single GPU equivalent? How would those get classified?

Initial suggestion described it as two seperate categories:

-Single GPU
-Multi GPU (dual, triple, quadriple)

In the latter in other words.

ether.real said: There should be consistency between the CPU rankings and the GPU rankings.

In my opinion consistency should never be the main reason to change or not change anything. I can give you as much examples of why the current rankings are not consistent as you can give me why they are probably.

TiN said: +1 for change, and to go more : there is the same ranks with CPU sockets, not cores :)

As for Wprime: noted on the agenda.

ether.real said: All sports and competitive leagues are "Pay to Play". You are not going to be able to equalize the dollar influence unless you yourself are willing to buy top end hardware for everyone.

I agree with you. However, I also believe that if there's an easy way to reduce the dollar influence we should at least consider it. It may not change that much for the so-called top dogs, but if it changes something for the less fortunate enthousiasts I'm already happy.

Thor941 said: Isn't it possible to make a simulation on the top 100 scorers or I do'nt know how many ...

Next meeting is on monday and it's on the agenda.

komadyret said: This should also be followed by a increase in the hardwareboints ceiling from 300 to lets say 500 points?

Good suggestion. We have already discussed this possibility in the staff section as well and it came to our attention that increasing the hardware points is in fact a very short-term solution as:

- Hardware categories are added by the day
- For one hardware category, there are 7 (CPU) or 6/12 (GPU) new rankings
- For each newly added benchmark there are countless new hardware points available

Whether or not we increase the hardware points cap for the individual league, there will always be a point where the cap forms an issue. Raising the cap will work just fine for a while, but at a certain point we will return to the same situation as we are in today. Raising it again will work fine again ánd bring us back to the same situation. Eventually, we will have a hardware cap that is higher than the amount of points you can win in global categories. The problem is:

The Overclockers League is supposed to be filled with those who show extreme skill using extreme cooling on extreme hardware. The top results, in other words.

This is the main reason why we set the hardware cap in the first place. If we increase the cap so much that it doesn't have it effect anymore, why bother having it in the first place.

komadyret said: Or set in place a limit on gpu separate from cpu hardware points?

That is a very interesting idea!

komadyret said: Making a third gpu-global-class will seriously shift the weight of the points in the gpu-direction. I'm not sure it's a good thing to make the cpu's so much less important in the big points ...

Again, excellent remark. Creating single, dual and more GPU categories is in itself no problem, but it will most definitly shift the weight of the points to GPU benchmarking. In particular: multi-gpu benchmarking. In other words, instead of making it a more balanced game, we would be making it a more expensive game.

bazx said: Single GPU rank and a x2 rank both with global points

Single GPU and Multi GPU with X2 going into the Multi GPU

Dualist said: As you are surjesting you either have to buy one card or four, so much for making it cheaper and more acessable.

The other solution would be exactly the same situation, but then forcing even more to buy four. This is how it works when you buy ONE videocard:

- Current system: compete in single-card category, so you have to buy a very expensive dual-gpu card. You participate in 50% of the GPU global categories.
- Suggested system: compete in single-gpu category, so you can choose between a couple of less expensive single-gpu cards. More expensive will be better for higher scores, but less expensive doesn't mean no points instantly. The gap between top-dollar and middle-dollar is less big. You participate in 50% of the GPU global categories.
- Multi-category system: compete in single-gpu category, so you can choose between a couple of less expensive single-gpu cards. More expensive will be better for higher scores, but less expensive doesn't mean no points instantly. The gap between top-dollar and middle-dollar is less big. You compete in 25% of the global categories.

In other words, if you want to increase your ranking in the Overclockers League (or actually: any league) you will have to invest in a second multi-gpu solution as well. However, if you have a multi-category system (dual, triple, quadriple) you need a solution for each category. This would be even MORE expensive than the one suggested in the newspost.

Comment from rbuass (Team Adrenaline) - link
Time: Oct 30, 2009

rbuass said: My opinion is that errors have much more serious than that of GPUs.
Many people make points with notebook hardware and equipment that do not need or overclocking.
Many people do items of equipment that no competitors (the person struggles to earn points overclocked a CPU in LN2 and for example, takes place 5 or 6 and get the same score of a person who overclock a laptop that only has it) .
What should happen is that the points of hardware should be studied so that they have value.
For example, overclock a GTX 275 and the person to reach 1200 Mhz and it gets points even if you have your 300 points of hardware.
I believe it is better to hit the points of hardware to modify the overall points (which in my opinion should separate the number of PCI ex used).
Sorry the broken english...i hope you understand me

And one more thing i think.
Some people gain a lot of points without do it the art of overclocking; for exemple a power server runs 16 cores (or more) easy in stock but broken power overclockers that used LN2 at a Core i7 at more than 5200 Mhz (for example).
What the user that haves a server did?
Only runs the Wprime at stock frequency, and easy broke the overclocker that fight to reach run at 5200 Mhz in LN2 (and the risk, and the work, and the spended time)...and he gain a lot of points easy.
I think there are many things to change before change GPU method to allow the points.
Its only my opinion.
;)

Comment from r1ch (Bench Tec UK) - link
Time: Oct 30, 2009

rbuass said: *snip*

Fair enough point of view, but not the right place to talk about it. Please concentrate on the GPU situation in this thread. :)

ether.real, massman just justified the CPU comparison 10x better than I could. Your "no borders oc team" are the only ones in 5 pages to disagree at this point.

Comment from bazx (Bench Tec UK) - link
Time: Oct 30, 2009

why does it have to be a choice?

unless there are multiple rankings that take into consideration all levels and methods of GPU benching you will not get consensus here

i am all for Single GPU ranking but not at the expense of other benchers who have worked hard to obtain there points

if we had separated the cards out a year and a half ago i would have been happy but the ship has sailed on that as so many have now got points from the x2 cards

is there a technical reason why we cannot have multiple GPU ranks?

Comment from K404 - link
Time: Oct 30, 2009

The revised approach to hardware points.. whenever the idea is developed and ready to be shared with community... will make or break it for me :)

The X2 cards (and the level of competiton in the categories for single card global) take a lot of the flak for other (mid-high-mid range) GPU categories which gives a lot of guys breathing space to do well with water or dice (ok... less available to everyone but still cheap) I know everyone understands this and I dont need to re-phrase it but I dont want to see people having to use really old crap (or coldbugged) hardware to get hardware points :)

LN2 should never be the minimum level needed to compete for any worthwhile points. I loved seeing Stummer take top 3 in 3DM03 with a watercooled 4870X2. :)

Controversial....
Cost of getting to the top....show skill, develop LN2 ability, get a sponsor.. then the cost of X2 cards doesnt matter, coz they're free.

The phychology of it... who wants top scores but doesnt want to be picked up by a company if the opportunity was there?

Again... might be a topic for another thread. Those last 2 sentences arent meant to offend

Comment from rbuass (Team Adrenaline) - link
Time: Oct 30, 2009

r1ch said: Fair enough point of view, but not the right place to talk about it. Please concentrate on the GPU situation in this thread. :)

ether.real, massman just justified the CPU comparison 10x better than I could. Your "no borders oc team" are the only ones in 5 pages to disagree at this point.

No friend...i posted it cause i think there are many things to change before change GPU method.
Cause i think is not "fair play" to gain lot of points without overclocking...
And...i intend we have to keep method and fix this kind of "not fair play" before.
;)

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 30, 2009

r1ch said: Fair enough point of view, but not the right place to talk about it. Please concentrate on the GPU situation in this thread. :)

ether.real, massman just justified the CPU comparison 10x better than I could. Your "no borders oc team" are the only ones in 5 pages to disagree at this point.

Ya, we chatted about this a bit, had some disagreement, but for the most part agree on the same thing now. I still think you guys are chasing the wrong problem. The absolute worst thing about the GPU rankings is that CPU clocks matter too much. Splitting up the Single/Dual GPUs is all brass on the titanic until the CPU problem gets fixed.

Comment from knopflerbruce (PURE) - link
Time: Oct 30, 2009

The hardware boints need some fixes, too. Limit at500 instead of 300 would be nice. Most of the top global OC'ers have alot of globals anyway, only two top 20 OC'ers have a little bit below 300 boints. IMO it looks like they get their 300 HWBoints, and that's it. If the limit was 500, I'm sure most of the would just get 500.

Prehaps it's a good idea to announce it like 2 months before the limit is raised, so the global folks get a chance to bench more gear and get close to the new limit?

Comment from psynapse (411 Overkill) - link
Time: Oct 30, 2009

I'm swayed by the arguments both ways.

I'm leaning in favour of the proposed change, however.

Comment from by_xfile (HwProject.net Italy OC Team) - link
Time: Oct 30, 2009

I think it's a good idea, will be more competition and more fun

Comment from rbuass (Team Adrenaline) - link
Time: Oct 30, 2009

"Prehaps it's a good idea to announce it like 2 months before the limit is raised, so the global folks get a chance to bench more gear and get close to the new limit"

100% ok
to me....its sure

But i don t allow that people can gain more than 10 points easy without overclocking or competition with some hardware (notebooks, hardware thats no competitor).
Is more than a guy can gain with a Phenom II 955 at 6250 Mhz (for exemple) running benchs...its not fair play to me
See above:

No points
CPU-Z - 6244.29 mhz - rbuass (Team Adrenaline) - (Phenom II X4 955 BE @ 6244mhz on LN2) 0 points

2 points
CPU-Z - 1600.25 mhz - rbuass (Team Adrenaline) - (Athlon TF-20 @ 1600mhz ) 2 points

I can t understand i gained 2 points without overclock and with the phenom at more than 6240 mhz LN2 cooled i have zero points....lol

Comment from K404 - link
Time: Oct 30, 2009

Sometimes Iv wondered if a category should need more than..say.. 20 results from different users in.. say...3 different teams to be awarded points? :)

About unlocked CPUs... GPUs.... what about a rule that for HWB category benching... the component must have the stock hardware capabilities of that chip. y'know... 800SP/ 3 cores/ 256KB cache etc etc

Im not sure how to open things up so the unlocked GPUs/ CPUs can shine give benefit to the user

Comment from Iron (HwProject.net Italy OC Team) - link
Time: Oct 31, 2009

I always hated multi gpu cards or multiple cards in sli or xfire... but that's a personal idea.
It's not a personal idea the fact that if I put under ln2 a 4890, here it comes a guy with a 4870x2 that crushes me with everything at stock speeds.
That's completely wrong, for me.
So, the fact is: how many categories should we put in the game, speaking about multi-gpus?
I think that single gpu, dual gpu and multiple gpu will do the job: in fact, if you can afford to buy 3 5870, you have even the $$ to buy another one... speaking about mainstream vgas it's the same.
Speaking about benchmark scores, as you said before, there's not so much differences between triple gpu configurations and quadriple gpu configurations.

In other words:
- Single GPU category.
- Dual GPU category.
- Multiple GPU category. ;)

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 31, 2009

|ron said: I always hated multi gpu cards or multiple cards in sli or xfire... but that's a personal idea.
It's not a personal idea the fact that if I put under ln2 a 4890, here it comes a guy with a 4870x2 that crushes me with everything at stock speeds.
That's completely wrong, for me.
So, the fact is: how many categories should we put in the game, speaking about multi-gpus?

How is this any different than someone else coming with a 5870 that crushes you at stock speeds?

Comment from Iron (HwProject.net Italy OC Team) - link
Time: Oct 31, 2009

It's the new generation vs the old one, so it's normal.
Your example is like complaining if someone with a tuned Ferrari from the 80's is beaten by someone with a new Ferrari Enzo... there's no match, but it's obvious and normal.
Our case is about driving a race with a 4cylinders car vs a V12... you can put efforts in tuning the poor 4cil engine but if someone will slightly tune the V12... good bye! ;)

Comment from CaNNon (411 Overkill) - link
Time: Oct 31, 2009

This is some long needed change so I'm +1 here.
Dividing gpu's up so it bus vs bus and then number of cores vs number of cores is the only fair solution.

I think his would be a improvement in the data base too giving more accurate searched info.
Like if searching for a fast single gpu card you really want to find just that.

Also this would show speed and performance differences between single pcb solutions and using full multi card setups much better. My point here is Hwbots future, things should in away the data can useful to anyone who trips over the data base. This will end up getting new users and maybe even some free PR.
As a example I could see the data of single pcb with dual gpu vs dual cards using the same gpu being of use to reviewers or review sites.

[rant]
If a guy has earned points overclocking let him show those points with no limits! I mean I don't see any reason one of the leagues should be valued higher that the other.
[/rant]

Comment from TinTin@OCQQ (PURE) - link
Time: Oct 31, 2009

Now we can see low HWpoints on single card single GPU category because of low ranking in Global.
I would like the Dual GPU card to be other category. Then we will see more LN2 on single card. :D

Comment from BradleyKZN (Team South Africa) - link
Time: Oct 31, 2009

Im all for it. I was actually thinking about this yesterday, and i think this would be a great move.

Comment from VasGTO (PURE) - link
Time: Oct 31, 2009

Well it's at least an improvement... I hope that someday when you have 10,000+ active users you will have a much more complex algorithm... To truly have a fair comparison among GPU's you'll need to use the part of the HWBot competition engine and limit the cpu speed (just like most gpu competitions). I can see right now it's probably not feasible, but when this site (or another) has millions of submissions then you can have global pionts for gpu's with cpu limits of 1.0ghz, 2.0ghz, 3.0ghz, 4.0ghz, 5.0ghz ... etc. That combined with single gpu, dual gpu, multi gpu, and then further combined with single cpu, multi cpu.... Oh my, I'm getting a headache just thinking about it :)

Drunk pirate brainstorming again ;) .... btw, what's a construction bot vs. kitchen bot vs. a maintenance bot ?

Comment from Splave (OC Alliance) - link
Time: Oct 31, 2009

What this would do is make ln2 a must for any type of global points which would do the opposite of giving lower end hardware users a chance.

A 5870 on ln2 would cost 3x as much as a 295 on air. So dont make the change for the money aspect. Unlocking hw points to 500 would be a great idea and would allow people to slum up more old hardware and compete if they dont have major funds or get free hardware.

r1ch said: Fair enough point of view, but not the right place to talk about it. Please concentrate on the GPU situation in this thread. :)

ether.real, massman just justified the CPU comparison 10x better than I could. Your "no borders oc team" are the only ones in 5 pages to disagree at this point.

Who the hell are you buddy? We all get our opinions here. CPU talk is more relevant than you know. If you did this with gpu's how could you not follow suit with cpu's. This would allow people with low end hardware like a e8400 compete in wprime for example. Isnt that what you want? Or should there be a double standard.

Im reading alot of posts that just say. I agree, sounds good, good idea. Why dont you post why you think it is a good idea so there can be discussion ;)

Comment from Eeky NoX (Union Francophone Hwbot) - link
Time: Oct 31, 2009

The way to go ;) I follow that idea with enthousiasm ^^
...more fairly competition and also more equality between the Leagues, will make the game increasing frankly his popularity imo

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Oct 31, 2009

Sometimes Iv wondered if a category should need more than..say.. 20 results from different users in.. say...3 different teams to be awarded points? :)

This is crap. If I have something that nobody else has benched, I expect to be rewarded for it.

Comment from knopflerbruce (PURE) - link
Time: Oct 31, 2009

Mr.Scott said: This is crap. If I have something that nobody else has benched, I expect to be rewarded for it.

+1

If not, then nobody will bench non-mainstream HW. Plus, the idea is to reward skill too - and you can show skill even if there is no competition, so the 2 points you gt are well deserved.:)

Comment from BenchBrothers.de (AwardFabrik) - link
Time: Oct 31, 2009

pro changing !!

it's great that this discusion is up again,
i found that decision unfavorable since it was made.

These cards performance comes by their number of GPUs, not by the number of the used PCIE Slots - as it is ranked now.
so why all the other (real) singleGPU cards have to compete against that unequal opponents?

HWBot is about competing - thats the reason why till today all the Gulftown results are not ranked and thats the reason why multi GPU Cards should be rated like native SLI or CF Setups.

It would also prevent that in some 3DMarks (like currently 03 & 05) the "singleCard" WR gets more points than the "multiCard" (and overall) WR.

since a long time I had the feeling that multiGPU Cards are prefered by most of the people to gain "easier" points (as it is possible to say that) than to spend the same time e.g. for a gtx280.
great to see that my feeling is confirmed by the statistics.

Yes we can change that rule - pro changing!
Multi GPUs need to be compared to Multi GPUs - everything else is not a real competition - thats not a secret watching the cards performance!

Comment from BenchBros (HardwareLuxx Germany) - link
Time: Oct 31, 2009

pro changing also!

(if my Teammates agree :) )

Comment from DonBanana (HardwareLuxx Germany) - link
Time: Oct 31, 2009

Pro.

Best decision recently.

Comment from diaazul - link
Time: Oct 31, 2009

This issue exactly hits the point on which I am wondering for a long time now.

In my opinion, it would be much better, if you distinguish between single GPU and multi GPU instead of single pci-e slot and multi pci-e slot.

At the moment, single GPU video cards have to compete with multi GPU cards ands thats somehow strange, isn't it?

Comment from Splave (OC Alliance) - link
Time: Oct 31, 2009

this is going to render the gtx-295 and 4870x2 as useless cards. Your going to need 2 of them now to compete.
2x4890s will smash a 4870x2 in a multi gpu category, as well as 2x285s will smash a gtx-295.
Why dont we just disallow posting scores with dual gpu vid cards. Problem solved.

Comment from r1ch (Bench Tec UK) - link
Time: Oct 31, 2009

Splave said: What this would do is make ln2 a must for any type of global points which would do the opposite of giving lower end hardware users a chance.

I think someone who has puts the skill and effort into benching with LN2 on the best single-gpu card should receive more points than someone who benches the best dual-gpu card on air. ;) That is the single biggest reason for this change IMO.

Splave said: Who the hell are you buddy?

I could say the same thing, but please don't resort to personal insults. :)

Splave said: We all get our opinions here. CPU talk is more relevant than you know. If you did this with gpu's how could you not follow suit with cpu's. This would allow people with low end hardware like a e8400 compete in wprime for example. Isnt that what you want? Or should there be a double standard.

I was not suggesting for one second that you don't get an opinion, but this is a thread about the GPU situation. There are too many differences between the way GPU's work, the drivers, the benchmarks to say categorically that they should just be the same. PLEASE read massman's post a few pages back to see why. Please say why you disagree with massman's reasoning if you still have the same opinion.

If you're bothered by the CPU situation, reply to one of the countless threads there already is on the CPU categories. :)

Splave said: Im reading alot of posts that just say. I agree, sounds good, good idea. Why dont you post why you think it is a good idea so there can be discussion ;)

Maybe because they read they agree with the "Pros" that massman detailed in the news post? If you want to promote discussion, how about bringing up reasons why those "Pros" are not valid reasons to change. ;)

Comment from knopflerbruce (PURE) - link
Time: Oct 31, 2009

This thread is about the boint algorithm, just because most folks talk about GPUs doesnt make it a GPU only thread. The wPrime discussion is highly relevant to the topic.;)

If an own category for multiGPU-cards should be created, I think we should at least separate multi-socket and single-socket scores for wPrime. Perhaps that's a better solution than to split by number of cores, as that would create more mess than anyone can imagine... :p

Comment from K404 - link
Time: Oct 31, 2009

Mr.Scott said: This is crap. If I have something that nobody else has benched, I expect to be rewarded for it.

knopflerbruce said: +1

If not, then nobody will bench non-mainstream HW. Plus, the idea is to reward skill too - and you can show skill even if there is no competition, so the 2 points you gt are well deserved.:)

Silverware not enough? Are points the only thing that drives you?

IMHO.. without competition, where is the comparison of skill? Anyone can install 3DMxx and hit run. Dont even need to overclock. 20 results isnt asking for much across the Western world. Thats 4 guys.

If a submission category has no results... clearly no-one is benching it anyway!

Comment from komadyret (411 Overkill) - link
Time: Oct 31, 2009

I'm gonnna put my 2 cents into this too;
If someone care to bench a certain hardware-set, but no one else cares, should the one that puts an effort into it be punished for taking a few hours tweaking a non-overclockable hardware-set (of which there are a few) to give decent benchmarks? I'd rather say that the effort should be worth a little reward to complete the hwbot database.

If you are offended by seeing someone get those 10-20 points, go out and get that hardware yourself and see if you can do any better instead of whining about it!

imho...

Comment from Thor941 (JMax-Hardware) - link
Time: Oct 31, 2009

komadyret said: I'm gonnna put my 2 cents into this too;
If someone care to bench a certain hardware-set, but no one else cares, should the one that puts an effort into it be punished for taking a few hours tweaking a non-overclockable hardware-set (of which there are a few) to give decent benchmarks? I'd rather say that the effort should be worth a little reward to complete the hwbot database.

If you are offended by seeing someone get those 10-20 points, go out and get that hardware yourself and see if you can do any better instead of whining about it!

imho...

+1

Comment from CaNNon (411 Overkill) - link
Time: Oct 31, 2009

knopflerbruce said: This thread is about the boint algorithm, just because most folks talk about GPUs doesnt make it a GPU only thread. The wPrime discussion is highly relevant to the topic.;)

If an own category for multiGPU-cards should be created, I think we should at least separate multi-socket and single-socket scores for wPrime. Perhaps that's a better solution than to split by number of cores, as that would create more mess than anyone can imagine... :p

I would agree, a quad core is a quad core and 2 quad cores are not a quad core.
Whats fair in having them compete in the same category?

Yep it may be a mess at first but the future looks to be going to more and more multi core vs GHZ might it just be better to move that way now?

Comment from DrNip (OC Alliance) - link
Time: Oct 31, 2009

I'm not trying to veer off course but the suggestion of increasing the hardware points is something I always pondered on too. I never understood why there was only a limit of 300 points. That could be a way to combat the situation as many of overclockers have older hardware that doesn't gain them any global points but have reached their 300 point hardware point ceiling. I say why have a ceiling? You don't have one on global points. That might be a little bit of influence on people breaking back out all the old hardware they have which in return equals more participation.

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Oct 31, 2009

drnip said: I'm not trying to veer off course but the suggestion of increasing the hardware points is something I always pondered on too. I never understood why there was only a limit of 300 points. That could be a way to combat the situation as many of overclockers have older hardware that doesn't gain them any global points but have reached their 300 point hardware point ceiling. I say why have a ceiling? You don't have one on global points. That might be a little bit of influence on people breaking back out all the old hardware they have which in return equals more participation.

+1
Here, here. :celebration:

Silverware not enough? Are points the only thing that drives you?

Yes. If silverware was all I wanted, I'd just bench PCmark04.

Originally Posted by komadyret
I'm gonnna put my 2 cents into this too;
If someone care to bench a certain hardware-set, but no one else cares, should the one that puts an effort into it be punished for taking a few hours tweaking a non-overclockable hardware-set (of which there are a few) to give decent benchmarks? I'd rather say that the effort should be worth a little reward to complete the hwbot database.

If you are offended by seeing someone get those 10-20 points, go out and get that hardware yourself and see if you can do any better instead of whining about it!

imho...

Right on. :)

Comment from SoF (AwardFabrik) - link
Time: Oct 31, 2009

I personally do not like the idea of a change as it makes you need more hardware laying around. To get up in booth rankings I first bought 4870X2 and then I bought a second one. Considering the new rules I would have to buy 4890 or 5870 additionally. Single cards are single cards, what is going to happen in 2-3 years maybe when all GPUs are multicores somehow...do we need to change the rules again then?
Back in the day the rule was decided as it is now I accepted it and bought whatever was needed for the rankings. If it would change now I would face a HUGE loss in points and a big investment to find a good card to climb the new rankings. I wanna keep benching my X2 which would be TOTALLY useless with the new rules :( All 295 and X2 users would face a big problem and I hope it will not happen :o

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Oct 31, 2009

bazx said: why does it have to be a choice?

The more global categories, the less meaningful the points are you gather for the Overclockers League. If you have 2 categories, there are only 2 categories you need to bench hard in. If there are 20, you need to bench hard in all 20.

In addition, if we increase the amount of global categories for 3D, the balance between 2D and 3D skill for the Overclockers League will lost completely. Now it's already 12 versus 7, we don't want to increase that to 24 versus 7 I presume?

bazx said: is there a technical reason why we cannot have multiple GPU ranks?

No, we can rank results based on number of sockets, number of cores and number of cards.

rbuass said: No friend...i posted it cause i think there are many things to change before change GPU method.

Let us focus on one problem at the time :).

ether.real said: Splitting up the Single/Dual GPUs is all brass on the titanic until the CPU problem gets fixed.

There's very little HWBot can do about the CPU problem. This is a characteristic of the benchmarks used; Vantage Xtreme is one we could add, but it won't change the cpu problem in all other benchmarks.

Changing to a single GPU system does in fact solve the cpu problem a bit for at least half of the 3D rankings.

ether.real said: How is this any different than someone else coming with a 5870 that crushes you at stock speeds?

It's already a difference in money. Also, it's a new generation; not really a fair comparison. We can't ban innovation.

vasgto said: btw, what's a construction bot vs. kitchen bot vs. a maintenance bot ?

I think it's some custom post-count name or so. A forum feature, nothing to be bothered about. :)

knopflerbruce said: This thread is about the boint algorithm, just because most folks talk about GPUs doesnt make it a GPU only thread. The wPrime discussion is highly relevant to the topic.

It's mainly to gather opinions about the suggestion HWBoints change. I prefer to stick to this topic ... hehe :).

drnip said: I say why have a ceiling? You don't have one on global points. That might be a little bit of influence on people breaking back out all the old hardware they have which in return equals more participation.

We have a cap because the Overclockers League was designed to rank overclockers on extreme overclocking skill. This means:

- Extreme cooling (LN2)
- Extreme hardware (High-end systems)
- Extreme results (Best scores in the benchmark)

Without a cap, this league would be dominated by people who just bench very much and although I'm not at all saying that this doesn't require skill, it's not what the Overclockers League is designed for. Note that the amount of available hardware points increase spectacularly as hardware categories increase.

Comment from MMauro DuDu (Crysis.it Overclocking Team) - link
Time: Oct 31, 2009

i'm agree this opinion

Comment from DrNip (OC Alliance) - link
Time: Oct 31, 2009

Massman said:

We have a cap because the Overclockers League was designed to rank overclockers on extreme overclocking skill. This means:

- Extreme cooling (LN2)
- Extreme hardware (High-end systems)
- Extreme results (Best scores in the benchmark)

Without a cap, this league would be dominated by people who just bench very much and although I'm not at all saying that this doesn't require skill, it's not what the Overclockers League is designed for. Note that the amount of available hardware points increase spectacularly as hardware categories increase.

Well as it stands it is setup on who has the newest, greatest hardware which out dates itself every few months. It is so far lopsided that it takes money to play, or sponsors which took money to get, once you get past a certain point. I am all about getting more people involved and there are more small fry benchers out there than big fry benchers. Give us small timers some more room to play. Increase our maximum hardware points. The big dogs can bench older hardware as well and get those points too. What I would like to see is points not givin to people benching their stock Dells, HP's and what not. That takes no skills at all. But who am I, just another small fry in the bunch.

I thought the hardware points increased by more submissions of that hardware. So we will say GPU-A has 200 submissions and GPU-B has 2000 submissions. GPU-B would have higher hardware points than GPU-A.

In conclusion I say bring an official poll with an agree or disagree answer and see how it pans out this time. If nothing were to change and it remain the same I would be alright with it as there is no perfect solution and ya'll have done a great job in getting it where it is today. :)

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Oct 31, 2009

|ron said: It's the new generation vs the old one, so it's normal.
Your example is like complaining if someone with a tuned Ferrari from the 80's is beaten by someone with a new Ferrari Enzo... there's no match, but it's obvious and normal.
Our case is about driving a race with a 4cylinders car vs a V12... you can put efforts in tuning the poor 4cil engine but if someone will slightly tune the V12... good bye! ;)

How is a 4870X2 less obvious and normal than a 5870? Next, are we going to complain that a 4890 beats a 4850? A card is a card and it should stay that way.

Comment from Splave (OC Alliance) - link
Time: Oct 31, 2009

Unsubscribed.
Goodluck with the discussion. Either way im still down with hwbot.

Comment from CaNNon (411 Overkill) - link
Time: Nov 1, 2009

drnip said: Well as it stands it is setup on who has the newest, greatest hardware which out dates itself every few months. It is so far lopsided that it takes money to play, or sponsors which took money to get, once you get past a certain point. I am all about getting more people involved and there are more small fry benchers out there than big fry benchers. Give us small timers some more room to play. Increase our maximum hardware points. The big dogs can bench older hardware as well and get those points too. What I would like to see is points not givin to people benching their stock Dells, HP's and what not. That takes no skills at all. But who am I, just another small fry in the bunch.

I thought the hardware points increased by more submissions of that hardware. So we will say GPU-A has 200 submissions and GPU-B has 2000 submissions. GPU-B would have higher hardware points than GPU-A.

In conclusion I say bring an official poll with an agree or disagree answer and see how it pans out this time. If nothing were to change and it remain the same I would be alright with it as there is no perfect solution and ya'll have done a great job in getting it where it is today. :)

Maybe if we take that one step farther drnip, a more reasonable or simpler approach could be something along the line of increasing the value of global points like increasing it's payout ratio? If there is a fair number that could be used to balance this.

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Nov 1, 2009

Massman said:
There's very little HWBot can do about the CPU problem. This is a characteristic of the benchmarks used; Vantage Xtreme is one we could add, but it won't change the cpu problem in all other benchmarks.

Changing to a single GPU system does in fact solve the cpu problem a bit for at least half of the 3D rankings.

I fail to see how changing to a single GPU system fixes the CPU problem. Single GPUs are still incredibly CPU limited in the GPU tests, and the CPU tests are 25% of the overall score as it is in Vantage P, while they are only 5% of the score in Vantage X.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 1, 2009

drnip said: I thought the hardware points increased by more submissions of that hardware. So we will say GPU-A has 200 submissions and GPU-B has 2000 submissions. GPU-B would have higher hardware points than GPU-A.

This is exactly how it works.

Actually, it should be X unique users rather than number of submissions.

Comment from komadyret (411 Overkill) - link
Time: Nov 1, 2009

Massman said: This is exactly how it works.

Actually, it should be X unique users rather than number of submissions.

Haha! The thought hit me a year or so ago that hardwarepoints may increase if I spammed one category with a few dozen submissions ;)
(But if you look at my submission you'll see I didn't, I have mostly one submission for each piece of hardware as I like to keep my list clean and easy to browse, deleteing all but the best result)

Comment from K404 - link
Time: Nov 1, 2009

I know that old results are good for building a bigger picture of the hardwares performance but it would be nice if there was a way of making those results invisible in the active submissions for each user. Even if a user opted to "give the score up completely"- it is still included with the hardware results but is not associated with the user who made it in the first place (another way of keeping an old result off our score/submission sheet)

Comment from Iron (HwProject.net Italy OC Team) - link
Time: Nov 1, 2009

ether.real said: How is a 4870X2 less obvious and normal than a 5870? Next, are we going to complain that a 4890 beats a 4850? A card is a card and it should stay that way.

A 4890 is a 4870 with a newer stepping, nothing else.
a 5870 is something far distant from a 4x00 series.
A 4870x2 is 2xRV770... it's not a newer generation card, it's not a slightly tuned 4870, it's two cards in one, so it should be put in another category.
Same thing speaking about cpus in multicore: it doesn't make sense that someone has wprime WR, done with a Cluster pc composed by 64CPUs etc... it simply doesn't make sense for me ;)

Comment from DrNip (OC Alliance) - link
Time: Nov 1, 2009

Well if you don't want to take the cap off the hardware points I think they should be lifted some, like 500 as suggested. I also think that the dual GPU cards should be counted as dual and not single. Same goes with any other hardware like this.

I wish there was a way to classify the overclockers like there is in every other sport. Pro, non pro, novice and what not. This would prolly be impossible though but a thought.

Comment from franknos90 (PcBrain OC Team) - link
Time: Nov 1, 2009

I agree with these new possible rules.. I think that these new rules could improve the challenge between the clockers..

Good work

Comment from ether.real (NBOC :: No Borders OC Team) - link
Time: Nov 1, 2009

|ron said: A 4890 is a 4870 with a newer stepping, nothing else.
a 5870 is something far distant from a 4x00 series.
A 4870x2 is 2xRV770... it's not a newer generation card, it's not a slightly tuned 4870, it's two cards in one, so it should be put in another category.
Same thing speaking about cpus in multicore: it doesn't make sense that someone has wprime WR, done with a Cluster pc composed by 64CPUs etc... it simply doesn't make sense for me ;)

And a 5870 is essentially the same technology as a 4870X2 crammed onto a single chip.

Comment from Iron (HwProject.net Italy OC Team) - link
Time: Nov 1, 2009

So, according to you, everything done by ATi from raddy 2900xt to present day is the same thing... -.-
2900xt is improved in 3870, 3870 improved in 4870, 5870 is a 4870x2...
If you think that 2 VGAs with roughly the same processing units, built on a different fabrication process, with one gpu instead of 2 (with all of the components doubled) etc etc, are the same, well I think that every CPU from Intel today is equivalent to a PIII because they have something in common... nnaahh, I totally disagree ;)

Comment from knopflerbruce (PURE) - link
Time: Nov 1, 2009

K404 said: Silverware not enough? Are points the only thing that drives you?

IMHO.. without competition, where is the comparison of skill? Anyone can install 3DMxx and hit run. Dont even need to overclock. 20 results isnt asking for much across the Western world. Thats 4 guys.

If a submission category has no results... clearly no-one is benching it anyway!

For instance... there are very few results of Opteron 156, but you can get an idea of how good the score is by comparing with similar and more popular chips, like Opteron 150, FX55 etc.

There are very few categories completely without competition. Some server chips and models that were never released, but that's it - so the problem you talk about is nearly non-existant.

I do this for points and cups, and I can't find any arguments whatsoever that it's not "right" to do that.

20 results aren't that many, but do you want to bench a socket 940 server in the near future? If not, then maybe you see that it would make alot of chips unusable for boint collecting... which is not fair to those who put money and effort into that type of platforms - even if it's just a minority of the bot users.

Comment from der8auer (PC Games Hardware) - link
Time: Nov 1, 2009

In my opinion it would suck to split it up... keep it like it is now :)

What would happen to my poinst? I submitted results with 1x 9800GX2 and 2x 9800GX2. Would i loose the points of the 1x 9800GX2 scores because its both multi gpu? This would suck really much!

Comment from der8auer (PC Games Hardware) - link
Time: Nov 1, 2009

Okay thats bad :(

Comment from CaNNon (411 Overkill) - link
Time: Nov 2, 2009

Ouch, how many guys in a similar spot jmke?

Comment from knopflerbruce (PURE) - link
Time: Nov 2, 2009

The 3dmark01 singlecard ranking will be turned upside down without the dual-GPU cards. Not that it means anything, just an interesting observation;)

Comment from CaNNon (411 Overkill) - link
Time: Nov 2, 2009

It's best to know both sides going in, but it still needs doing.

Comment from _Datura_ (XtremeSystems) - link
Time: Nov 2, 2009

Counting the number of GPU's makes more sense. But this sort of problem also exists with the CPU's.

A stock dual socket Pentium 3 1 GHz setup will also beat my heavily overclocked single Pentium 3 CPU in wPrime.

Comment from Dragon Oc Team (HwMania-OC-Team - HwMania.org) - link
Time: Nov 2, 2009

jmke said: yes splitting up single and dual socket systems is something that can be discussed at a later point, I don't see anything stopped this move;
but it should only be viable for multi-cpu benches, aka: PCmark05 ranking, Wprime ranking;
for SuperPi and other benchies they should not get a multi-socket ranking...

if hwbot code can support this function, i think this is the best idea

Comment from Autokiller677 (PC Games Hardware) - link
Time: Nov 2, 2009

I think, it's good to change single card to single gpu, because there's no difference between one who benches a 4870X2 and one who benches 2*4870's, but for the one with the 4870X2 it's much easier to gain points.

The very best solution would be to create different categories for single, dual, triple and quad setups, and moving dual gpu cards to the multigpu category ist the first step in this direction.

jmke said: yes splitting up single and dual socket systems is something that can be discussed at a later point, I don't see anything stopped this move;
but it should only be viable for multi-cpu benches, aka: PCmark05 ranking, Wprime ranking;
for SuperPi and other benchies they should not get a multi-socket ranking...

If this could be realised, it would be great.

Comment from der8auer (PC Games Hardware) - link
Time: Nov 2, 2009

It's just not fair.

Masterwana posted this at PCGHX Forums:
"
ATI
3850X2 = 122
3870X2 = 3896
4870X2 = 6999
4850X2 = 261
Mobility 3870X2 = 7
Mobility 4870X2 = 21

Nvidia
7900 GX2 = 82
7950 GX2 = 1480
9800 GX2 = 3758
GTX 295 = 3950

20576 Submissions "

A lot of guys spend so much time, money and hard work in these scores. It would suck so hard if they don't "count" anymore :[

Comment from K404 - link
Time: Nov 2, 2009

jmke said: yes splitting up single and dual socket systems is something that can be discussed at a later point, I don't see anything stopped this move;

Ummm....you were quite vocally against this. I've brought up the possibility more than once?

Comment from Gomeler (PURE) - link
Time: Nov 2, 2009

I'd like to see single GPU, dual GPU, and, multi-GPU. The multi-GPU category will nearly always be a strictly money game($1000 minimum in GPUs right now) with very few LN2 benchers. The single-GPU category will be a skill/cooling game($350-$400 GPU) where those who use exotic cooling will be handsomely rewarded. The dual-GPU category will be a nice blend between the two($450-800 GPU or GPUs) where those who are willing to juggle two-three pots will have an edge.

Those of you that don't have LN2 readily available might think it is unfair that you can't easily compete near the top without either dumping a lot of money into GPUs or LN2 but that is the path that the hardware manufacturers have pushed us along.

wPrime needs a socket classes bad, those clustered machines are even more big money than the GPUs ;)

Comment from onkel-bill (PC Games Hardware) - link
Time: Nov 2, 2009

Hello,
my english isn´t really good, but I hope you guys understand what I mean...
In my opinion, there must be distinguish in GPU not in card. Also in quantity of GPU.
So the guys which the X2-cards will lose theire points in the single card category, but will get some in the new 2-GPU-Category. (There an amount of work is in it, it would be a punishment, if they lose all points, because these points going to 3-way or 4-way GPU setup.)
And the high scorer are certainly know, how to get points with a X2. They cooling their cards anyhow with LN2, water isn´t a callenge for these guys. They will be surly on top in this new category.
But so, I think, it will be fairer.
Regards from germany
onkel-bill

Comment from komadyret (411 Overkill) - link
Time: Nov 2, 2009

I know I've said what I had to say in this thread, but I don't think all the argument here hold water...

Those of you whining about loosing points if x2 cards are moved to the multi-gpu class; how would you feel if a manufacturer decided they'd like to make a monster-card? Are you willing to spend those $1000 for a 4gpu, or $3000 for a 8gpu card to grab a few global points i the "single-card category" as the insanity escalates?
And don't throw me that comparison with CPUs that have multiple cores. GFX cards have hundreds of cores already. A 2-gpu card is best compared to a 2-cpu-socket board, not a 2core cpu. (and this cpu-buisness is a similar discussion I'm sure we'll take later)

Bench that gx2/x2, earn those hardware points by benching against similar configurations and don't expect a free ride by exploting a hole in the rule-set. If you're good at what you do, you should enjoy the challenge. What's the satisfaction of grabbing 10-20-30 global points if you can't grab a single hardwarepoint for that piece of hardware? Doesn't it mean that many of those who cares to spend the cash will probably beat you? I know that I don't stop beating my benching hardware until I've earned at least a tenth of a hardware point, or killed it in the process of trying!

And one more thing, I think global points should be awarded for the world top scores, without regard for the number of gpu's or cpu's. I think that global points should be just that. Like in the CPU class. But for hardware points there should be classes based upon number of gpus and cpu-sockets. Bench for the silverware and the hardware-glory, but be prepared to spend the big money if you aim for globals!

But this is just my opinion. Thousands of benchers would loose a lot of points if this was done, as most single-card (not only x2/gx2 owners) would loose almost all global points, so it'll probably never happen.

Comment from [UC OC] Karmakazi (PURE) - link
Time: Nov 3, 2009

I totally agree. I am against making GX2s/X2s/etc into multi card. If you created a new ranking for dual gpu that would be OK. But all the time money and LN2 spent benching these cards would be a major loss if points are taken away. I dont have a big budget to buy 4x5870 4x285classified etc etc.

I think the fairest way to do this would be to make 3 ranking categories..

1.) single card & single gpu
2.) dual card or Dual gpu single card
3.) multi card no limits.

As someone posted earlier, what happens when someone makes the X4 or X8 monster card. That would totally mess up single card. But having the single, dual, and multi categories would make it more economical for users to compete (which is what this should be about), and fair to the people who have spent alot on dual gpu cards. just my 2 cents

Jamie

der8auer said: It's just not fair.

Masterwana posted this at PCGHX Forums:
"
ATI
3850X2 = 122
3870X2 = 3896
4870X2 = 6999
4850X2 = 261
Mobility 3870X2 = 7
Mobility 4870X2 = 21

Nvidia
7900 GX2 = 82
7950 GX2 = 1480
9800 GX2 = 3758
GTX 295 = 3950

20576 Submissions "

A lot of guys spend so much time, money and hard work in these scores. It would suck so hard if they don't "count" anymore :[

Comment from rbuass (Team Adrenaline) - link
Time: Nov 3, 2009

I only allow the change if change the multi core CPU too.
i think 2 x GTX285 is not the same category than a GTX 295...
But...i think is not fair play 2 X HD5870 together 3 X HD5870 together 4 X HD5870.
To me the number of slots is the correct.
i am contra the changes.

Comment from CaNNon (411 Overkill) - link
Time: Nov 3, 2009

der8auer said: It's just not fair.

Masterwana posted this at PCGHX Forums:
"
ATI
3850X2 = 122
3870X2 = 3896
4870X2 = 6999
4850X2 = 261
Mobility 3870X2 = 7
Mobility 4870X2 = 21

Nvidia
7900 GX2 = 82
7950 GX2 = 1480
9800 GX2 = 3758
GTX 295 = 3950

20576 Submissions "

A lot of guys spend so much time, money and hard work in these scores. It would suck so hard if they don't "count" anymore :[

You know that can also be looked at as 20576 times that single gpu was unfairly beaten by a dual-gpu solution? ;)

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 3, 2009

Hey guys,

Thanks for the insane amount of feedback. Last night we discussed all remarks in the HWBOT meeting we're trying to see what's feasible in terms of coding as well as 'concept'.

:D

Comment from komadyret (411 Overkill) - link
Time: Nov 3, 2009

You're welcome, Massman.
Not sure we helped much, but at least we had a good discussion going ;)

Comment from der8auer (PC Games Hardware) - link
Time: Nov 3, 2009

CaNNon said: You know that can also be looked at as 20576 times that single gpu was unfairly beaten by a dual-gpu solution? ;)

Thats also true but it's just to late to change. Too much scores would be taken away...

Comment from TheGreenFoX (overclocking.dk) - link
Time: Nov 3, 2009

Could it be possible to give extra points for fx best air results with each hardware?

Comment from Thor941 (JMax-Hardware) - link
Time: Nov 3, 2009

TheGreenFoX said: Could it be possible to give extra points for fx best air results with each hardware?

how can you prove it is air!?

Comment from CaNNon (411 Overkill) - link
Time: Nov 3, 2009

der8auer said: Thats also true but it's just to late to change. Too much scores would be taken away...

I due understand about the mess and loss to many members! But what if a card gets released with more that 2 cores? This is something that would become really lopsided and hurt even more bencher scores.

Comment from K404 - link
Time: Nov 3, 2009

What if the changes were only implemented for future GX2 cards? All the existing GX2 scores will fall down the (single card) global rankings until they stop being a big problem.

I see it in the same way that existing old scores (any category... 2006, 2007 scores etc) still stand as they were within the rules set at the time of submission.

Comment from knopflerbruce (PURE) - link
Time: Nov 3, 2009

K404 said: What if the changes were only implemented for future GX2 cards? All the existing GX2 scores will fall down the (single card) global rankings until they stop being a big problem.

I see it in the same way that existing old scores (any category... 2006, 2007 scores etc) still stand as they were within the rules set at the time of submission.

It would make the database look weird, but I see your point. IMO it would be most fair to move the scores a couple of months after a major launch (GT300?).

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 3, 2009

That would be a mess to implement. : /

Comment from Brian y. (PURE) - link
Time: Nov 4, 2009

Gomeler said: I'd like to see single GPU, dual GPU, and, multi-GPU. The multi-GPU category will nearly always be a strictly money game($1000 minimum in GPUs right now) with very few LN2 benchers. The single-GPU category will be a skill/cooling game($350-$400 GPU) where those who use exotic cooling will be handsomely rewarded. The dual-GPU category will be a nice blend between the two($450-800 GPU or GPUs) where those who are willing to juggle two-three pots will have an edge.

Those of you that don't have LN2 readily available might think it is unfair that you can't easily compete near the top without either dumping a lot of money into GPUs or LN2 but that is the path that the hardware manufacturers have pushed us along.

wPrime needs a socket classes bad, those clustered machines are even more big money than the GPUs ;)

I agree with this totally. Very well put.

Comment from Linuxfan (XtremeSystems) - link
Time: Nov 4, 2009

IMO 3 catagories for the GPUs is the best for now. Maybe if GPUs change and we get more cores like CPUs are going then we would need to change it, but for now that would work the best.
We would have great global competition for the single GPU dual cards, as well as the single cards that most people have access to, and the multi GPUs would still have their global competition as well. Sounds great to me!

For the CPU issue I think there should be a way to categorize them into single socket and multi socket solutions, especially because of the Wprime scaling. It would definitely allow harder benching of lower CPUs than core i7s in Wprime, and would be more fair IMO.
Just my thoughts. :)

Comment from K404 - link
Time: Nov 4, 2009

Having 3 GPU categories gives 3 global leaderboards so it will cost MORE money to compete at the top level. 3 categories will benefit the guys with GPU sponsorship

Comment from TheGreenFoX (overclocking.dk) - link
Time: Nov 4, 2009

Thor941 said: how can you prove it is air!?

Temperature, pics of setup... many ways, but no real way to PROVE it... But its kinda the same with ES chips etc... no way to prove it

Comment from komadyret (411 Overkill) - link
Time: Nov 4, 2009

TheGreenFoX said: Temperature, pics of setup... many ways, but no real way to PROVE it... But its kinda the same with ES chips etc... no way to prove it

ES chips are identifiable by stepping, not always but often, and are also often run on non-default multi (and quite commonly doesn't clock that much better than retail as far as I've seen)

If you will try to divide on cooling, I guess many of my future best results (on air, as I have supply problems on dice/ln2 and tight funding) will be forced into the wrong category (Dice/SS) unless I make a movie of the benching?
Really J/K, but I easily exploit my advantage of having sub-zero ambients for benching in my garage part of the year and can see questions raised ;D

Comment from knopflerbruce (PURE) - link
Time: Nov 4, 2009

More GPU categories will make the 3d part of HWBot too big. Atm it's 7 vs 12 (CPU vs GPU), and it will be 7 vs 18 if another set of categories will be made. That's simply too much.

Comment from knopflerbruce (PURE) - link
Time: Nov 4, 2009

jmke said: we need to app a few more 2D benchies, I hear you :)

A long benchmark would be nice, even superpi 32m is nearly over before it starts:p 30min+, and it would be nice if the test would last equally long on all platforms. This can be done for instance by letting the score be the number of calculations made (divided by some number if necessary) within a fixed time frame.

We also need something that's memory intensive, superpi32m used to be that, but isn't anymore (135mb or what it uses is nothing compared to the 2gb we find in most systems today).

Comment from der8auer (PC Games Hardware) - link
Time: Nov 4, 2009

K404 said: Having 3 GPU categories gives 3 global leaderboards so it will cost MORE money to compete at the top level. 3 categories will benefit the guys with GPU sponsorship

+1.
A 3rd categorie should not be an option.

I think we should stay at the 3D discussion and after that solve the multi CPU, wPrime problem and so on...

Comment from K404 - link
Time: Nov 4, 2009

knopflerbruce said: A long benchmark would be nice, even superpi 32m is nearly over before it starts:p 30min+, and it would be nice if the test would last equally long on all platforms. This can be done for instance by letting the score be the number of calculations made (divided by some number if necessary) within a fixed time frame.

We also need something that's memory intensive, superpi32m used to be that, but isn't anymore (135mb or what it uses is nothing compared to the 2gb we find in most systems today).

Im all different kinds of agreeing to this :D I dont mind long benches at all as long as they're nice to watch. Vantage isn't. Seeing as dice + LN2 benchers will have to be at their rigs for the bench duration it should ideally be watchable.

Id like to see some burst CPU activity, some prolonged, some towards-2GB RAM useage (I would say more but the best 775 systems are 2x1GB) or somehow give score boosts appropriately between speed and volume.

Time dependent benches dont work. AM3 and wP are almost writeoffs now and SPi is heading the same way.

Comment from K404 - link
Time: Nov 4, 2009

Ummmmmmm..... yes *embarressed face* I like watching Spi32M

Im never going to have sex again, am I? :p

Comment from knopflerbruce (PURE) - link
Time: Nov 4, 2009

jmke said: SuperPI Loops are nice to watch? :D

Lmao. It's like if the time is .016s different you can go from :mad: to :woot: instantly:D Or the other way around.

Comment from komadyret (411 Overkill) - link
Time: Nov 4, 2009

Bench a Pentium2 with Superpi32m... You have NO idea how many times you can re-estimate an end-result by looking at the loops for 2-3 hours ;) :D

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Nov 5, 2009

komadyret said: Bench a Pentium2 with Superpi32m... You have NO idea how many times you can re-estimate an end-result by looking at the loops for 2-3 hours ;) :D

I hear that. I bailed on the SPI32m I had running on some Pentium I's. I just couldn't take it any more.

Comment from CaNNon (411 Overkill) - link
Time: Nov 5, 2009

k404 said: ummmmmmm..... Yes *embarressed Face* I Like Watching Spi32m

Im Never Going To Have Sex Again, Am I? :p

Lol!

Comment from 69_Goat (OC Alliance) - link
Time: Nov 5, 2009

Gomeler said: I'd like to see single GPU, dual GPU, and, multi-GPU. The multi-GPU category will nearly always be a strictly money game($1000 minimum in GPUs right now) with very few LN2 benchers. The single-GPU category will be a skill/cooling game($350-$400 GPU) where those who use exotic cooling will be handsomely rewarded. The dual-GPU category will be a nice blend between the two($450-800 GPU or GPUs) where those who are willing to juggle two-three pots will have an edge.

Those of you that don't have LN2 readily available might think it is unfair that you can't easily compete near the top without either dumping a lot of money into GPUs or LN2 but that is the path that the hardware manufacturers have pushed us along.

wPrime needs a socket classes bad, those clustered machines are even more big money than the GPUs ;)

I almost agree with this, but think it should be taken 1 step further and have a separate catagory for each number of GPU's. Furture proof this thing, now. How long is it going to be before we have the capabliity to bench a monster card with 6 or 8 or more GPU's on it? Do we want to dump those scores into the multi-gpu catagory and/or re-visit this issue again at that time?

Same goes for the CPU discussion. The time to make these changes is now.

If this change is needed to make things fair, then make it fair. As an X2 bencher, I feel it is just as unfair to dump my single card scores into the multi-gpu catagory, as it is to have them lumped in with the single gpu scores.

If this change is needed to improve participation and give low budget benchers a chance at points, then make more catagories. Single, dual, tri, quad, etc.
It wont matter how they are arranged, as the top spots will always be filled by hardcore LN2 benchers, as it should be.

Comment from knopflerbruce (PURE) - link
Time: Nov 5, 2009

69_GTO_Judge said: I almost agree with this, but think it should be taken 1 step further and have a separate catagory for each number of GPU's. Furture proof this thing, now. How long is it going to be before we have the capabliity to bench a monster card with 6 or 8 or more GPU's on it? Do we want to dump those scores into the multi-gpu catagory and/or re-visit this issue again at that time?

Same goes for the CPU discussion. The time to make these changes is now.

If this change is needed to make things fair, then make it fair. As an X2 bencher, I feel it is just as unfair to dump my single card scores into the multi-gpu catagory, as it is to have them lumped in with the single gpu scores.

If this change is needed to improve participation and give low budget benchers a chance at points, then make more catagories. Single, dual, tri, quad, etc.
It wont matter how they are arranged, as the top spots will always be filled by hardcore LN2 benchers, as it should be.

The only way to make this fair for 2d benchers would be to lower the amount of boints given to each 3d benchmark. If you double the number of benchmarks with a card, you'll have to divide HW + global boints by 2, as well. Or multiply 2d boints by 2. Not sure how good this solution is, it may hurt singleGPU rankings alot - which are the most popular 3d ranks (maybe even overall, I dunno).

Comment from CaNNon (411 Overkill) - link
Time: Nov 6, 2009

I'm not sure why they need to be equals? any reason they need to be?

Comment from knopflerbruce (PURE) - link
Time: Nov 6, 2009

CaNNon said: I'm not sure why they need to be equals? any reason they need to be?

Not equal, but not more unequal than they are now.

Comment from rbuass (Team Adrenaline) - link
Time: Nov 6, 2009

Any news about the result?
I dont agree the changes but i think everybody is waiting the decision
;)

Comment from K404 - link
Time: Nov 10, 2009

Another not-thought-out idea....

what if the Hardware points that contributed to global were:

a % of the users global points
a % of the No.1 users global points

? :)

a set no. of hardware points doesnt have any real context... its just a number. If the hardware points-as-global was a %, it looks after itself :)

Comment from by_xfile (HwProject.net Italy OC Team) - link
Time: Nov 10, 2009

I honestly thought you had already reached a verdict on what to do

Comment from komadyret (411 Overkill) - link
Time: Nov 10, 2009

K404 said: Another not-thought-out idea....

what if the Hardware points that contributed to global were:

a % of the users global points
a % of the No.1 users global points

? :)

a set no. of hardware points doesnt have any real context... its just a number. If the hardware points-as-global was a %, it looks after itself :)

Not a bad idea at all K404. Perhaps not in the exact way you lay it out here, but to have a hw points as a mathematical dependency of globals would be a nifty way to solve it :)

Comment from Predator (Team OCX) - link
Time: Nov 10, 2009

Gomeler said: I'd like to see single GPU, dual GPU, and, multi-GPU. The multi-GPU category will nearly always be a strictly money game($1000 minimum in GPUs right now) with very few LN2 benchers. The single-GPU category will be a skill/cooling game($350-$400 GPU) where those who use exotic cooling will be handsomely rewarded. The dual-GPU category will be a nice blend between the two($450-800 GPU or GPUs) where those who are willing to juggle two-three pots will have an edge.

Those of you that don't have LN2 readily available might think it is unfair that you can't easily compete near the top without either dumping a lot of money into GPUs or LN2 but that is the path that the hardware manufacturers have pushed us along.

wPrime needs a socket classes bad, those clustered machines are even more big money than the GPUs ;)

i pretty much agree with this idea so far :) in every single point

i would also encourage to give more weight to hardware points, i mean, imagine you are benching single card for trying to get the max out of global points for that category, let's say, 3D05, and you bench with your 4870 GPU, you push really hard and you do a quite huge investmnet with LN2 cooling which involves pots and all that gear stuff, then you achieve let's say no.3 or no.4 in the rankings, then next gen hits the market and whoever at stock cooling pulls your global score down badly , now you are battling in hardware points as your old fashiones GPU cannot have a chance agaisnt the new gen, that's why i would give more weight to that hardware category, let's say, instead getting aproximately 15 points for the 1st in hardware points let's double to 30, i mean of course in those who are very popular, not in every single one VGA/CPU

that's my 2 cents :)

Comment from lakeside2189 (OC Alliance) - link
Time: Nov 12, 2009

drnip said: Well if you don't want to take the cap off the hardware points I think they should be lifted some, like 500 as suggested. I also think that the dual GPU cards should be counted as dual and not single. Same goes with any other hardware like this.

I wish there was a way to classify the overclockers like there is in every other sport. Pro, non pro, novice and what not. This would prolly be impossible though but a thought.

This is really the only thought if you want to keep people interested. I bench with a x2 and if I loose my points then so be it. The bottom line is I might have benched with a x2 but did it on air and H20 in a case at 4.500-4.625..After I finished I used the same PC to email my family or my wife did her ebay thing. I know it's impossible to separate Pro from Novice to everyday PC kinda guy...The down side is if you turn this into a all Pro LN2 thing you are going to loose alot of people that really have fun with this..

I am not trying to down the Pro's at this..I love to watch them. I also know that a guy trying to push 4.6+ in his everyday PC is also needed...

I guess I really don't understand how making this change fair or not is going to help the guy who likes to bench with his or her's PC...in a case...that they use everyday...
Also could take the multi GPU single card everyday users out of the picture..And alot of upcoming multi GPU single cards..I can tell you I won't be buying the new x2 as planned.

Do what you need to do now but consider how to break this down for different type's of benchers..
When I go to the drag races I love to watch the super pro's but also the stock cars...

For those of us that are considered stock, running 4.6+ and a x2 or 295. This is not Stock to us.
I am sure there is no way to separate this but in the future if you want full participation it would be well worth figuring out...

Sorry about getting off the point...Just thought it needed to be said...

Comment from by_xfile (HwProject.net Italy OC Team) - link
Time: Nov 12, 2009

lakeside2189 said: This is really the only thought if you want to keep people interested. I bench with a x2 and if I loose my points then so be it. The bottom line is I might have benched with a x2 but did it on air and H20 in a case at 4.500-4.625..After I finished I used the same PC to email my family or my wife did her ebay thing. I know it's impossible to separate Pro from Novice to everyday PC kinda guy...The down side is if you turn this into a all Pro LN2 thing you are going to loose alot of people that really have fun with this..

I am not trying to down the Pro's at this..I love to watch them. I also know that a guy trying to push 4.6+ in his everyday PC is also needed...

I guess I really don't understand how making this change fair or not is going to help the guy who likes to bench with his or her's PC...in a case...that they use everyday...
Also could take the multi GPU single card everyday users out of the picture..And alot of upcoming multi GPU single cards..I can tell you I won't be buying the new x2 as planned.

Do what you need to do now but consider how to break this down for different type's of benchers..
When I go to the drag races I love to watch the super pro's but also the stock cars...

For those of us that are considered stock, running 4.6+ and a x2 or 295. This is not Stock to us.
I am sure there is no way to separate this but in the future if you want full participation it would be well worth figuring out...

Sorry about getting off the point...Just thought it needed to be said...

this line of thought I'm agree with you, because for many users this is pure entertainment with a PC all day, and not PC dedicated to WR or bencmarck

Comment from knopflerbruce (PURE) - link
Time: Nov 12, 2009

Why on earth would you want to create an own category for PCs at stock speeds? The whole idea is that you overclock the rig to the max.

If you don't OC you can't expect to get a good score either. The scenario you put up is like having a race in some neighborhood, just that your car has to follow any speed limit along the way. How fun is that?;)

Comment from K404 - link
Time: Nov 12, 2009

It would be good if every year a couple of machines could be tested by trusted members that use common specs and have stock scores put up so theres a baseline of performance? :)

Comment from knopflerbruce (PURE) - link
Time: Nov 12, 2009

K404 said: It would be good if every year a couple of machines could be tested by trusted members that use common specs and have stock scores put up so theres a baseline of performance? :)

...and what are you going to do with those numbers?;)

Comment from Iron (HwProject.net Italy OC Team) - link
Time: Nov 12, 2009

After reading sooo much comments, I slightly changed my idea... or better, I'm not sure that there is a right way of thinking and a wrong one.
In other words: all depends on what you (the staff) think about what hwbot should be.
- Should it be a place where overclocking skills count for the main part?
- Should it be a place where a pure score at a benchmark counts more than the rest?

So, in the first case, you should put global points divided for single gpu, global points for 2gpus and global point for 3+ gpus.
In the second case then... it's already in that way that hwbot works today.

My personal support would go to the first case ;)

Comment from K404 - link
Time: Nov 12, 2009

knopflerbruce said: ...and what are you going to do with those numbers?;)

Well it would be good to see how each bench suite scores with popular hardware at stock :) Maybe also good to see what boost over stock the same hardware will net a user in whichever bench once they overclock

A semi-aside. 42K in Vantage is good... 136K in 3D01 is good... but whats the context NOW? In 6 months... a year... the scores will be meaningless. It will be good to see how hard people are pushing to get the numbers that inspire us now, in the context of what the same gear is capable of at stock :)

Comment from lakeside2189 (OC Alliance) - link
Time: Nov 12, 2009

knopflerbruce said: Why on earth would you want to create an own category for PCs at stock speeds? The whole idea is that you overclock the rig to the max.

If you don't OC you can't expect to get a good score either. The scenario you put up is like having a race in some neighborhood, just that your car has to follow any speed limit along the way. How fun is that?;)

I am not saying STOCK speeds!!.I am saying there are some of us that use our everyday PC and clock them on air and H20 at 4.6-4.7 for benching..That is not stock!!!! If i am ranked 750ish out of 20,000 I can't be the only one that pushes his everyday PC..

Bots needs numbers(people) just like any other forum or site. If you just had the hardcore pros you would have less than 5000 members probably.

I am not saying run stock by NO means...Even with a X2 i had to push my PC to score 31,000+ at 06...4.625 to score in 03..Maybe 4.6-4.7 is stock to the hardcore guy but alot of us think we are doing something...If nothing else we are here, having fun, members of Bots, and clocking...Unless you have a 4.7 stock CPU???? This is where the different levels come in that Dr.Nip referred too...

I am saying there are several levels of benching and in the end bots will want and need these benchers.. I am proud of my 80-130 points I have scored on Bots..I know it's not Pro stuff but good stuff none the less and I had to clock to get them. They we're not done running STOCK..It takes all the levels to make Bots really work..That is why they have 20,000 members..

Comment from knopflerbruce (PURE) - link
Time: Nov 12, 2009

lakeside2189 said: I am not saying STOCK speeds!!.I am saying there are some of us that use our everyday PC and clock them on air and H20 at 4.6-4.7 for benching..That is not stock!!!! If i am ranked 750ish out of 20,000 I can't be the only one that pushes his everyday PC..

Bots needs numbers(people) just like any other forum or site. If you just had the hardcore pros you would have less than 5000 members probably.

I am not saying run stock by NO means...Even with a X2 i had to push my PC to score 31,000+ at 06...4.625 to score in 03..Maybe 4.6-4.7 is stock to the hardcore guy but alot of us think we are doing something...If nothing else we are here, having fun, members of Bots, and clocking...Unless you have a 4.7 stock CPU???? This is where the different levels come in that Dr.Nip referred too...

I am saying there are several levels of benching and in the end bots will want and need these benchers.. I am proud of my 80-130 points I have scored on Bots..I know it's not Pro stuff but good stuff none the less and I had to clock to get them. They we're not done running STOCK..It takes all the levels to make Bots really work..That is why they have 20,000 members..

Are you asking for an amateur league?

Comment from komadyret (411 Overkill) - link
Time: Nov 12, 2009

More or less my thoughts when reading that post...

Not sure it's easily achievable tho... not even sure I'd like it

Comment from knopflerbruce (PURE) - link
Time: Nov 12, 2009

komadyret said: More or less my thoughts when reading that post...

Not sure it's easily achievable tho... not even sure I'd like it

It's impossible to implement... who can decide who is an amateur and who is not? And how do we verify that the scores are done on "non-expert"-setups?

I like the idea (to reward amateur scores more), but there are better ways to do that. Maybe it's possible to award all entries with a tiny bit of boints, like .1. It's at least something, run all benchmarks and get a boint or two.

Maybe it's possible to award the first 20(?) results for all users with 2 "rookie boints" pr result. It definately won't screw up any ranks, and it may encourage more people to bench. Some fine tuning would be nice, like maybe the first 20 results that don't receive any boints will get the rookie boints (no need to award scores that get either global or hwboints with this little extra motivation).

Comment from lakeside2189 (OC Alliance) - link
Time: Nov 12, 2009

knopflerbruce said: Are you asking for an amateur league?

I'm not really asking for anything...I am just stating facts where alot of us come from...I really don't want anything to change..If it does then so be it...I was just agreeing with a post about different leagues but understand that can't happen...

Comment from K404 - link
Time: Nov 13, 2009

Will HWB admins make a decision on X2/GX2 cards before the new ATI dual-die cards are released?

I believe thats an important question :)

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 13, 2009

When are they due?

The next hwbot staff meeting is Monday, I think chances are high we will come to a consensus. : ) There will be no major engine release while the MSI P55 competition is running though, so the hwboints rev3 engine will most likely be released 1st of January.

Comment from K404 - link
Time: Nov 13, 2009

NDA 18th Nov

Release.....well for cards-in-hands and not vapourware...ask TSMC :p

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 13, 2009

richba5tard said: I think chances are high we will come to a consensus.

So it's doable in terms of coding? :-)

Comment from MTP04 (XtremeSystems) - link
Time: Nov 18, 2009

These dual gpu cards, in a way, is turning extreme benching into semi-extreme +air/water, meaning LN2 on cpu and gpus on air and sometimes water, which is perfectly fine, sometimes I do it too BUT in return, with the way the algorithm is setup now, we are seeing less and less subzero single gpu cards. So, for seeing more extreme setups with gpu higher clocks, I am in favor of changing the algorithm to the number of gpus rather than to the number of cards.:)

Comment from komadyret (411 Overkill) - link
Time: Nov 18, 2009

Very good point there too MTP04 :)

Comment from Splave (OC Alliance) - link
Time: Nov 19, 2009

MTP04 said: These dual gpu cards, in a way, is turning extreme benching into semi-extreme +air/water, meaning LN2 on cpu and gpus on air and sometimes water, which is perfectly fine, sometimes I do it too BUT in return, with the way the algorithm is setup now, we are seeing less and less subzero single gpu cards. So, for seeing more extreme setups with gpu higher clocks, I am in favor of changing the algorithm to the number of gpus rather than to the number of cards.:)

yes but this is counter intuitive to the point being made of:
-making global points more viable for everyone even with less $$

Comment from thebanik (OC Alliance) - link
Time: Nov 20, 2009

Splave said: yes but this is counter intuitive to the point being made of:
-making global points more viable for everyone even with less $$

+1, It pains to mod a $500-600 card just to see it go bust during a bench session. Also personally speaking, I do not have a separate bench rig and regular rig. I buy my rig in a way which I can afford with my meager salary, as well as gets me some good boints when I use DICE/LN2 on the CPU.......

Comment from rbuass (Team Adrenaline) - link
Time: Nov 20, 2009

I think for exemple 2 X GTX285 is more powerfull than one GTX295.
I think its the number of slots (graphic cards and not GPU).
Is not fair compete 4 x HD5870 at same category than 2 X HD5870, but GTX295/HD4870X2 is only 1 videocard.

Comment from thebanik (OC Alliance) - link
Time: Nov 20, 2009

rbuass said:
I think its the number of slots (graphic cards and not GPU).
Is not fair compete 4 x HD5870 at same category than 2 X HD5870, but GTX295/HD4870X2 is only 1 videocard.

yup this I guess if possible would be simply great for every1 and nobody should complain.........

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 20, 2009

thebanik said: yup this I guess if possible would be simply great for every1 and nobody should complain.........

Yes but then rankings would be dominated by dual gpu, single socket videocards... which is the main reason for complaints now. :p

Comment from BenchZowner (Hellas Overclocking Team) - link
Time: Nov 20, 2009

In my opinion ( yeah I know, opinions are like assh0les, everybody has one :p ) the best way to do the 3D Benchmarks rankings is...
you're not going to like the amount...but here it comes:

1) Single GPU ( 1 GPU )
2) Single VGA - 2 GPU
3) 2 GPUs ( 2* VGAs )
4) 3 GPUs ( 3* VGAs )
5) 4 GPUs ( 4* VGAs )
6) Dual VGAs - 4 GPUs ( 2* VGAs )

This way, the ranks will be fair for everyone, and everybody's going to have a chance to get some decent points out of his hardware.
Having the 2 GPUs ( Single Card ) and 2 GPUs ( 2* VGAs ) in the same ranking will be unfair since it's going to be just like now, a "5970" dominating the ranking, and the other SLI/CF setups lagging behind.

Comment from K404 - link
Time: Nov 20, 2009

But thats 6 global leaderboards, even if your categories 2 + 6 will mostly be dominated by one card at a time and will almost be an extension of the hardware category :p Category 5 will be ATI only. Not fair competition.

Whoever gets the best support from GPU companies will do best- the cost of competing globally, if someone is buying their own hardware, gets retarded ridiculous.

Comment from BenchZowner (Hellas Overclocking Team) - link
Time: Nov 20, 2009

K404 said: But thats 6 global leaderboards

That's not an excuse :p

K404 said: even if your categories 2 + 6 will mostly be dominated by one card at a time

Maybe in 3D Mark03 and/or 05.
But in 2k1SE & 06 & AquaMark 3 they could be trading blows.

K404 said: and will almost be an extension of the hardware category :p Category 5 will be ATI only. Not fair competition.

Category 5 = eVGA GTX 285 + Classified E762 :p
And I have a good feeling that in the future we'll see more 4-way SLI compatible nVIDIA cards.

K404 said: Whoever gets the best support from GPU companies will do best

The same applies everytime.
If you can get the hardware and you have the skills the ranking is yours.

However, the difference with these rankings will appeal the "normal" overclockers and users, as it will be easier to gain some points with lower end and midrange cards.
Now in the single card category you can't get any points ( apart from HWboints ) with a 9400GS for example.
Why ? Mostly because the dual-GPU cards dominate the charts and then are followed by the best single-GPU card and then... you're over the global rank awarding points limit and you get 0 points.

The big boys remain big, while the newcomers and the average joe gets a chance to get some points and a motive to try harder and maybe even get into extreme cooling.

Comment from K404 - link
Time: Nov 20, 2009

Its too much.

I dont believe it will be easier for less extreme guys. LN2 will be absolutely necessary for single GPU points. The gap between LN2 cooled and watercooled will be just as big, but there will be more categories for the gap to manifest itself.

Take all the LN2 cooled 260s, 275, 280s and 285s.... 5870, 5850, thats probably the top....300? accounted for for a single card category. Then the LN2 cooled 9800GTX+, GTS512MB, 4890......they'll still be capable-ishhhh

Watercooled GPUs, often still needing an LN2-cooled CPU- start fighting for rank No.301. Whats that... 3 global points?

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 20, 2009

Thank you for all the feedback! After thorough consideration, we are currently basing hwboints rev3 on the following concept:

1. Split up of rankings based on number of cores for both videocards and processors.
Currently we have single socket and multisocket rankings for videocards, and one ranking for processors. In rev3, we will no longer base the ranking on the amount of sockets, but the total amount of cores. To avoid the problem that you would have to participate in 8 (eg.) 3DMark05 rankings, members will only recieve "global points" once for each benchmark application: only for their highest ranked submission regardless of the amount of cores.

Eg. if a user is 3rd with 2x HD4780, 2nd with HD4780X2, 4th with 1x GTX285 and and 5th with 4x GTX270 in the global ranking of 3DMark01, he will only receive hwboints for being 2nd place in the 2 cores ranking (HD4780X2).(*) All his submissions are applicable for hardware points.

(*) Actually, this would only be true of there would be equal amount of participants in each ranking. If there are a lot less participants in the 2 cores ranking than the 1 core ranking, your 1 core submission might end up receiving global points.

Same logic applies to processors. wPrime would have a 1 core, 2 core, etc ranking, but you will only receive points for the submission that would receive the highest points.

Benefits:
+ very fair and competetive rankings, based on amount of cores (1, 2, 3, ...)
+ no expensive hardware needed if you want to get first in the single/dual core ranking
+ no global points grinding possible as you will only get points for your best submission, regardless of the amount of points

The worldwide records page will still be based on the highest score, unregardless of how many cores where used.

2. More hardware points for extremely competitive hardware
In hwboints rev2 only the top 20 members get points for having the fastest hardware in it's category, even if there are 1000 people using this hardware. This is unrewarding if you managed to get in the top 20 percentile, and still do not recieve any points.

More points will be rewarded to more than 20 people when hardware is used by more than 100 overclockers. Less points will be rewarded if hardware is used by less than 10 people (1 instead of 2 points for first place).

Benefits:
+ If you compete in a very competetive hardware ranking, and you have a great, but not top 20 oc, you will still earn some points for your hard work.

3. Consistency in calculating how many people participate in rankings

The global and hardware points are based on how many people participate in the rankings. Eg. if only 10 people would participate in 3DMark01, you would only get a few points for being first. However, if 1000 people would participate, you would get a lot more. The same counts for hardware points, except we made the mistake of counting the amount of submissions, instead of participants. For hardware points, we also want to change it to how many people participate, instead of how many submissions.

Benefits:
+ consequent behavior, more fair

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 20, 2009

Yes & yes.

Comment from Thor941 (JMax-Hardware) - link
Time: Nov 20, 2009

When is it in place?

January?

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 20, 2009

HWBOT version 3 (currently 2.1) will be released first week of January. It will include hwboints version 3, and lot's of other structural changes.

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 20, 2009

My wife will go on vacation without me. :(

FREEEEEDOM! :D

Comment from knopflerbruce (PURE) - link
Time: Nov 20, 2009

richba5tard said: Thank you for all the feedback! After thorough consideration, we are currently basing hwboints rev3 on the following concept:

1. Split up of rankings based on number of cores for both videocards and processors.
Currently we have single socket and multisocket rankings for videocards, and one ranking for processors. In rev3, we will no longer base the ranking on the amount of sockets, but the total amount of cores. To avoid the problem that you would have to participate in 8 (eg.) 3DMark05 rankings, members will only recieve "global points" once for each benchmark application: only for their highest ranked submission regardless of the amount of cores.

Eg. if a user is 3rd with 2x HD4780, 2nd with HD4780X2, 4th with 1x GTX285 and and 5th with 4x GTX270 in the global ranking of 3DMark01, he will only receive hwboints for being 2nd place in the 2 cores ranking (HD4780X2).(*) All his submissions are applicable for hardware points.

(*) Actually, this would only be true of there would be equal amount of participants in each ranking. If there are a lot less participants in the 2 cores ranking than the 1 core ranking, your 1 core submission might end up receiving global points.

Same logic applies to processors. wPrime would have a 1 core, 2 core, etc ranking, but you will only receive points for the submission that would receive the highest points.

Benefits:
+ very fair and competetive rankings, based on amount of cores (1, 2, 3, ...)
+ no expensive hardware needed if you want to get first in the single/dual core ranking
+ no global points grinding possible as you will only get points for your best submission, regardless of the amount of points

The worldwide records page will still be based on the highest score, unregardless of how many cores where used.

2. More hardware points for extremely competitive hardware
In hwboints rev2 only the top 20 members get points for having the fastest hardware in it's category, even if there are 1000 people using this hardware. This is unrewarding if you managed to get in the top 20 percentile, and still do not recieve any points.

More points will be rewarded to more than 20 people when hardware is used by more than 100 overclockers. Less points will be rewarded if hardware is used by less than 10 people (1 instead of 2 points for first place).

Benefits:
+ If you compete in a very competetive hardware ranking, and you have a great, but not top 20 oc, you will still earn some points for your hard work.

3. Consistency in calculating how many people participate in rankings

The global and hardware points are based on how many people participate in the rankings. Eg. if only 10 people would participate in 3DMark01, you would only get a few points for being first. However, if 1000 people would participate, you would get a lot more. The same counts for hardware points, except we made the mistake of counting the amount of submissions, instead of participants. For hardware points, we also want to change it to how many people participate, instead of how many submissions.

Benefits:
+ consequent behavior, more fair

I assume the GPU submission bug is included in that third point?

It would be nice to see what kind of boint vs number of submissions list you're thinking of. I quite agree that when I run these old Opteron server chips on air while watching TV, untweaked and more or less without any effort at all I shouldn't be rewarded with a gazillion boints, but in the moment you get competition I think you should raise it more or less back to norwal levels. Like 2 submissions => 1.5 boints, 3 submissions => 2 boints, and then keep the old system for ranks with more than 3 submissions.

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 20, 2009

I think he means the v2 system does not take into account that there is a single and multiple sockets ranking for gpu related benches, so even if there is little competition in the dual socket ranking, you will get a lot of points.

Yes, it will be fixed too.

Comment from knopflerbruce (PURE) - link
Time: Nov 20, 2009

Is it possible to calculate how the HW masters rank will look with the new rules?:p

GPU bug fix = I'll gain alot (because all others bench 3d more than me). 2 => 1 boint for gold cups in no competition categories = I'll lose alot as I've got more of that sort of scores than anyone.

Comment from knopflerbruce (PURE) - link
Time: Nov 20, 2009

HW boint cap won't be raised this time... or?

Comment from PeterStoba (Bench Tec UK) - link
Time: Nov 20, 2009

Will we all lose points when this comes into affect if we've benches cards with only a few results or will it apply from January onwards?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 20, 2009

PeterStoba said: Will we all lose points when this comes into affect if we've benches cards with only a few results or will it apply from January onwards?

The loss in points will not be 'just' a loss in points. All the VGA rankings' popularity weight will decrease, everyone will lose points.

Comment from PeterStoba (Bench Tec UK) - link
Time: Nov 20, 2009

Yeah, I understand that all people will loose points if it comes into play, was just wondering if it would apply to current scores or only new ones, was a stupid quiestion really.

Comment from Neur0mancer (OC Forums) - link
Time: Nov 20, 2009

I say leave it alone. Separating the points system does not help the poor overclocker. (I should know I am one)
Why is RAM disk not allowed (using software to run physical hardware) not allowed when Hyperthreading is? (using firmware to tell software that there is more cores than there really is?)

Hyperthreading helps in a most of the 3D benchmarks as well as wPrime. (I am running i7 now :) too)

I would love to see a vantage eXtreme section, but not everyone can afford the money to pick up a monitor capable of running it. I would have to go dig out my $200 monitor in order to run it :) (CRT baby!) But I would :)

If you split dual GPU cards from their current placing. You will then have to split dual gpu from multi card also. Otherwise all the 5980, 295, 4870X2 etc, will be dominated by people running 2 of them.

Does not make sense to me.

You want to help out poor Overclockers? Raise HW point maximum. That does not punish the well off, since they can buy old hardware too :)

Comment from [UC OC] Karmakazi (PURE) - link
Time: Nov 20, 2009

I am dissapointed to see this go into effect, but hey I guess its going to hurt everyone so it doesnt really matter...

Comment from knopflerbruce (PURE) - link
Time: Nov 20, 2009

TheKarmakazi said: I am dissapointed to see this go into effect, but hey I guess its going to hurt everyone so it doesnt really matter...

Yeah... it'll be a bit strange to see the points dropping, but the fact is that the GPU boints have been much higher than they should've been - say you have 20 results in total for 3dmark01 for some GPU, 10 for single-card and 10 for multi-card. Of these 10, only 5 scores get boints because the other 5 are worse scores from the same users. The fact is that the way it is now, the HWBot engine will think the popularity is 20 for both singlecard and multicard, when it should really be 5 in both cases, rewarding the scores with MUCH more boints than what's correct.

Pretty sure I'd be 2nd in the HW masters rank if this was corrected now, without the other changes:p

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Nov 21, 2009

My teams gonna get killed with the change. :p

Comment from thebanik (OC Alliance) - link
Time: Nov 21, 2009

do we really need such a major revamp????

Comment from Thor941 (JMax-Hardware) - link
Time: Nov 21, 2009

Hi,

I just don't understand one thing :

- is there any interets keeping all the scores we have made with the new version? I mean let's say I have made 4 scores in one benchmark with a 4870, I always keep the best one on the bot and delete the other ones myself.

Hope I am clear enough ...

Comment from komadyret (411 Overkill) - link
Time: Nov 21, 2009

Looks good to me.
It'll certainly hurt me, but it looks fair at first glance, much better that the current setup.
And with a little tweaking over the next few months, I'm sure this will be streamlined and well working. And it's good to see cpu getting back some of the importance, as gpu-globals will be more balanced.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 21, 2009

thebanik said: do we really need such a major revamp????

Revision 2 was designed back in 2007, the hardware scene has changed quite a lot of these 2 years, multi-gpu and multi-cpu wasn't taken into account.

Comment from proste71 (United Team of Poland) - link
Time: Nov 21, 2009

richba5tard said:

2. More hardware points for extremely competitive hardware
In hwboints rev2 only the top 20 members get points for having the fastest hardware in it's category, even if there are 1000 people using this hardware. This is unrewarding if you managed to get in the top 20 percentile, and still do not recieve any points.

More points will be rewarded to more than 20 people when hardware is used by more than 100 overclockers. Less points will be rewarded if hardware is used by less than 10 people (1 instead of 2 points for first place).

Benefits:
+ If you compete in a very competetive hardware ranking, and you have a great, but not top 20 oc, you will still earn some points for your hard work.

3. Consistency in calculating how many people participate in rankings

The global and hardware points are based on how many people participate in the rankings. Eg. if only 10 people would participate in 3DMark01, you would only get a few points for being first. However, if 1000 people would participate, you would get a lot more. The same counts for hardware points, except we made the mistake of counting the amount of submissions, instead of participants. For hardware points, we also want to change it to how many people participate, instead of how many submissions.

Benefits:
+ consequent behavior, more fair

very well, fair anough I think ;)

Unfortunatelly i have already deleted many hardware scores not in top 20 :(
Big mistake...I could earn some points....

Comment from rbuass (Team Adrenaline) - link
Time: Nov 23, 2009

When can we know how will our new scores?
I hope also unfair are corrected by separating the categories of CPUZ (single, dual, quad) cause it is easier to get high scores with a pentium 4 or celeron that with a quad core, Wprime and are also separated by the VGAs number of slots occupied it is not just compete against 4 VGAs X 2 VGAs at same category.
I think everybody is anxious to know the new scores.
Hugs

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

rbuass said: When can we know how will our new scores?
I hope also unfair are corrected by separating the categories of CPUZ (single, dual, quad) cause it is easier to get high scores with a pentium 4 or celeron that with a quad core, Wprime and are also separated by the VGAs number of slots occupied it is not just compete against 4 VGAs X 2 VGAs at same category.
I think everybody is anxious to know the new scores.
Hugs

The multi-GPU categories will be changed a bit.

The "Cedar Mill-problem" you're talking about isn't really a problem. Sure, you will get a higher frequency with those CPUs, but you will also get better scores in ANY other test with an i7 than with those cellys. Which is the same thing. This is why we have hardware boints, too;) If you use technology that can't compete with whatever is on the top of the global ranks, you can fight for those boints instead. Just like I can't get 100 global boints no matter how good I tweak my AMD K8 superpi runs:p

Socket separation will be implemented too, AFAIK - but ONLY for multithreaded benchmarks like wprime. Maybe PCMark05 as well?

Comment from K404 - link
Time: Nov 23, 2009

1. Split up of rankings based on number of cores for both videocards and processors.
Currently we have single socket and multisocket rankings for videocards, and one ranking for processors. In rev3, we will no longer base the ranking on the amount of sockets, but the total amount of cores. To avoid the problem that you would have to participate in 8 (eg.) 3DMark05 rankings, members will only recieve "global points" once for each benchmark application: only for their highest ranked submission regardless of the amount of cores.

Eg. if a user is 3rd with 2x HD4780, 2nd with HD4780X2, 4th with 1x GTX285 and and 5th with 4x GTX270 in the global ranking of 3DMark01, he will only receive hwboints for being 2nd place in the 2 cores ranking (HD4780X2).(*) All his submissions are applicable for hardware points.

(*) Actually, this would only be true of there would be equal amount of participants in each ranking. If there are a lot less participants in the 2 cores ranking than the 1 core ranking, your 1 core submission might end up receiving global points.

Same logic applies to processors. wPrime would have a 1 core, 2 core, etc ranking, but you will only receive points for the submission that would receive the highest points.

Benefits:
+ very fair and competetive rankings, based on amount of cores (1, 2, 3, ...)
+ no expensive hardware needed if you want to get first in the single/dual core ranking
+ no global points grinding possible as you will only get points for your best submission, regardless of the amount of points

The worldwide records page will still be based on the highest score, unregardless of how many cores where used.

I...err.... dont understand this bit.

Isnt it just the same as it is now?.... but the X2 scores are ALL going into the multi-GPU category? If I submit scores with 2x 4780 and 4 "GTX270" I would only get 1 global rank, like it is now.

BUT... this bit: "members will only recieve "global points" once for each benchmark application: only for their highest ranked submission regardless of the amount of cores."

sounds like all the single-core stuff (5870, GTX285) competes directly against all the Tri-SLI, CF-X setups for global... making 3 or 4-way setups the only way to rank globally

Is there still just 1 global leaderboard for multi-GPU configs? We just have the X2 cards in there too? (and X3 if they appear in future)

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

revision 2:
gpu: single socket and multisocket ranking, and you can get points for both single as multisocket
cpu: 1 ranking for all

worldrecords page: any amount of sockets

revision 3:
gpu: 1 core, 2 core, 3 core, ... 8 core ranking. You can get global points only in one ranking (the one that gives you the most points).
cpu: 1 core, 2 core, 3 core, ... ?? core ranking. You can get global points only in one ranking (the one that gives you the most points).

worldrecords page: any amount of cores

You can submit to as many rankings you want, the bot will automatically award points to the submission that's worth the most points. You can get hardware points in each ranking.

This approach has many benefits:
- interesting, fair rankings (comparing apples to apples, single core processors will be interesting to bench, same for single core videocards)
- You don't need to submit to 1x, 2x, 3x, ... ranking to gain more points, as only your best submission will count for points anyway.

Comment from K404 - link
Time: Nov 23, 2009

Got ya! ok, that makes sense :D Thanks RB

In some ways, this will discourage multi-way systems :D

I get it now

There IS gonna be a big drop in some peoples global points!

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

K404 said: Got ya! ok, that makes sense :D Thanks RB
In some ways, this will discourage multi-way systems :D

Yes, but it does make sense for an overclocker league. This site is based on overclocking skills, not finding the supercomputer with the most cores. :)

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

Yes, you can get hardware points in each ranking. If there are only a handfull of people using the same hardware, it won't be much though. Eg if there are less than 5 people using 3xHD5780, you will only get one point... and a golden cup. :)

Comment from K404 - link
Time: Nov 23, 2009

...and theres cheaper ways of getting hardware points :p

Comment from BenchBrothers.de (AwardFabrik) - link
Time: Nov 23, 2009

first I need to say that I'am with that decision,
but I do have one question.

the result will be that everybody with will lose GlobalPoint caused by the fact that with rev.3 every 3DMark will only give 1x GlobalPoints in contras to rev.2 where each 3DMark gave 2x Globalpoints (1x Single- & 1x MultiSocket).

So the Ratio of Global- to HWPoints will change dramatically because the possibility of making HWPoints will rise with the different Core-Number Rankings while the number of rankings where I can get Globalpoints will becomes less.

the 300p limit for the Userranking might be unchanged,
but it could effect the teamrankings alot.

was this the plan or will the amount of Globalpoints I get for one good Score be raised to fix that "problem"?

well - I hope some of you guys know what I'am trying to explain... ;)

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

Good point masterchorch. It is the idea that people who have world records (like Hipro5, K|ngp|n, Andre Yang, SF3D, etc) remain at the top, and don't need to go scouting for hardware points to defend their position.

Comment from BenchBrothers.de (AwardFabrik) - link
Time: Nov 23, 2009

jmke said: more dynamic :)

more dynamic? ;)
well, I hope the cap will not be increased - that would make it even worse...

just tell us dude :D

[QUOTE=richba5tard;40519]Good point masterchorch.[/Quote]

great. u know what i mean :banana:

first it is exactly what you wrote,
the people who push the current hardware must be clearly at the top spots.
but I also think of the the teams and what they earn from their best overclockers.

without eg. increasing the globalpoints,
all teams would earn more from those "oldschool goldgrinders" than from their wellknown highend benchers - and that can't be the idea I think?!

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

masterchorch said: without eg. increasing the globalpoints,
all teams would earn more from those "oldschool goldgrinders" than from their wellknown highend benchers - and that can't be the idea I think?!

While there will be more rankings where you can achieve hardware points, "oldschool goldgrinders" will not get more points (I think), because we will assign less points to unchallenged hardware (eg 1 point for being first if there are only 5 other with that hardware), and more hw points for very challenged hardware (eg 20 points for first Q6600, 8800GT, etc).

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

richba5tard said: While there will be more rankings where you can achieve hardware points, "oldschool goldgrinders" will not get more points (I think), because we will assign less points to unchallenged hardware (eg 1 point for being first if there are only 5 other with that hardware), and more hw points for very challenged hardware (eg 20 points for first Q6600, 8800GT, etc).

Actually it's 1 boint for first place if there are less than 10. But that means 1 boint for 1st even if it's 1,2,3...9? That's what it says in the big post about the changes...

Edit: maybe we should double the weights, or something. I bet alot of people will hate to lose all those boints, even if it happens to everyone.

Comment from BenchBrothers.de (AwardFabrik) - link
Time: Nov 23, 2009

it's not one oldschool score that will earn more points for the owner and the owners team, it's the amount of possible categories to earn hwpoints - and that amount will increase with rev.3.

if you compare Stummerwinter (who is global first) and turrican (who is hwpoint leader), than you see the "problem" which I mean and which is getting worse with hwbot rev.3

Stummerwinters team gets 1829 Point from him (and he is HWBot Leader), while
Turricans team gets 3051 Points from him.

It not about punish the oldschoolbencher or do not appreciate the hugh number of hours they spend for oldschoolbenching,
but I think HWBot need to make a decision about the HW- and GlobalPoint Ratio.

As in my Example it is unfavorable at the moment,
but it is definitely gets worse with Rev.3.

my opinion is that it must worth more to bench current hardware and not to aggravate the proportionality even more.

will there be another discusion if Useraccounts can still be abused as "subteam-accounts" by small groups of people?

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

You think it will be worse based on the fact that there will be more rankings to bench in, but you do not seem to take into account that you get less points if it's an "easy run" (less contestants). Anyway, it's hard to make claims until the rev3 algorithm has been written, we will show the effect before we release rev3.

Maybe we should think about limiting the amount of hw points a user can donate to it's team.

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

masterchorch said: it's not one oldschool score that will earn more points for the owner and the owners team, it's the amount of possible categories to earn hwpoints - and that amount will increase with rev.3.

if you compare the Stummerwinter (who is global first) and turrican (who is hwpoint leader), than you see the "problem" which I mean and which is getting worse with hwbot rev.3

Stummerwinters team gets 1829 Point from him, while
Turricans team gets 3051 Points from him.

It not about punish the oldschoolbencher or do not appreciate the hugh number of hours they spend for oldschoolbenching,
but I think HWBot need to make a decision about the HW- and GlobalPoint Ratio.

As in my Example it is unfavorable at the moment,
but it is definitely gets worse with Rev.3.

my opinion is that it must worth more to bench current hardware and not to aggravate the proportionality even more.

...you can't limit hardware boints for the teams as well, we've already got an insane cap in the global ranks, that's more than enough. IMO a good team should include both types of benchers, and not just get dominated by whatever team that has the most LN2 highend-benchers. Right now we have a decent mix.

Creating a limit that you suggest will remove even more skill from the lowend benching, which is not fair. It takes alot of knowledge to get scores on older gear, too - there's definately skill in other places than the top 100 of the global rank;)

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

richba5tard said: You think it will be worse based on the fact that there will be more rankings to bench in, but you do not seem to take into account that you get less points if it's an "easy run" (less contestants). Anyway, it's hard to make claims until the rev3 algorithm has been written, we will show the effect before we release rev3.

Maybe we should think about limiting the amount of hw points a user can donate to it's team.

LOL. You're not serious?? There are more things than global boints in this place.

Who does no-one acknowledge lowend stuff:( Sounds like everyone want to put pillows under the i7-benchers asses these days...

Comment from BenchBrothers.de (AwardFabrik) - link
Time: Nov 23, 2009

I am just asking,
my aim was maybee to get some algorithms for rev.3 out of you guys,
but if it is not ready yet I do have my answer.
and I know that you guys do have a little time to think of our thoughts and ideas. ;)

knopflerbruce said: ...you can't limit hardware boints for the teams

i wrote nothing about doing so,
I was asking about ideas to solve the problem (which is one in my opinion) that today the teams get more benefit from "grinders" than from highend global top10 benchers - and that "problem" will be reinforced.

all I want is a dialogue with you guys ;)

and what about that 'Useraccounts abused as "subteam-accounts" by small groups of people' thing?

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

I feel like a whore trying to please anyone. :p At least they get payed. : /

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

richba5tard said: I feel like a whore trying to please anyone. :p At least they get payed. : /

If you want to please anyone, let the HWBoint contribution to teams be unlimited. The global cap is already REALLY low (what am I? 260th with 300?). Plus, this new system will lower the amount of HWboints awarded much more than global.

Adding a team limit means you're not "pleasing everyone", then you're just making globals much more important than they need to be.

DON'T create this new boint limit => less work for you. I should be your hero:D

PS: it has been mentioned a million times when people have complained about the global limit that "you can still fight for your team, there is no cap there". Then what?

Edit: I compared the total contribution for the top 5 teams, and got these numbers:

Global: 30508.6
Hardware: 42566.1

...and this is before the changes already mentioned, which should bring them even closer. I don't really see the issue here - sure some folks bench alot of stuff, but I don't think it's right to punish them for that more than by more or less kicking them out of the global rank.

Edit2: You can also award global boints to more than top 5000, there are enough members here that raising that limit to top 1000 can be defended.

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 23, 2009

Massman & me talked a lot about raising the hw cap for members (to sth like 500), but we first need to see what the effect is of the other changes, before we change this.

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

richba5tard said: Massman & me talked a lot about raising the hw cap for members (to sth like 500), but we first need to see what the effect is of the other changes, before we change this.

OK, np.

Is it possible to see how this new hardware boint algoritm may look like, btw? 1st place in a category with no other results = 1 boint, right?

Comment from Monstru (lab501.ro) - link
Time: Nov 23, 2009

I like the HW cap just how it is :D

Comment from proste71 (United Team of Poland) - link
Time: Nov 23, 2009

richba5tard said:
Maybe we should think about limiting the amount of hw points a user can donate to it's team.

I don't think it's a good idea ... what's the reason for ?

Comment from PeterStoba (Bench Tec UK) - link
Time: Nov 23, 2009

richba5tard said:
Maybe we should think about limiting the amount of hw points a user can donate to it's team.

Not sure many people would like that, would change alot of rankings!

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 23, 2009

richba5tard said: Maybe we should think about limiting the amount of hw points a user can donate to it's team.

Not sure whether limiting the amount of points is a good solution. Before we start making the crowd go crazy (I bet any sentence in line of "limit ... or remove ..." will do that, hehe), let's see how the global boints change works out.

For all we know, the 300 boints cap turns out to be perfect :)

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

Massman said: Not sure whether limiting the amount of points is a good solution. Before we start making the crowd go crazy (I bet any sentence in line of "limit ... or remove ..." will do that, hehe), let's see how the global boints change works out.

For all we know, the 300 boints cap turns out to be perfect :)

Since both the globals AND the hwboints go down, I think the 300 cap will be comparable to a higher one with today's algorithm. 500 may even be too high, I have a feeling top 10 won't be much more than that:p

Comment from proste71 (United Team of Poland) - link
Time: Nov 23, 2009

Massman said:

For all we know, the 300 boints cap turns out to be perfect :)

exactly ;)

Comment from proste71 (United Team of Poland) - link
Time: Nov 23, 2009

Massman said: Not sure whether limiting the amount of points is a good solution.

exactly, imo very bad idea

Comment from K404 - link
Time: Nov 23, 2009

Maybe a bit late for more ideas but.....seeing as having hardware points limit as a % of global (self-global or the No.1 spot global?) is being considered...

what about a Teams hardware points being a fixed % of their global?

The better you do...the better ya do.

Would put an emphasis on global benching. Dunno if thats good or bad. Pretty much everything discussed means more LN2.

I think the changes proposed so far are BIG changes. You guys realise that, dont you? :p

Comment from knopflerbruce (PURE) - link
Time: Nov 23, 2009

SInce we all know everyone will lose alot of boints, I think maybe we should compensate for that by changing some constants in the algorithm. Rev 2 looks like this:

* global points (top 5): (50,40,35,30,25 * application weight) * 0.6
* global points (other): ((-20/145) * your rank + (600/29) * application weight) * 0.6
* hardware points (top 5): (5,4…1 * hardware weight) * 2
* application weigth: 0.5 (<500 participants) - 2 (2000 or more participants)
* hardware weigth: 0.2 (1 participant) - 1 (100 or more participants)

IMO, to avoid people getting mad because they lose half their boints or something like that maybe it's a good idea to raise the constants 0.6 and 2 to something higher, maybe 0.9 and 3. it won't affect the relative scores, if both gets multiplied by the same number, and it will the new scores look a bit less depressive.

Comment from bazx (Bench Tec UK) - link
Time: Nov 23, 2009

when do your think the single GPU rank will be ready for use?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 23, 2009

bazx said: when do your think the single GPU rank will be ready for use?

We aim for the big package release early January.

Comment from road-runner (PURE) - link
Time: Nov 23, 2009

jmke said: for team benching I vote against any kind of limits
even the lowest of benchies should count for the team

+1 on against any kind of team limits...

Comment from anvil (Union Francophone Hwbot) - link
Time: Nov 23, 2009

Hi Guys,

Sounds like a great idea to me !

Things had to change, I hope you will eventually change the rules.

Cheers !

Anvil

Comment from Gomeler (PURE) - link
Time: Nov 24, 2009

This thread is under discussion on my home forum and one of our members brought up a great idea. Instead of limiting it to the best global in a benchmark, how about we make it 2? This way it'll still make it interesting to run multiple GPU setups but guys on a budget can still compete.

Comment from Kal-EL (OC Alliance) - link
Time: Nov 24, 2009

richba5tard said:

Maybe we should think about limiting the amount of hw points a user can donate to it's team.

Makes sense to me and certainly fits into the scheme of things we have in place currently. Why cap hardware to 300 individually and then remove the cap on a team scale? Doesn't make sense. The cap should carry over to your team submissions as well.

I think the hardware grinders with 2000+ hardware points should be listed and acknowledged in a league or standings of their own.

Pitting global benchers against benchers with years of hardware under their belts and hw boints upwards of 1500 boints would be a futile endeavor.

I think the fair thing would be to create a league that is geared towards these chaps with tons and tons of hardware benched under their belts. Expecting new teams and new benchers to enter a game that is ruled by extent of your hardware warehouse is unreasonable. We won't get any new blood into this game and foster any new fever.

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Nov 24, 2009

Just wanted to thank Knopflerbruce for sticking up for us "old school" guys. My team stands to take it in the ass with the proposed changes. It hardly seems fair.

Comment from Kal-EL (OC Alliance) - link
Time: Nov 24, 2009

knopflerbruce said: SInce we all know everyone will lose alot of boints, I think maybe we should compensate for that by changing some constants in the algorithm. Rev 2 looks like this:

* global points (top 5): (50,40,35,30,25 * application weight) * 0.6
* global points (other): ((-20/145) * your rank + (600/29) * application weight) * 0.6
* hardware points (top 5): (5,4…1 * hardware weight) * 2
* application weigth: 0.5 (<500 participants) - 2 (2000 or more participants)
* hardware weigth: 0.2 (1 participant) - 1 (100 or more participants)

IMO, to avoid people getting mad because they lose half their boints or something like that maybe it's a good idea to raise the constants 0.6 and 2 to something higher, maybe 0.9 and 3. it won't affect the relative scores, if both gets multiplied by the same number, and it will the new scores look a bit less depressive.

I think if we can stretch a bit further down the totum pole with boints, it'll stir new interest in those first timers that bench their nub pants off and get big fat zero's because their hardware has been benched to death by the rest of the gang. It's in all of our interest to keep people interested and involved in overclocking, not stomp it out. Foster the Fever.

Comment from knopflerbruce (PURE) - link
Time: Nov 24, 2009

jmke said: at one point we thought about giving every score at HWbot "something" in the global ranking; even if it was 0.1 boint, I still think this should be considered:D

+1

I also made a "rookie boint" suggestion for the HW masters rank as well, gve a small number of boints (maybe just one, it's better than nothing) for the first few submissions. Perhaps something that would help folks to get started:)

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 24, 2009

0.1 points for everyone but the lowest oc? To make him feel extra bad? :D

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 24, 2009

Hmm, processors with unlocked cores will be an issue for rev 3. It's unfair to rank an X3 processor in the 3xcores category, it the 4th core is unlocked.

Comment from knopflerbruce (PURE) - link
Time: Nov 24, 2009

richba5tard said: Hmm, processors with unlocked cores will be an issue for rev 3. It's unfair to rank an X3 processor in the 3xcores category, it the 4th core is unlocked.

It will mess up the X2 rank, too. Phenom II X2 550... And the singlecore rank, with the Sempron 140.

Perhaps you can create categories for unlocked chips... like 140 @ x2, and PII 550 @ x3 and PII 550 @ X4.

There will be a few more ranks, but let's not forget most models are locked - this only applies to SOME Athlon II and Phenom II-rankings.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 24, 2009

I don't know. If you buy a dual core, unlock the two extra cores ... you have a performance advantage. Even in 24/7 setups.

Comment from Kal-EL (OC Alliance) - link
Time: Nov 24, 2009

jmke said: at one point we thought about giving every score at HWbot "something" in the global ranking; even if it was 0.1 boint, I still think this should be considered:D

As much as I despise drug dealing pimps, they got it right. First boints on me, when you're hooked, you're gonna pay to get you're next fix and will be doing all kinda nasty tricks (v-mods, thermal mods, 2nd mortages etc) to get you're next fix.

Comment from knopflerbruce (PURE) - link
Time: Nov 24, 2009

jmke said: unlock extra pipelines, you have a performance advantage. Even in 24/7 setups.
you might have misread "unlock" with "overclocked"; nothing talking about that;)

those who unlocked Radeon 9500 en Geforce 6200 rule the roost in those benchmarks; same thing happening with those CPUs with unlocked cores
the principle is the same; manufacturer disables cores/pipelines, either because they failed to pass quality check, or they need more products in a certain SKU; either way, unlocking them is part of the OC experience; either you are lucky and you get working cores, working pipelines, or you are out of luck and they don't. No different from having luck with CPU/GPU that clocks high or doesn't clock at all.

I don't see how is an issue for HWBoint rev3.
Dual Core Athlon X2 with unlocked cores already rules the ranking there too
in rev3 it will also rule that same category.

there won't be a one core Athlon X2 ranking, nor three core Athlon X2 ranking.

no it's not. be consistent. If you wanted this to be FAIR, unlocking pipelines and cores should have been BANNED from the start. don't change the rules halfway through the game; we have always allowed unlocking pipelines and cores; rev 3 doesn't change any of this ranking; sure in global 3 core ranking, the UNLOCKED core CPUs will have an advantage, same advantage as having a good clocking CPU, supersweet LN2 install, good BIOS, vmodded motherboard, custom PSU etc etc. If you want to rank high, you'll have to get lucky with HW sample

The difference here is that we don't divide the ranks by pipelines.

Is it so hard to make separate rankings? We're talking about 18 new ones as of today (Sempron 140, Athlon 5000+, X3 400e, 405e, 425, 430 Phenom X3 700e, 705e, 710, 720, 720 BE, 740 BE, 545, 550 and 555 (last 3 needs two new each because they can be unlocked to 3 or 4 cores). This is an easy fix to this problem, no "core questions" will be asked then.

Edit: not even sure all those Athlon II x3's come with Deneb core in the first place

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 24, 2009

jmke said: it doesn't "mess the ranking" any more than unlocked VGA does RB.

Yes it does, as an unlocked VGA (eg Radeon 9500 non pro to 9700 pro) only causes a mixup in the 9500 non pro ranking, not in the whole single core gpu ranking. That's a huge difference.

I'm in favor of making seperate rankings for unlocked processors. Eg both Sempron 140 (=> goes in 1 core ranking) and Sempron 140 (unlocked) (=> goes in 2 core ranking).

Comment from knopflerbruce (PURE) - link
Time: Nov 24, 2009

jmke said: there should be a minimum amount of submissions of a separate hardware before it should get its own category, in other cases it should be grouped in a family or generation ranking.... future plans :D

This will only cause people to post in the wrong categories.

I wonder where I'll post my FX 64 scores, then. Higher binned than FX 62, so not fair to put it there... and there should ALWAYS be a category where you can post scores of a CPU here at HWBot:) This is a database, too - and then you need to compare apples to apples, and not mushrooms or whatever.

Comment from Oj0 (PURE) - link
Time: Nov 24, 2009

knopflerbruce said: I wonder where I'll post my FX 64

ATT Jonathan Horne
32 Briga... :D :D :D

Comment from K404 - link
Time: Nov 24, 2009

jmke said: there should be a minimum amount of submissions of a separate hardware before it should get its own category, in other cases it should be grouped in a family or generation ranking.... future plans :D

Thats getting close to saying that points should only be awarded once theres a certain number of submissions.

Cant offer 0.1 points for every submission to encourage people to bench then deny them the category if its not popular enough :p

Comment from knopflerbruce (PURE) - link
Time: Nov 24, 2009

jmke said: not really no:) since it would go into Athlon FX category based on socket xxx
if there are more than xxx entries it will be automatically added to a separate category

anyway, it's doable, but requires some major overhaul
it would allow us to reward the best Athlon S939 CPU , the best S462 Athlon XP CPU, the best Athlon CPU, the best AMD CPU, etc

How exactly do you want to automatically add the new CPU? Like I post it in the FX-62 category, then when there are enough results suddenly HWBot "Knows" that it's indeed an FX-64? How do you want to do that?

Awarding the best CPU pr socket is still doable I think, you just need to add some socket ID to the database.

Comment from Dragon Oc Team (HwMania-OC-Team - HwMania.org) - link
Time: Nov 24, 2009

this ranking is very interesting, but the top result may gain something about 300/ 400 points

Comment from knopflerbruce (PURE) - link
Time: Nov 24, 2009

jmke said: ah, the miracles of coding :)
HWbot will have all the different models listed; you'll be able to post under them; but they won't be their own category until they meet a certain quota.

we'd have more than two rankings and points awarded.
- Global HWboints
- Global Brand HWboints (AMD/NVIDIA/ATI/INTEL/...)
- Global Model Range HWboints (Radeon/Geforce/Quadro/Athlon/Geode/...)
- Hardware Family HWboints (Athlon XP, Athlon 64 , Core 2 Duo, COre i5, Radeon 9 series, Geforce 8xx series)
- Hardware Model HWboints (3100+, 8800 GT, FX-64, ...)

yes it's a lot of extra's; not they won't each scale to 100 HWboints per score; yes it will require a complete restructuring of the hardware database; the HWBoint system and HWbot, it won't be there for Rev4 either I think.

btw, this is not my suggestion alone, Massman is also a strong follower of this train of though. More competition, less grinding, keeping bench gear affordable, rewarding scores which are awesome, even if they are done with hardware that doesn't rank high globally.

anyway;.... future music :D

practical example:

now don't tell me you wouldn't like to get a bit recognition for that FX on Ice? :)

I like the thought (as it shows that skill can be found in other areas than the newest tech, by rewarding boints for it), but if you have HW family boints and HW model boints, how can this be anti-grinder-adjustments?

...I would also like to point out that without the grinders there would be a LOT less decent results with alot of HW, we're the ones pushing old tech to new heights:)

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Nov 25, 2009

I would also like to point out that without the grinders there would be a LOT less decent results with alot of HW, we're the ones pushing old tech to new heights:)

+1
I'd also like to know where the line will be drawn on what you consider "current tech"? In a year or two when i7 is just another memory, will all the i7 HW points drop to 1 also?
My team is based on old tech, hence the name Classicplatforms. If you're making it so we can't be competitive here, tell me now, so we don't waste any more of each others time.

Comment from Stone_age (Classicplatforms.com) - link
Time: Nov 25, 2009

Might I suggest you sell some of your code so that someone can start another similar database for the masses who don't have the deep pockets or sponsorship to compete here after all this is implemented?

Comment from Gomeler (PURE) - link
Time: Nov 25, 2009

Sounds to me like you guys are trying to turn hwbot into a "everyone is a winner" platform. Guess we'll see how Rev3 works and see if people flock elsewhere.

Comment from road-runner (PURE) - link
Time: Nov 25, 2009

Looks like someone needs to open up another website similar to this one with different rules...

Comment from Gunslinger (PURE) - link
Time: Nov 25, 2009

How about this idea>>>
highestclocked.besthardware.wins.org

Comment from Stone_age (Classicplatforms.com) - link
Time: Nov 25, 2009

You guy's from PURE are ranked #1, and rightfully so. These new rules could cause knopflerbruce to lose over 1000 boints, which could hurt your team. They already killed ours. Everyone quit benching after reading this thread and are already talking about selling all their hardware, and the future of our website is very bleak also, as the competition here was the glue that was binding our gang together.
I know the topic for this thread is Opinions please, but everyone knows what they say about opinions....

It's been fun.

Comment from road-runner (PURE) - link
Time: Nov 25, 2009

I am sure not in a hurry to buy anything else, looks like its time to sell...

Comment from Barton (Classicplatforms.com) - link
Time: Nov 25, 2009

Hey admins and mods - RB, JMKE, Turrican, Massman, and others,

I must speak against the proposed "new rules". Not just for the obvious reason that they will kill the small team that I'm on, but also because it appears to me that the new rules will be counter-productive.

IMO, the new rules will discourage, rather than encourage, people from become active participants on HWBot. Under the "old rules" there was a system of awarding points that allowed everyone a fair chance to compete on a reasonably equal basis, or as we would call it in my country, a "level playing field".

To change the rules now in the manner proposed is simply not fair. Those of us with older hardware and with limited financial resources, will lose any chance to compete with those members with more money who can afford $500 video cards, $400 processors, $350 motherboards, $300 ram kits, and unlimited amounts for upgrades and extreme cooling.

Under the existing rules, members with limited incomes can compete with a chance to score a reasonable amount of total points by being both skilled and persistent, even if we are not wealthy.

Under the new rules, we will lose many of our hard won accumulated points and will have no ability to compete in any reasonable manner in the future. Hence many of us are likely to permanently leave the site. That's not something I want to do, but it is something that is very likely once the reason for being an active member here is eliminated.

All that said, let me ask you to abandon the proposed new rules and keep the existing rules that now apply to HWBot competitions. The present rules are not broken; there is no need to change them.

Thanks for listening. I hope you read this. Let us know what you decide.

Comment from zila1 (Classicplatforms.com) - link
Time: Nov 25, 2009

You're gonna lose a lot of good folks over this boys. I may not compete all that much but I do like to come here and watch the extraordinary talents of the guy with the older rigs that can really make them fly and this has been a great place for those fellas to exhibit those skills.

Your "new rules" will knock them out of contention which doesn't seem fair at all. The only guys left will be the ones who can afford top of the line parts on LN2. I don't know too many folks that can keep up with a hardware change every few months or so just to remain competitive.

That having been said, your new rules are going to force folks that really want to be able to compete here to leave and take the competition elsewhere.

You've got to ask yourself, is it really worth it?

I don't mean to say anything wrong but that's just my two cents.

I really hope it all works out for the better and the great competition here can continue.

Comment from ZorchThatCPU (Overclockt) - link
Time: Nov 25, 2009

I have to agree with the guys from CP. I have no intention of going LN2, or any intention of building an i7 rig just to jeopardize it with extreme mods and cooling. That's throwing money out the window. And with all due respect to the heavy hitters who do use LN2 and other extreme cooling, I don't think that makes them MORE talented. I think it removes the need for having talent. It lets the hardware do the work instead of the person doing the overclocking.

In any case, my favorite platform to bench has always been 754. Who cares about 754, right? Well, I do. And what I do is no less difficult or interesting than what guys do with their $3000 rigs, in my opinion and the opinion of a lot of others.

Anyway, I'm not looking to step on anyone's toes, nor to offend. But the people with outrageous amounts of money already have the Boints slanted well in their favor, so that their rewards for the same amounts of effort are much greater. I personally see no need to further marginalize those of use who favor cheaper (but just as entertaining) systems. There's no reason not to encourage people to continue breaking records in Socket A into the foreseeable future, because it's fun.

Basically, the new rules would eliminate a vast segment of your fan base, while adding no one to replace them. You will not attract large masses of people who want to drop thousands of dollars on a bench rig, because the demographic is already tapped out.

Thanks for hearing my two cents worth.

Comment from Dragon Oc Team (HwMania-OC-Team - HwMania.org) - link
Time: Nov 25, 2009

i like new rules very much because we can make point just with a 5870, and we don't need 3 of them

Comment from Oj0 (PURE) - link
Time: Nov 25, 2009

Should read all posts and not just the comments

Comment from Oj0 (PURE) - link
Time: Nov 25, 2009

Apologies :)

Comment from Kal-EL (OC Alliance) - link
Time: Nov 25, 2009

ZorchThatCPU said: I have to agree with the guys from CP. I have no intention of going LN2, or any intention of building an i7 rig just to jeopardize it with extreme mods and cooling. That's throwing money out the window. And with all due respect to the heavy hitters who do use LN2 and other extreme cooling, I don't think that makes them MORE talented. I think it removes the need for having talent. It lets the hardware do the work instead of the person doing the overclocking.

Thanks for hearing my two cents worth.

Respectfully, this is a completely innacurrate assessment of extreme cooling overclockers. If anything, having to pour, manage temperatures and key in settings should be viewed as much more challenging and needed a wider skill set than overclocking on ambient air temps.

Those gentlemen at the top of the rankings are extremely skilled. The cooling is just one part of the puzzle. These guys tweak, tweak, tweak and tweak some more, they do voltage modifications, monitor everything imaginable and meticulously pick apart every aspect of a benchmark and system component to bleed out every last bit of performance possible. The exception here is that they must do it with many more variables to the equation. None of these guys have money coming outa there earballs to just rent truckloads of LN2 either. They tune as efficiently and quickly as humanly possible to save scripts for meals.

I think that comparing an overclocker that finds the most obscure un-benched and cheap piece of hardware to one of these guys isn't a valuable comparison.

How can someone who is unwilling or suffers from an insufficient budget have expectations to compete in this extreme league? It's an entirely different animal from the "old" hardware grinders.

It's a complicated dilema to solve. I applaud those involved for making earnest efforts to problem solve in a fair fashion. I surely won't spout off about how I'm gonna take my ball, walk off the field and not play if you don't do it my way.

Comment from knopflerbruce (PURE) - link
Time: Nov 25, 2009

jmke said: there won't be a point cap for HWboints that go towards a team score in Rev3 either.
Quote RB where he says it is a fact that he will implement that? ....

incorrect. that is SOOOOOOO incorrect a statement. What point of "rewarding popular hardware more" are you guys not getting? It will require skill and persistence to accumulate Hardware Points... but here's the change:

- In Rev2 you bench unpopular hardware to get your 2 points/score
- In Rev3, if you bench unpopular hardware, you'll get less points

- in Rev2 if you bench popular "expensive/none expensive" hardware you can earn up to 15 HWboints if you rank in TOP 20.
- in Rev3 if you bench popular "expensive/none expensive" hardware you can earn up A LOT MORE than 15 HWboints if you rank in the top AND it will trickle down much further.

- In Rev2 ONLY the top 20 scores get hardware points, no matter how popular the category is.
- In Rev3 the TOP X will get hardware points, X will be dynamic and will depend on the amount of entries. So in case of that 8800 GTX, instead of only TOP 20; it will more likely be the TOP 100 that get hardware points, and the TOP score will get more HWboints, and not limited to 15 as it is now.

so no.. we are not "killing" anything. We are "fixing" an obvious flaw which is present in the current system: not giving enough points to popular hardware. Rev3 is beneficiary to those not benching expensive hardware.

Yeah, the team HWboint limit will NOT happen.

The lower-boints-for-unbenched-HW-change is fair enough, 2 boints is without a doubt too much. Now it's like this:

(Results - boints awarded today):

1: 2.0, 2: 2.1, 3: 2.1, 4: 2.2, 5: 2.2, 6: 2.3, 7: 2.4, 8: 2.5, 9: 2.5, 10: 2.5, 11: 2.6 12: 2.6, 13: 2.7

With the change:

1: 1.0, 2: 1.0, 3: 1.0, 4: 1.0, 5: 1.0, 6: 1.0, 7: 1.0, 8: 1.0, 9: 1.0, 10: 1.0, 11: 1.1(?), 12: 1.2(?), 13: 1.3(?).

IMO, that "1.0"-sequence is too long. I've been competing in these kind of categories for some time now, and when there are more than 5 results you'd usually have to fight at least a little bit to get the #1 spot. Plus, many (and I mean MANY) categories don't have more than 10 participants, this change will make like half of the ranks 1.0-boint-for-first-place-ranks. Too many. Set the limit to 5, and there will be less complaints. When most ranks have more results, you can always raise it in rev. 4. No need to turn everything upside down.

The underlined part sounds very nice, wasn't aware of that change:)

Comment from [UC OC] Karmakazi (PURE) - link
Time: Nov 25, 2009

I like every proposed change... except.... not getting separate globals for single/multi card. I know "multi" card will now encapsulate three different rankings (dual, triple and quad gfx setups). Wouldnt it be more consistent to get single card globals AND a set of globals in only ONE of the three multi card categories (whichever score is best from those three or whatnot)?

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 25, 2009

knopflerbruce said: Yeah, the team HWboint limit will NOT happen.

The lower-boints-for-unbenched-HW-change is fair enough, 2 boints is without a doubt too much. Now it's like this:

(Results - boints awarded today):

1: 2.0, 2: 2.1, 3: 2.1, 4: 2.2, 5: 2.2, 6: 2.3, 7: 2.4, 8: 2.5, 9: 2.5, 10: 2.5, 11: 2.6 12: 2.6, 13: 2.7

With the change:

1: 1.0, 2: 1.0, 3: 1.0, 4: 1.0, 5: 1.0, 6: 1.0, 7: 1.0, 8: 1.0, 9: 1.0, 10: 1.0, 11: 1.1(?), 12: 1.2(?), 13: 1.3(?).

IMO, that "1.0"-sequence is too long. I've been competing in these kind of categories for some time now, and when there are more than 5 results you'd usually have to fight at least a little bit to get the #1 spot. Plus, many (and I mean MANY) categories don't have more than 10 participants, this change will make like half of the ranks 1.0-boint-for-first-place-ranks. Too many. Set the limit to 5, and there will be less complaints. When most ranks have more results, you can always raise it in rev. 4. No need to turn everything upside down.

The underlined part sounds very nice, wasn't aware of that change:)

This is very good feedback, thank you! The idea is to start giving more points as soon as it gets more competitive. If +- 5 people is the turning point, we should start handing out more hwboints when there are more than 5 participants, not 10.

Anyway, we have started coding on hwbot rev3. We will make a public version available long before the algorithm tuning is final. :)

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 25, 2009

TheKarmakazi said: I like every proposed change... except.... not getting separate globals for single/multi card. I know "multi" card will now encapsulate three different rankings (dual, triple and quad gfx setups). Wouldnt it be more consistent to get single card globals AND a set of globals in only ONE of the three multi card categories (whichever score is best from those three or whatnot)?

I think the benefit of this approach is that it would reflect the current rankings more, not that it's actually a more fair point awarding mechanism. The pointing mechanism would force people to compete in at least two rankings per videocard, instead of just trying to be best in one specific ranking.

And one remark: there would not be a single/dual/triple/quad videocards ranking, but single core, dual core, triple core, ... octa core videocard ranking. # sockets <> # cores :)

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 25, 2009

Oh my god, what happened here?

richba5tard said: Yes it does, as an unlocked VGA (eg Radeon 9500 non pro to 9700 pro) only causes a mixup in the 9500 non pro ranking, not in the whole single core gpu ranking. That's a huge difference.

I'm in favor of making seperate rankings for unlocked processors. Eg both Sempron 140 (=> goes in 1 core ranking) and Sempron 140 (unlocked) (=> goes in 2 core ranking).

You're making this more complex than it should be. If you split up rankings for the unlockable cpu's, people will be using ONE product to submit in MULTIPLE categories, which should not be the case. Btw, splitting up rankings based on #cores is only necessary for PCMark05 and Wprime. SuperPI and PiFast don't benefit from multi-core designs.

Also, will you be treating hyperthreading as feature or core? If as feature => intel rules the rankings, if as core => hyperthreading is less performant than native dual-core or dual-socket design.

I don't know why people are hating the unlocked cores; it's extra performance you get from the same chip.

jmke said: btw, this is not my suggestion alone, Massman is also a strong follower of this train of though. More competition, less grinding, keeping bench gear affordable, rewarding scores which are awesome, even if they are done with hardware that doesn't rank high globally.

anyway;.... future music :D

I was, untill I realised the amount of coding and work that would go into all this. Making sure that every single product is in the right core/family/subfamily, balacing the points and so on.

Future music indeed, no ETA-music :).

Stone_age said: Might I suggest you sell some of your code so that someone can start another similar database for the masses who don't have the deep pockets or sponsorship to compete here after all this is implemented?

Yes, you may suggest, no, we won't sell the code.

Before you start panicking, please have a bit of patience. Don't jump to conclusions based on what you think the ranking will be. We'll run a test-server to see what kind of effect this revision has.

By the way, this new revision is ALL for those who don't have deep pockets. Just re-read the part where single GPU's are now ranked seperatelly.

Gomeler said: Sounds to me like you guys are trying to turn hwbot into a "everyone is a winner" platform. Guess we'll see how Rev3 works and see if people flock elsewhere.

Sounds like we want to make it more about competition, less about '$$-and-boint'.

GunslingerOCS said: How about this idea>>>
highestclocked.besthardware.wins.org

Yes. Over-clock-ing. Clock ~ clock frequency ... over ~ higher must be better. Higher clocks must be better.

Stone_age said: They already killed ours. Everyone quit benching after reading this thread and are already talking about selling all their hardware, and the future of our website is very bleak also, as the competition here was the glue that was binding our gang together.
I know the topic for this thread is Opinions please, but everyone knows what they say about opinions...

Djeez.

1) thanks for referring to RB and me as 'they'
2) who killed what now? The revision hasn't even been coded yet. Talk about jumping to conclusions
3) If you feel that your opinion is not taken into account, and thus of no importance, don't comment. We're reading every single comment here from the absolute beginning.

Comment from Barton (Classicplatforms.com) - link
Time: Nov 25, 2009

Massman, the biggest problem is changing the long established rule where the minimum point count for the top score in any bench is now 2.0 and scaled from that.

Dropping that to 1.0 just doesn't seem fair. The guys who bench newer, more popular and more expensive hardware already earn many times the points we do.

To cut the points of those of us who earn the least to start with just isn't, IMO, fair play.

Comment from Stone_age (Classicplatforms.com) - link
Time: Nov 25, 2009

HawaiiSuperman said:

How can someone who is unwilling or suffers from an insufficient budget have expectations to compete in this extreme league? It's an entirely different animal from the "old" hardware grinders.

It's a complicated dilema to solve. I applaud those involved for making earnest efforts to problem solve in a fair fashion. I surely won't spout off about how I'm gonna take my ball, walk off the field and not play if you don't do it my way.

I don't think anyone is taking their ball and walking off the field. Depending on the final definition of "Popular" the ball may be getting pulled from our hands, and we may be getting kicked off the field. When I first visited HWbot, I thought to myself: wow, I can pull out all this old hardware destined to be recycled, and have some fun with it. I never expected then or now to compete with the extreme overclockers. I am not in they're league nor do I have any illusions of ever being in they're league.
I myself am financially challenged. The cost of a basic Core i7 system to compete in the league here, not including cooling, represents more than 1 month of income. That's something I, and the majority of my team cannot afford to do.

So, if the rule change means we can no longer participate here, then so be it.

Regardless of the outcome, It has been fun.

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 25, 2009

Barton said: Dropping that to 1.0 just doesn't seem fair. The guys who bench newer, more popular and more expensive hardware already earn many times the points we do.

Sorry, this is an incorrect statement. If you bench newer, popular hardware and you end up being 21nd out of 150 people, you get 0 points. While if you bench an Intel Pentium 60 in PCMark05, you get 2 points without overclocking. How is this fair?

You are correct that if you buy a very expensive dual gpu, single socket videocard, you can achieve quite some points in the single socket gpu ranking. But this is a major issue we tackle in rev3.

Comment from knopflerbruce (PURE) - link
Time: Nov 25, 2009

Massman said: Oh my god, what happened here?
You're making this more complex than it should be. If you split up rankings for the unlockable cpu's, people will be using ONE product to submit in MULTIPLE categories, which should not be the case. Btw, splitting up rankings based on #cores is only necessary for PCMark05 and Wprime. SuperPI and PiFast don't benefit from multi-core designs.

Also, will you be treating hyperthreading as feature or core? If as feature => intel rules the rankings, if as core => hyperthreading is less performant than native dual-core or dual-socket design.

I don't know why people are hating the unlocked cores; it's extra performance you get from the same chip.

Not splitting by number of active cores will result in quad core domination in dual and tri-core ranks... the downside by having multiple categories for the same model is as you say - you can submit more scores with the same CPU.

But... there are very few CPUs available that allows you to do exactly that. IMO a rank based on active cores will be very "clean", which is more important than those few who get a few more boints from that little loophole.

Hyperthreading => difficult decision. I would lean towards counting threads and not cores, as HT is performance wise closer to adding a core, than to the performance without HT. But this will make the loophole alot bigger, so maybe it's better to count cores after all?

...on the other hand, there may be some interesting fights between Phenom dual cores and i3/i5/i7 w/HT if you just count number of cores enabled by default.

Comment from richba5tard (hwbot.org benching team) - link
Time: Nov 25, 2009

Stone_age said: When I first visited HWbot, I thought to myself: wow, I can pull out all this old hardware destined to be recycled, and have some fun with it. I never expected then or now to compete with the extreme overclockers. I am not in they're league nor do I have any illusions of ever being in they're league.

I don't get why you think you can no longer compete with cheap, old hardware in rev3? I'm a socket A bencher myself, and I don't see why rev3 would be worse. If anything it's better, because the single cpu core ranking would be more interesting.

Comment from knopflerbruce (PURE) - link
Time: Nov 25, 2009

Barton said: Massman, the biggest problem is changing the long established rule where the minimum point count for the top score in any bench is now 2.0 and scaled from that.

Dropping that to 1.0 just doesn't seem fair. The guys who bench newer, more popular and more expensive hardware already earn many times the points we do.

To cut the points of those of us who earn the least to start with just isn't, IMO, fair play.

The effort needed to gain 17 boints from a rarely benched category is alot less than what you need to do to be able to fight for 18th place in some popular categories of some other piece of HW.

...but maybe HWbot is overcompensating for this? We will see when the test server is up;)

Comment from der8auer (PC Games Hardware) - link
Time: Nov 25, 2009

Okay, i understand the new global rankings of rev3 and im a bit pissed of but all will lose points and it makes it more competitiveso thats not that tragic...

But i didn't understand one thing about the hardware ranking:
Will i earn hwboints for scores with 1x GTX295 and hwboints for 2x GTX295? Just how it is at the moment?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 25, 2009

der8auer said: Will i earn hwboints for scores with 1x GTX295 and hwboints for 2x GTX295? Just how it is at the moment?

rev3 would be points for [max(1 x gtx295; 2 x gtx295)]

Comment from knopflerbruce (PURE) - link
Time: Nov 25, 2009

Massman said: rev3 would be points for [max(1 x gtx295; 2 x gtx295)]

Does this apply to hardware boints as well? Not just global?

Comment from der8auer (PC Games Hardware) - link
Time: Nov 25, 2009

Thats what i still don't understand :D

Comment from Ol'Bud (Classicplatforms.com) - link
Time: Nov 25, 2009

You all may need to remind yourselves that the data base you are building does not depend entirely on the members ability to purchase the newest hardware components to bench and report results.

It is mostly based on the members understanding of the mathematical relationships between the cpu the memory and I.C.'s on the motherboards and their persistence and patience to continue pushing hardware beyond manufacturing specifications.

Penalizing any members by lowering the accumulated scores would show a nefarious side to any website that engaged in such practices.

I hope the owner/admin here has clear thinking and abilities of issue spotting and problem solving and does not make for a lose-lose situation to occur here.

Comment from Stone_age (Classicplatforms.com) - link
Time: Nov 25, 2009

Massman said:

Djeez.

1) thanks for referring to RB and me as 'they'
2) who killed what now? The revision hasn't even been coded yet. Talk about jumping to conclusions
3) If you feel that your opinion is not taken into account, and thus of no importance, don't comment. We're reading every single comment here from the absolute beginning.

I was not referring to any one person, I was referring to the rules. I apologize to both RB and you. I have nothing but respect for the whole HWbot team and your coding skills. This is a fantastic site unlike any other in the world.
But, you cannot possibly take everyone's opinion into account. The bottom line is you must do as you must for the greater good. And if the greater good spells the end for teams like ours, then we'll have to accept that and move on.

Comment from knopflerbruce (PURE) - link
Time: Nov 25, 2009

Stone_age said: I was not referring to any one person, I was referring to the rules. I apologize to both RB and you. I have nothing but respect for the whole HWbot team and your coding skills. This is a fantastic site unlike any other in the world.
But, you cannot possibly take everyone's opinion into account. The bottom line is you must do as you must for the greater good. And if the greater good spells the end for teams like ours, then we'll have to accept that and move on.

No need to end anything...

Comment from Stone_age (Classicplatforms.com) - link
Time: Nov 25, 2009

Thanks for that final answer jmke.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 25, 2009

It's not like we are removing hardware points? If anything, we're increasing the chance for points as we're aiming for a 0.1p minimum. I really fail to see how you can make such giant statements based on almost no information about how it affects the ranking; the complexity of the HWBoints algorythm makes it very, very difficult to predict the actual loss of points, let alone predicting 'the end of a team'.

knopflerbruce said: Does this apply to hardware boints as well? Not just global?

Only at the level of global points. So, only to affect the Overclockers League. That's why the hardware points distribution may have to be altered; but that's something we need to decide when we see the new revision in practice.

der8auer said: Thats what i still don't understand :D

Ok, an example:

VGA: GTX285
Benchmark: 3Dmark03

your scores:

Single gpu: 1st = 100 points
Dual gpu (2xGTX285): 50th = 20 points
Triple gpu (3xGTX285): 30th = 40 points
Quad gpu (4xGTX285): 10th = 80 points.

=> Single GPU score gets most points => This is the points you get for 3DMark03 global.

Comment from K404 - link
Time: Nov 25, 2009

Im not enirely sure all of what im about to say will flow properly but.. try and stick with it :D

In the proposed Rev3, the points awarded for hardware that isnt popular will drop. It doesnt matter if its old or new. Popular hardware= well rewarded for good scores.

If some guys are benching unpopular hardware, there is less competition. You might not need to OC much or maybe even at all to rank well.

Why SHOULD you be well awarded for those scores? If you want to show that benching older gear is fun, start spreading the word. Other guys will be in a similar boat no matter what country they're in. The better you show old gear to be fun, the more guys will pick it up, the more scores will come in, the more points the good scores will get.

Again, I gotta question why you choose to bench this stuff? Companies are less likely to take notice- they cant use it for PR purposes. You aren't aiming at World No.1 spots. You do it for fun and for rankings? The points, even now, are non-existent (comparitively) the carrot dangling in front of you is the fun and satisfaction? Who cares how many points are awarded for what you do. Theres a ranking and maybe silverware available.

If the category is popular, Its totally possible the points awarded wont change much. If its not popular, why did you bother benching it in the first place?

This has nothing to do with old Vs new. Its popular Vs unpopular. Did anyone ever feel euphoric for winning a race where no-one else was taking part? Would you expect to be offered a job just because you turned up for interview?

Comment from der8auer (PC Games Hardware) - link
Time: Nov 25, 2009

Massman said:
Ok, an example:

VGA: GTX285
Benchmark: 3Dmark03

your scores:

Single gpu: 1st = 100 points
Dual gpu (2xGTX285): 50th = 20 points
Triple gpu (3xGTX285): 30th = 40 points
Quad gpu (4xGTX285): 10th = 80 points.

=> Single GPU score gets most points => This is the points you get for 3DMark03 global.

Ah okay now i understand it :)

Thanks for your explanation.

Comment from ZorchThatCPU (Overclockt) - link
Time: Nov 25, 2009

HawaiiSuperman said: Respectfully, this is a completely innacurrate assessment of extreme cooling overclockers. If anything, having to pour, manage temperatures and key in settings should be viewed as much more challenging and needed a wider skill set than overclocking on ambient air temps.

Those gentlemen at the top of the rankings are extremely skilled. The cooling is just one part of the puzzle. These guys tweak, tweak, tweak and tweak some more, they do voltage modifications, monitor everything imaginable and meticulously pick apart every aspect of a benchmark and system component to bleed out every last bit of performance possible. The exception here is that they must do it with many more variables to the equation. None of these guys have money coming outa there earballs to just rent truckloads of LN2 either. They tune as efficiently and quickly as humanly possible to save scripts for meals.

I think that comparing an overclocker that finds the most obscure un-benched and cheap piece of hardware to one of these guys isn't a valuable comparison.

How can someone who is unwilling or suffers from an insufficient budget have expectations to compete in this extreme league? It's an entirely different animal from the "old" hardware grinders.

It's a complicated dilema to solve. I applaud those involved for making earnest efforts to problem solve in a fair fashion. I surely won't spout off about how I'm gonna take my ball, walk off the field and not play if you don't do it my way.

In no way am I saying the LN2 crowd aren't excellent at what they do. On the contrary, guys like Andre Yang, KingPin, and KnopflerBruce probably know more about OCing than I ever will bother to learn. But by the same token, I have met up with KB in the 754 categories numerous times, and when his runs weren't on extreme cooling, I've won about as often as I've lost. And believe me, I have nothing but respect for the guy, who also seems very nice in the limited interactions I've had with him. And I would guess that his 754 runs were older or at least are not a priority for him. But it meant a lot to me to be in the same league with him, even if it's the minor leagues.

But anyway, it's just one man's opinion, but those kind of situations seem like the most "true" overclocking competitions... you're on air, I'm on air, and we're using comparable hardware, now let's see who can tweak for the win! In a perfect world, everybody would have to use EXACTLY the same hardware, down to the same case, with identical airflow. The tweaking skill would be the only determining factor. Obviously, that's never going to happen.

But it still seems kind of silly to basically let people "buy" points by using the most expensive hardware. Why not just skip the overclocking and have everybody just post up their bank account balance? Then we could see who would win the extreme overclocking competitions without ever bothering with all that pesky overclocking.

Anyway, I haven't been benching much lately anyway, because I've been busy with other things. And I'm nobody in particular around here. Nobody would particularly miss me if I didn't come around. But since opinions were solicited, I gave mine.

Props to the folks who do the hardcore stuff, but it would be nice to have the accomplishments of everyone else rewarded as well.

Comment from BenchZowner (Hellas Overclocking Team) - link
Time: Nov 25, 2009

For tweaking skills there are clock per clock competitions, limited frequency competitions, etc.

Having all the money in the world won't make you a top overclocker.
Trust me, it isn't as simple as you seem to think it is.

Comment from Bones (Classicplatforms.com) - link
Time: Nov 25, 2009

After reading through this thread, I have to say losing a great deal of points isn't what I'd want to see either since I earned them all the hard way by benching the crap out of my hardware - BSOD's, crashes, ect too numerous to keep count of is the price I've paid to get what I currently have.

Here's a suggestion:
Why not create a seperate "Class" or "Division" for the older hardware that's being affected?

This would allow folks that either want or simply have to use the older stuff to be competitive yet shoudn't affect the guys on LN2 with the newer stuff. This concept is used for several competitive sports such as Drag Racing. You have a "Pro" class, "Pro Modified" class, "Top Fuel" class, ect and these classes can have a points system within themselves similar to what we have now.

Place the hardware that's considered "Old" in these so entries with this hardware will only go in this class and let the points count there where the newer hardware benchers won't be bothered by it. As to who could run, of course if you have it and want to run it, go for it.

What do you guys think?
Could this be a viable solution?

Comment from knopflerbruce (PURE) - link
Time: Nov 25, 2009

Fiddling around with the different constans I mentioned earlier can be done on the test server.

This is a psychological problem, losing boints feels bad even if it's the same for all:p

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Nov 26, 2009

I guess I'm just not understanding. I'll just wait to see how it plays out before I shoot my mouth off any more. I owe you guys that much. The entire staff has been good to us here. I'm not willing to just throw that away on a mis-understanding. Meanwhile, I'll just bench on. Thanks for letting me speak my mind.:)

Comment from Bones (Classicplatforms.com) - link
Time: Nov 26, 2009

jmke said: again. you're not reading this correct.
you'll only lose points if you benched hardware that belongs to a category with a not a lot of competition;

looking at your profile, you bench OLD hardware, but not UNPOPULAR hardware. Thus you stand to GAIN points. Not lose them.

might want to read through the thread again, especially paying attention to WHAT rev3 is about.

Yes, I understand and I can see your point.

I'm content to see how it plays out and go from there guys. I don't envy you having to come up with the changes to make things better as a whole. I was making a possible suggestion that hopefully would give a fresh idea or perhaps a different perspective while explaining my reasons for suggesting it in the first place.

Perhaps it could be something to use for future revisions of the site - Who knows?

You guys have been great and it's appreciated. Looking foward to see how it all goes from here.

Comment from zila1 (Classicplatforms.com) - link
Time: Nov 26, 2009

Thanks for all the responses and taking the trouble to explain it all. It really is appreciated. I too just want to see how it all plays out.

Comment from thebanik (OC Alliance) - link
Time: Nov 26, 2009

Easier said that done???

Though I believe getting hardware points is easy and far more cheaper, hence I also started benching old hardware and got mixed results, but it really takes time and patience and skills to really bench the old hardware and get to medals.

Now talking about the point of points awarded to popular hardware, and since you picked 8800GT, if you may look out for its top 100 results. Even the 100th result is from a SS, so where would the average overclocker go????

jmke said: there won't be a point cap for HWboints that go towards a team score in Rev3 either.
Quote RB where he says it is a fact that he will implement that? ....

incorrect. that is SOOOOOOO incorrect a statement. What point of "rewarding popular hardware more" are you guys not getting? It will require skill and persistence to accumulate Hardware Points... but here's the change:

- In Rev2 you bench unpopular hardware to get your 2 points/score
- In Rev3, if you bench unpopular hardware, you'll get less points

- in Rev2 if you bench popular "expensive/none expensive" hardware you can earn up to 15 HWboints if you rank in TOP 20.
- in Rev3 if you bench popular "expensive/none expensive" hardware you can earn up A LOT MORE than 15 HWboints if you rank in the top AND it will trickle down much further.

- In Rev2 ONLY the top 20 scores get hardware points, no matter how popular the category is.
- In Rev3 the TOP X will get hardware points, X will be dynamic and will depend on the amount of entries. So in case of that 8800 GTX, instead of only TOP 20; it will more likely be the TOP 100 that get hardware points, and the TOP score will get more HWboints, and not limited to 15 as it is now.

so no.. we are not "killing" anything. We are "fixing" an obvious flaw which is present in the current system: not giving enough points to popular hardware. Rev3 is beneficiary to those not benching expensive hardware. Please grasp this :)

practical example1:
* You buy a $50 8800 GT now to bench; you check 3D01 ranking: http://hwbot.org/listResults.do?gpuModelId=1233&applicationId=1&filterUser=true&filterBlocked=true&sli=false&limit=100 ; even if you bench it with LN2 and high end CPU you'll only get 15 HWboints. In rev3: since there are over 4000 scores with 8800 GT the popularity is very high and thus more than TOP 20 will get awarded, also the maximum HWBoint that you can earn is also increased; it won't be 15 HWboints, but noticeably more than that. Outcome: in Rev3 you get more points for benching popular (older) hardware.

practical example2:
* You buy a $500 Radeon HD 5970 now to bench; you'll get plenty of Global Points; but little to no hardware point as there are not many people benching that hardware. In Rev3: you'll get plenty of Global Points, but less than Rev2, because now you're also competing in the global ranking with 4xRadeon HD 5870 scores, LN2 overclocked GTX 285 and what not. Outcome: In Rev3 you'll get less points for benching high end hardware at stock speeds

[SIZE="3"]NOW ADD 1 + 2: does that seem like Rev3 is rewarding CHEAPO HARDWARE OR EXPENSIVE HARDWARE[/SIZE].... you tell me.

Comment from Kal-EL (OC Alliance) - link
Time: Nov 26, 2009

Competition has gotten pretty stiff since phase builders have picked up the pace lately and been more accessible to customers.

Comment from tiborrr (OC-Lab) - link
Time: Nov 26, 2009

Well, thank god i build my own cooling equipment :)

Comment from thebanik (OC Alliance) - link
Time: Nov 26, 2009

jmke said: an SS unit that can handle the output wattage of 8800GT can be had for less than $100 on second hand market:)

As such I am cool with whichever direction hwbot goes, and I would not leave for sure unlike what some other benchers said. But I really am very very skeptical about this change. Usually I do somehow manage to buy 1 top end card every 6-8-10 months hence the slight shift to single cards getting more score sounds pretty good to me since I also use DICE/LN2 from time to time. Nothing much to lose here atleast.

But now coming to the $100 SS. I am in India and only 3 LN2 benchers here. SS/cascade non-existent. I tried asking for QUOTE for SS/cascade from both US/Europe, and only the shipping price was in the range of 500-800$. Leave aside the price of the ss/cascade. Now certainly cant afford this much of moolah, and so is the case with many I believe who are not in US/Europe. I really have the feeling that with the new hardware points system, the poor bencher is gonna be killed. I hope your team proves me wrong, and hope thats big time wrong.......

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Nov 26, 2009

Bones said: Perhaps it could be something to use for future revisions of the site - Who knows?

It's something we can keep in the back of our head, that's for sure. But to work out the details of such system for the Rev3 is simply not possible, I'm afraid.

Comment from egm_xt - link
Time: Nov 26, 2009

I read pretty much the whole thread and for me I think change is always welcomed if we are willing to take a step back if does not work as well. When I looked at the way the site works I was amazed and really think it is very clever. It is always possible to improve and you are in a much better position to judge it. But I feel that the data base role that you have is really very well served by the point structure you have right now. It gives some (not really too much) incentive to look for obscure hardware and bench it. There is a lot of information when a processor or videocard that is empty is benched for the first time (actually from the information theory point of view much more than the 100th submission of a popular hardware). Looking from this angle 2 points is not really that much. Please think that this way you might be contemplating a bigger picture than just extreme overcloking skills. Of course that is a choice you have to make. For me anyway it is fine and I think you guys are doing a terrific job!

Comment from Bones (Classicplatforms.com) - link
Time: Nov 27, 2009

Massman said: It's something we can keep in the back of our head, that's for sure. But to work out the details of such system for the Rev3 is simply not possible, I'm afraid.

That's OK and at least my suggestion did bring something to consider in the future as a possible option. Nothing wrong with tossing a few ideas out there to get one's mind working on the possibilities.

Comment from _Datura_ (XtremeSystems) - link
Time: Dec 1, 2009

thebanik said: But now coming to the $100 SS. I am in India and only 3 LN2 benchers here. SS/cascade non-existent. I tried asking for QUOTE for SS/cascade from both US/Europe, and only the shipping price was in the range of 500-800$. Leave aside the price of the ss/cascade. Now certainly cant afford this much of moolah, and so is the case with many I believe who are not in US/Europe. I really have the feeling that with the new hardware points system, the poor bencher is gonna be killed. I hope your team proves me wrong, and hope thats big time wrong.......

Who said that pioneers would have an easy life :)

But I also think that the amount of OC'ers in India must be rising? With so many IT industrie there must be some potential for OC'ers!

Comment from Thor941 (JMax-Hardware) - link
Time: Dec 1, 2009

Thor941 said: Hi,

I just don't understand one thing :

- is there any interets keeping all the scores we have made with the new version? I mean let's say I have made 4 scores in one benchmark with a 4870, I always keep the best one on the bot and delete the other ones myself.

Hope I am clear enough ...

Hi!

Can anybody answer to this?

Thanks a lot ...

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 1, 2009

For you it may be interesting to use as information for future benchmarks. Keeping your scores allows you to check voltage/frequency/settings/subtest scores so that you have a target to aim at in the next session.

For us, it's interesting to do further analysis of the results. For instance to catch cheaters or for scaling analysis.

Comment from Thor941 (JMax-Hardware) - link
Time: Dec 1, 2009

yes ok but the point is : can I "re-post" my "old" scores (less than 1 year old) expecting to have points with the new revision, or is it forbidden?

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 1, 2009

Why would you repost them? The new hwboints revision would apply to all submission, also old submissions.

Comment from Thor941 (JMax-Hardware) - link
Time: Dec 1, 2009

yes but I have also deleted some HW score that were not in the top 20 anymore - If I understand well, with the new revision maybe the top 50 or 100 will have some (few points) - or there are some scores that I have not posted cause they were not in the to 20 of the catégory..

in fact the question : if I post these scores now hoping to have some points with the new revision, will thet be deleted as they are a few months old?

Thanks :)

Comment from Thor941 (JMax-Hardware) - link
Time: Dec 1, 2009

ok ...

thanks a lot :)

Comment from tiborrr (OC-Lab) - link
Time: Dec 1, 2009

I lost more than 65 hw points :mad:

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 2, 2009

That complaint should go in a different thread.

Comment from rbuass (Team Adrenaline) - link
Time: Dec 3, 2009

Dou you know when the Rev. 3 starts?

;)

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 3, 2009

It's scheduled for the first week of januari.

Comment from knopflerbruce (PURE) - link
Time: Dec 3, 2009

richba5tard said: It's scheduled for the first week of januari.

When is the test server up?:)

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 5, 2009

I hope to get a test server up to show the effect of hwboints rev3 by next weekend. Development is on track.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 5, 2009

Whoah!

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 5, 2009

We might need to tweak the amount of wPrime rankings though. :D

Comment from bazx (Bench Tec UK) - link
Time: Dec 5, 2009

that is a beautiful sight RB

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 5, 2009

I think we better do not show rankings with less than 10 (20? 50?) participants, otherwise they are a few meaningless ranks in multithreaded cpu bench apps.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 5, 2009

Whoah!

This looks so incredibly awesome!

Comment from CaNNon (411 Overkill) - link
Time: Dec 6, 2009

I like the formatting looks like it's going to be great!

Comment from knopflerbruce (PURE) - link
Time: Dec 6, 2009

I think it looks good :) Sure it's "depressing" with so few results for some categories, but it will probably go up after a while anyway.

Do I get any global boints for that 38.960s wp32m run I did earlier today btw?:D

Comment from oxyyy (Overclocking Team Austria) - link
Time: Dec 6, 2009

i love it! :)

So i have to bench some single Cards now!? :D

How will hwboints calculated in wprime?

1Core, 2Core,...

or

1Socket, 2Socket,... ?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 6, 2009

oxyyy said: So i have to bench some single Cards now!?

You bench whatever you're best at: 100th single GPU may be less rewarding than 20th dual gpu.

Comment from thebanik (OC Alliance) - link
Time: Dec 6, 2009

I am holding off my purchase of 5970 or 2 x 5870 till rev3 is live and we/I understand the scoring.....Afterall I can purchase a card only once or twice a year....and hopefully as promised guyz like me wont be simply killed by the new revision.....

Comment from der8auer (PC Games Hardware) - link
Time: Dec 6, 2009

Looking great so far :) Go on RB!

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 6, 2009

RB has had the first view on the new ranking, if I'm not mistaken ... uh-oh.

:D

Comment from komadyret (411 Overkill) - link
Time: Dec 6, 2009

Already? LOL

Can we expect a quick glance anytime soon, just out of curiosity?

Oh, never mind. Let him work out the code and get it ready, I can wait 4 more weeks :D

Comment from knopflerbruce (PURE) - link
Time: Dec 6, 2009

Massman said: RB has had the first view on the new ranking, if I'm not mistaken ... uh-oh.

:D

Will the mods get access to the server in some way?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 6, 2009

I guess they will.

Comment from K404 - link
Time: Dec 7, 2009

How does the new Hardware Masters League look? :p

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 7, 2009

Not important as the hardware boints algorythm has to be adjusted slightly.

Comment from knopflerbruce (PURE) - link
Time: Dec 7, 2009

Massman said: Not important as the hardware boints algorythm has to be adjusted slightly.

What adjustments are we talking about?

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 7, 2009

- more hw points when > 50 participants, less hw points when < 10 participants
- 0.1 ptn for your best submission, even if it's not in top ...

Comment from egm_xt - link
Time: Dec 7, 2009

One thing I could not figure out is what is the current algorithm logic when classifying hardware masters with only 300 points (no global points). The ranking seems to fluctuate all the time. Don´t you think it would be easier if it was ordered by the most hardware points exceeding 300? May be it could be addressed in the rev. 3. Thank you.

Comment from thebanik (OC Alliance) - link
Time: Dec 7, 2009

egm_xt said: One thing I could not figure out is what is the current algorithm logic when classifying hardware masters with only 300 points (no global points). The ranking seems to fluctuate all the time. Don´t you think it would be easier if it was ordered by the most hardware points exceeding 300? May be it could be addressed in the rev. 3. Thank you.

If I understood you correctly then, though they cannot exceed 300 points showing globally but this is the main hardware point by which the rank keeps fluctuating, http://hwbot.org/hardware.masters.do

Comment from egm_xt - link
Time: Dec 7, 2009

There are some people that have exactly 300 points just from hardware points (exactly my case :) ) I do not know what is the way they are placed in the global rank relative to each other that have also 300 points and could not figure it out. I only see that among these people they keep changing within themselves the place in the global rank. I suggested that the total hardware points could be the factor. It could as well be the date that the person reached the 300 point mark (the earlier ones get higher position) or whatever else. I just want to know the way it is right now and for me it seems that the total hardware points would be the better way of doing it and an algorithm revision could address this point.

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 10, 2009

I've got a first draft ready of hwboints rev3, which will be shown to massman later today. Depending on his feedback, a test server will be made public this weekend. Personally I think the results are looking good.

Overview of changes:
- ranking based on # of cores instead of single socket / multi socket for both cpu and gpu
- you can get global points only once for each benchmark application (your highest scoring submission) instead of twice
- global points distribution mostly unchanged, but everyone will get 0.1pts for his/her best submission per benchmark application.
- hardware points distribution changed: less points when there is little competition (min 1ptn for first place instead of 2), 3x more points if there are +300 competitors. (popular processors/videocards)
- point calculation is instantanious (!) instead of hourly, and can be retriggered on command for a specific submission.

Comment from BenchZowner (Hellas Overclocking Team) - link
Time: Dec 10, 2009

Very nice RB!

Comment from Dragon Oc Team (HwMania-OC-Team - HwMania.org) - link
Time: Dec 10, 2009

very nice

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 10, 2009

Something we will need to decide is what to do with unlockable cpu's. If you ignore the fact that many AMD cpu's can be unlocked, the single core rankings will be ruled by unlocked dual/triple core X2/X3's.

Comment from knopflerbruce (PURE) - link
Time: Dec 10, 2009

Sort by number of active cores. Otherwise it's no "real" single/dual/triple-core rank - just as it's not OK to submit sempron results with an unlocked core in the Country Cup.

Comment from Team Hardware Forest (Pc World Oc Team Italy) - link
Time: Dec 10, 2009

i'm agree this opinion

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 11, 2009

richba5tard said: I've got a first draft ready of hwboints rev3, which will be shown to massman later today. Depending on his feedback, a test server will be made public this weekend. Personally I think the results are looking good.

First time I looked at the rankings, it looked pretty good. None of the three rankings (overclocker, team, master) is completely different; not even classicplatforms is killed.

There were a couple of bugs, though, so the test server wasn't installed yesterday evening, possibly later this week or next week.

I did drop in rankings quite heavily ... :-/

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 11, 2009

I mean the overclockers rankings.

I need to shape up!

Comment from K404 - link
Time: Dec 12, 2009

Is 50 users really high enough for a tripling of points?

50 users is... y'know... 2 per country, if that. Thats nothing! (IMHO) how about 100 users?

50 users in one category could be 200 results (using the current results stats) if they dont submit for all available benches.

Categories like the E84, Q66, 8800GTX could be classed as tough as..say... the 7800GTX, 7900GT... which (IMHO) aint even close

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 12, 2009

K404 said: Is 50 users really high enough for a tripling of points?

50 users is... y'know... 2 per country, if that. Thats nothing! (IMHO) how about 100 users?

50 users in one category could be 200 results (using the current results stats) if they dont submit for all available benches.

Categories like the E84, Q66, 8800GTX could be classed as tough as..say... the 7800GTX, 7900GT... which (IMHO) aint even close

50 persons is indeed not much. I changed it to 200 people in the latest revision of hwboints, and the results are much better.

Comment from Barton (Classicplatforms.com) - link
Time: Dec 13, 2009

Do you have a link allow members to see the new rankings?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 13, 2009

Still some bugs left to iron out.

Comment from Iron (HwProject.net Italy OC Team) - link
Time: Dec 13, 2009

Massman said: Still some bugs left to iron out.

Oohh well, if you want me out, then I'll go.... :(:(

:D :D :D :D

Comment from Oj0 (PURE) - link
Time: Dec 13, 2009

So where's the test server? :D

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 13, 2009

It's down due to me still debugging it. :)

Comment from Dragon Oc Team (HwMania-OC-Team - HwMania.org) - link
Time: Dec 13, 2009

when we'll have pubblic test server?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 13, 2009

somewhere next week, I suppose

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 13, 2009

Depends whether I can work this night or not.

Comment from TiN (XtremeLabs.org) - link
Time: Dec 16, 2009

Btw, there is one more problem

If CPU's will be divided by sockets - a lot of results with not correct CPU socket number will move into multi-socket.
People who not careful with this will lost points.
Like those who think - wow, I have quadcore CPU so it's 4 cpus here.

So moderators will have more work here.

examples:

http://hwbot.org/compare.do?resultId=916824
http://hwbot.org/compare.do?resultId=781243
http://hwbot.org/compare.do?resultId=861567

and so on

Comment from Gomeler (PURE) - link
Time: Dec 18, 2009

A little birdy told me the test server is occasionally working. Any chance of seeing it? I'd like to see how bad of a beating I take with the new arrangement.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 18, 2009

The test server is constantly being debugged ... everything you've heard so far could actually be incorrect information.

Comment from Brian y. (PURE) - link
Time: Dec 19, 2009

Massman said: The test server is constantly being debugged ... everything you've heard so far could actually be incorrect information.

so that means that Massman won't be #1 in the new rankings?:D

J/K:p

Comment from CaNNon (411 Overkill) - link
Time: Dec 19, 2009

jmke said: HWbot mods get 500 HWboints bonus;)

Ouch.... straight for the kill! :)

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 19, 2009

guess what HWBot admins get ... :D

Comment from Kal-EL (OC Alliance) - link
Time: Dec 19, 2009

free paper products?
gift coupon at Toy Land?
firm handshakes?
Phenominal Cosmic Power?
rub and a tug?
a false sense of omnipotence?

j/k I'm guessing free hardware.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 19, 2009

HawaiiSuperman said: a false sense of omnipotence?

No, Banned!

J/K :p

Comment from CaNNon (411 Overkill) - link
Time: Dec 20, 2009

lol!

Comment from cowgut - link
Time: Dec 23, 2009

first off i love you guys and thanks for all the hard work.
now what i logged in for
i see from this post why you guys think you need to change things around.
http://hwbot.org/article/news/important_need_your_opinion
its your show and all so we have no choise but to follow your rules and such...but are you sure that it is a money thing?
i dont want to get off the track here by saying this but if you want an even playing field how come manufactures given hardware is able to gather any points at all...read the summissions 1000's of them thanking thier respected "doners".(lots in the x2 cards too)
is that a level playing field? or how about when say they have a gathering of the greatest in a world comp....how come those submissions are aloud?heck they dont buy any of the rig at all +free ln2.(not taking away from the o.c'ers themsleves they are there for a reason)
now lets put the shoe on the other foot...me being a true vga junkie that buys all my own cards what kind of message does this send me?is it take the points from the little guys and spead them around try to not mess with the big guys points as much as possible?

the x2 card rule as two gpu's: what happens in theses card rankings as per new v3.0 is that they will be put in base single cards ranking as x2 gpu which is sorta messed up since its 1 card.
now we all know the single gpu class will take sub zero cooling which is the standard now a days.....so im thinking is the new points system to further the big guys?
i like the fact now that the 59x2 cards are most all air/water is it too close from laymen to big guy?
are you guys mad when a little guy with paid for hardware even gets close to someone who has backing from the "man".
pc hardware is money as you well know its never going to change

"Overclockers league should be about skill, not money"...are you kidding? buy the right handpicked chip you can own all of am3 05 06 with a card on air.

i have more opinions but i'll wait for a reply
like i said i love you guys
thanks gino

Comment from cowgut - link
Time: Dec 23, 2009

i'm no rat and you know who i dont need to tell you
you can see in submissions whos butt they have to kiss by thanking them lol.

as for free ln2 did they not post scores from the msi oc? me and hundreds of other little guys pay are own way and dont want or need the man to help us even thou were not rich..
i'm not looking to start truoble with the best clockers in the world that get free stuff.(they deserve it)
but dont even say "even playing field" till you at least think about getting the free review hardware out of the points listings.

you know just go to a forum or two to see for yourself.... some cant even rerun cus they had to send the cards back..... plz dont denigh it

merry chirstmas /happy holidays guys

Comment from cowgut - link
Time: Dec 24, 2009

yeah i get alot of that,i tried to make it mo better to understand...i know you guys cant get by with just little guys
will see how it goes

Comment from cowgut - link
Time: Dec 24, 2009

no i could not log on under my user name for some reason

ok i we will see how it all pans out in a few weeks.

Comment from K404 - link
Time: Dec 24, 2009

For Rev3, would it be possible for all a users old results that dont get points be hidden in a users profile unless we navigate to "Show all"? (the results that have been moderated and cant be deleted)

I hate seeing my old results come up in my submissions page, im only interested in seeing my best at first-glace.

Comment from Dragon Oc Team (HwMania-OC-Team - HwMania.org) - link
Time: Dec 24, 2009

i have just seen the test server, but we have all the possible achivement now, like the helium achivement or the 5k submission achivement

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 24, 2009

How come you have access to test server? :) achievements in profile will be implemented Saturday

Comment from Dragon Oc Team (HwMania-OC-Team - HwMania.org) - link
Time: Dec 24, 2009

sent you a pvt

Comment from [UC OC] Karmakazi (PURE) - link
Time: Dec 25, 2009

jmke said:
rev3 will do the opposite of what you state here. FYI ;)

just to humor us, give a name of somebody you who gets hardware/LN2 for free and somebody who paid everything himself :)

He is quite right, PURE will barely make top 5 with the new revision. It wont be like that for long though hehe

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 25, 2009

I would say an opinion from the old school hardware. The GPU categories being divided by GPU count is absolutely clear and makes sense. But I can't say the same about the CPU's. I understand the reason why this decision was taken - it's the wPrime global Hall of fame where servers are a must. And I agree that this is a bad situation - in some means the same as dual GPU cards against single in the present single card Hall of Fame.

But lets put the global boints aside and look at the hardware points. What can one achieve in CPU-Z or Pifast using a bunch of CPU's? Nothing good to be exact. Does PCMark boost it's speed on a 48-way Dunnington? Nope. So were do we have dual socket systems?

I divide them in two categories - the workstation ones and the server ones. The first category is socket 7, socket 8, socket 370, slot 1, socket A. And now tell me - do you try to use the best test bed for a certain CPU and benchmark? Look at the Super pi 1M and Pifast. Everybody tries to use LP DK P45-T2RS for the first one and Rampage Extreme for the second. This can be called record preparation.

The same is the I-ram.

The server platforms - slot 2, socket 603/604, 771, 940, F have only a few results so according to the new formula they won't achieve many scores anyway. So they can't change the balance. Not talking about global points, I want to clear this again, only HW.

And the "workstation" - all of them are only dual socket (except socket 8, a very rare exception) and single core. So I assume that using a dual CPU system in PCMark and wPrime is rather a tweak, a record preparation, the same way I-ram is. After they are separated, they will get close to none points and overclocking them will make no sense. You have to find an overclocking MoBo, but twice the number of CPU's and get only a couple points. And a much weaker result in a single CPU will give more. Is this good? I think, that the first result will require more time, more effort and more money.

So this means, the rev.3 will kill this CPU class, if I understand the upcoming rules right. Look at this for example: http://hwbot.org/hardware/processor/pentium_3_800mhz?tab=hall%20of%20fame
Turrican made a good job in PCMark05 combining dual CPU, 3850 AGP, I-ram. His overclock is more with two CPU's than the uniprocessor frequency on the second place. Another dual CPU result in only fifth. That means that even the jump from 1 to 2 cores (not to mention 2 to 4 or 4 to whatever) is not a panacea. Same can be seen here: http://hwbot.org/hardware/processor/pentium_3_1.0ghz_coppermine?tab=hall%20of%20fame
If your video and HDD is weak, you can't be sure in being on top. So it is logical to separate PCMark05 by video and I-ram count too. No, I don't seriously mean this :D I just wanted to show that separating HW points by CPU count is no good.

So we only have wPrime left, same a the global points situation leading to this. Looking here: http://hwbot.org/listResults.do?cpuModelId=1015&applicationId=14&filterUser=true&filterBlocked=true&limit=100
The thirteenth result (Olesius) was done with one CPU and a lower frequency that the dual CPI first. When they will be separated, it will become sixth. I think it will get more points than Turricans being achieved much easier.

Then look at this one: http://hwbot.org/listResults.do?cpuModelId=1381&applicationId=14&filterUser=true&filterBlocked=true&limit=100
The 22-nd result (Dewa_OC) is made on the same frequency as the first with two CPU's. It will be sixth but from 35 compared to first of 16. As the the first place cost will rise faster from the number of submissions than now - will dual systems survive? Many would prefer to bench with one CPU compared to two.

The best one would be to separate only the global points rating by sockets and only for wPrime. Or make a much harder algorythm and connect the dual CPU score cost with both the uni and dual submissions. So if someone can achieve results with two CPU's on close speeds as with one, then he gets more than it should be like with one.

Many words are written, hope somebody reads this and at least one understands :D making me believe this wasn't a waste of time.

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 25, 2009

Antinomy said:
But lets put the global boints aside and look at the hardware points. What can one achieve in CPU-Z or Pifast using a bunch of CPU's? Nothing good to be exact. Does PCMark boost it's speed on a 48-way Dunnington? Nope. So were do we have dual socket systems?

Not sure whether you know, but in rev3 only multithreaded cpu apps are divided by #cores. So there is only one ranking for superpi, cpu-z, etc.

Your thought of seperating cpu rankings based on socket instead of cores is a valid approach. The downside would be that the global rankings would be ruled by extremely expensive i975's, while if we split them up by core, you would have a fair chance to get quite some global points with a 40$ Sempron 140. Imho the latter would make overclocking more fun, and focus more on the overclocking itself (cooling, tweaking, voltmodding) instead of buying a 800$ chip and hoping it's a good overclocker.

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 25, 2009

richba5tard said: Not sure whether you know, but in rev3 only multithreaded cpu apps are divided by #cores. So there is only one ranking for superpi, cpu-z, etc.

Nope, were would I read this? I only read the bits that you announce and use my painful brain :) this is good news. Cause it makes sense.

Your thought of seperating cpu rankings based on socket instead of cores is a valid approach. The downside would be that the global rankings would be ruled by extremely expensive i975's, while if we split them up by core, you would have a fair chance to get quite some global points with a 40$ Sempron 140. Imho the latter would make overclocking more fun, and focus more on the overclocking itself (cooling, tweaking, voltmodding) instead of buying a 800$ chip and hoping it's a good overclocker.

Now this one shows you don't get me. I repeat - I agree that the situation with the global wPrime is bad and everything you did there is correct. But... my post wasn't about global points, it's about Hardware ones.

The score is a sum of Global points + Hardware points - yes? So I agree with the separation of global points and agree with you absolutely but this isn't an answer to my question and opinion. My question was about Hardware points. Yes, I mistakenly wrote about socket separation in the end instead of core count. But the mistake should be obvious because of this:

But lets put the global boints aside and look at the hardware points.

Not talking about global points, I want to clear this again, only HW.

The best one would be to separate only the global points rating by sockets and only for wPrime.

The last QUOTE meant that the a good way would be the core separation being done only in global points, not HW.
So I hope there will be an answer on the hardware score situation.

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 25, 2009

That's a bad one :(

Comment from knopflerbruce (PURE) - link
Time: Dec 25, 2009

Well, by running two CPUs at stock you will beat any single-CPU rig no matter how high it goes. Not sure there's much skill in that...

Of course you have situations where really good results won't get credited as much as they deserve with this new revision (such as s7e9h3n's sub-22s Opteron 154 superpi 1m result), but what matters is that the new system brings an overall improvement, not that every score was rewarded 100% correctly.

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 25, 2009

knopflerbruce said: Well, by running two CPUs at stock you will beat any single-CPU rig no matter how high it goes. Not sure there's much skill in that...

Look at my first big post and thoughts on PCMark results. It shows your statement is incorrect. I-ram and GPU make more.

With wPrime - yes, you are write. I just wanted to say that the good old dual socket workstation class will be dead with the new rules. As it's much easier to take the golden cups in a popular integrated video and get more points than do hard work on overclocking a dual system.

But when you get a bunch of points by running at stock - this is certainly wrong, I agree.

Comment from Gomeler (PURE) - link
Time: Dec 26, 2009

Any chance of a public test server in the coming days? Trying to figure out the best way to game err compete with Rev3.

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 28, 2009

HWBOT rev3 test server

This is a test server. It can not handle heavy load, and is not without bugs. You may forward people to this topic, but you may not redistribute this link on other sites. (due to bandwith being very expensive for the test server) The data represents a production snapshot on 15th December. You may try to submit or alter old submissions, the database will be reset tomorrow anyway.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 29, 2009

Do note that the Teams ranking provided on the team ranking page isn't correct. There's a bug left to squash.

Comment from K404 - link
Time: Dec 29, 2009

Thanks for letting us see how it looks :D

Will rankings be put next to the relevant results so we dont need to find + expand the result that counts towards global? :)

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

There is room for improvent on that page, but we got little time left.

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Dec 29, 2009

I'm fairly disappointed. My team took it in the nose for roughly 1300 points. I was an idiot for thinking the older platforms might still have a chance to compete decently with the high cost equipment after the revision.:(

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

Nice look :)
And the category list (socket 478, for example) has the full page scrolled, not only the frame. That is very convenient for low res.
And there are visual disadvantages - I shouldn't report on them because this test server is only an example for the database or I'm wrong?
For example, when you list the hardware hall of fame, the name of the benchmark isn't separated from the results neither by font size nor by color. And why no [+] for the result details in the HOF?

Scottie, everything changes, you should know. I prefer to bench the old ones too. Are you here only for boints or for fun? ;)
The GF4 OC challenge shouldn't be the last - you can show your skills there. Or just make a private competition for the interested ones.

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Dec 29, 2009

Oh, I didn't say I would stop benching. I enjoy it. Just a little disappointed is all. I AM looking forward to the monthly contests though.;)

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

That's what I love to hear! :)

Comment from Bones (Classicplatforms.com) - link
Time: Dec 29, 2009

I've noted a few bugs in the new revision.

Saw in the 4000+ San Diego listings as an example, some results have lower points than results they actually beat and this isn't the only example I've seen. Another is the 3700+ Newark ( Mobile 64 3700+) listing has a few errors there too. Just browse around and you will see these.
May help explain why some took a monster hit in points as I did while others didn't get zapped so badly.

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

Bones, are you sure it's not because the results that are lower got not only hardware points, but global ones too? There is a separation by #CPU_cores for wPrime and PCMark, didn't you forget? I saw the same, but figured this one out quickly.

P.S. damn, I like your team for your nicknames :D

P.P.S. if there are less than 5 results, then the first place takes only 1.5 boints. Then (with 5) 1.8 and 2 (6). I don't get how many people get points. If it's a popular - it's 50% as richba5tard said. But what about the unpopular, say 1-40 results?

Comment from komadyret (411 Overkill) - link
Time: Dec 29, 2009

Had a quick peek at it now, and I'm pleased.
Took a bad hit i points, but nothing unexpected. And judging by the current results om the test server my team just about retained the same points as in rev2. A few members actually got a nice boost in points :)

What takes a little to get used to is the score-inspection. Having the hardware info spread out like that and with a icon/picture in there is alright I guess, but it was a little confusing at first glance.

A few of the results I've looked at gives me an impression that beeing 1st in a hardware is rewarded big above being 4th or 5th. (I've looked mostly at "uncommon" hardware) when there are few participants. I think I like this, it should urge us to push that little harder to gain those few extra points, and may let the smoke out of the hardware a little more often ;)

Good job, can't wait to see the finished product!

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

I can display points as global + hw, instead of total points... but that might be confusing to some too. Anyway, hwbot rev2 also shows total points, so why the confusion? :)

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

I think this is temporary, because they can't think that they can achieve globals on that hardware. But since there's a separate ranking for single-cores, they do.

What should be done - is to separate the benchmark name from the results. Also, to add enumeration of the results in HW HOF.
And maybe - draw a bold line showing were the results that achieve global points end up (after that only globals can be in the list).

Comment from jabski (Bench Tec UK) - link
Time: Dec 29, 2009

I like the new rev. nice work all who have worked on it.
One thing i found is the 'Stumbled on a Goldmine' acheivment goes to level 2. Is it possible to have level 3 and 4 ?

Many thanks

J

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

jabski, I think it goes up to lvl 4 atm, but achievements are no longer hard coded. We can add new levels / achievements as we go.

Updated the dev server:
- added "points" tab in user profile, which sheds a clear(?) light where your points come from.
eg http://94.225.192.196:8080/community/user/andreyang?tab=points
- fixed the team points algorithm, i think rankings are correct now
eg http://94.225.192.196:8080/rankings/overclocker/

Comment from Don_Dan (XtremeSystems) - link
Time: Dec 29, 2009

Yes, team points are working correctly now. :)

Splitting up global and hardware points for every result messes up the layout sometimes though.

"Points" tab is a great idea, I like it.

Comment from jabski (Bench Tec UK) - link
Time: Dec 29, 2009

Hi rich

thanks for your reply but i am afraid level 3 and 4 is not there. If you look at Turricans profile he is at level 2 but with 400+ gold cups

http://94.225.192.196:8080/community/user/turrican

or do the levels not go up in 100's ?

Many thanks

J

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 29, 2009

Jabski, you're assuming that Turrican has enough Gold cups to hit level 3 ... :)

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 29, 2009

http://94.225.192.196:8080/community/user/knopflerbruce

Check out this account :)

Comment from Don_Dan (XtremeSystems) - link
Time: Dec 29, 2009

Could someone please have a look at my Wprime 32M scores:

http://94.225.192.196:8080/listResults.do?userId=15466&applicationId=14

It says I get more points for my 2x submission on my x and 4x submission, but on the 2x submission it says I get more points for my x submission which is not true in fact.
Or am I missing something?

It seems to be working correclty for my 1024M scores though.

Also, this: "1146th out of 801, but no points as his 2x CPU submission earns more points." on my Pentium M 760 submission.

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

Don_Dan said: Could someone please have a look at my Wprime 32M scores:

http://94.225.192.196:8080/listResults.do?userId=15466&applicationId=14

It says I get more points for my 2x submission on my x and 4x submission, but on the 2x submission it says I get more points for my x submission which is not true in fact.
Or am I missing something?

It seems to be working correclty for my 1024M scores though.

Also, this: "1146th out of 801, but no points as his 2x CPU submission earns more points." on my Pentium M 760 submission.

That's a bug! Thanks don_dan :)

Comment from Don_Dan (XtremeSystems) - link
Time: Dec 29, 2009

I'm glad I could help! :)

Comment from Don_Dan (XtremeSystems) - link
Time: Dec 29, 2009

Something is wrong with the achievement "I have double vision":

I have double vision: 26/50

Submit 30 results using multiple graphics cards

( took this from stummerwinter's profile )

on Kingpin's page:

I have double vision: 30/50

Submit 30 results using multiple graphics cards

Comment from Oj0 (PURE) - link
Time: Dec 29, 2009

http://94.225.192.196:8080/rankings/overclocker/team/pure

Brings up a blank list for me :(

Comment from thebanik (OC Alliance) - link
Time: Dec 29, 2009

with new submissions on the Test server wont there be a recalculation of total points just yet???

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 29, 2009

Oj0 said: http://94.225.192.196:8080/rankings/overclocker/team/pure

Brings up a blank list for me :(

No problems here.

thebanik said: with new submissions on the Test server wont there be a recalculation of total points just yet???

What do you mean?

There's no point in submitting scores through the test server, the database will be reset when it goes public.

Comment from PeterStoba (Bench Tec UK) - link
Time: Dec 29, 2009

Oj0 is right, last night the team member rankings were showing blank, but they are fine now.

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

Massman said: Jabski, you're assuming that Turrican has enough Gold cups to hit level 3 ... :)

Massman said: http://94.225.192.196:8080/community/user/knopflerbruce
Check out this account :)

OMFG, I don't wanna know, how much you need to get level 4. Maybe kill all the HW masters and take their scores :D

But there's an old hint to get enough of goldies.

There is a mistake: "costom flag" - the ALT for team flags. Guess it should be "custom" :)

And for the scores page - when it's global and you see your score as "Global + HW" it's normal, cause there aren't much global rankings. But when listing the HW scores, the global points there being in front make mess. Maybe display only globals in global and only HW in HW? At least on the members page.

Comment from ZFeSS (Hardware Hackers) - link
Time: Dec 29, 2009

Massman said: No problems here.

Thanks for fixing. I've had a same problem few hours ago.

It will be wery useful if we can see the clocks in hardware HOF. I need to open each result in hardware cathegory to see them, it's incovenient.
At rev.2 we have a "+" function and can see CPU and Mem clocks for 2D and CPU and GPU clocks for 3D, if user add them, but in rev.3 I can't can't compare two-three results not opening the "details" in new browser tab or window. Will be some changes in this way, or it's final version of interface?

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

What do you mean with hardware HOF? (link)

It's not final, hwbot rev3 can be tweaked till tomorrow night.

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

http://94.225.192.196:8080/hardware/videocard/geforce_6150_se?tab=hall%20of%20fame - this one for example ;)

[+] button would be really nice.

Comment from Oj0 (PURE) - link
Time: Dec 29, 2009

Thanks, team rankings are working now.

Comment from ZFeSS (Hardware Hackers) - link
Time: Dec 29, 2009

richba5tard
http://94.225.192.196:8080/listResults.do?gpuModelId=1587&applicationId=2&filterUser=true&filterBlocked=true&sli=false&limit=100
No enumeration - I see that SF3D is third in ranking, but if I look down I need to count :) No VGA clocks. No "+" button. The last is not critical, but it was helpful.

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

And the HW rankings are not separated by #cores yet. I.e. by CPU count...

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

I made a change to the hwboint algorithm:
previously: 0.1 points minumum for (ranked) global submission, 0.0 points minimum for (ranked) hw submission
now: 0.0 points minumum for (ranked) global submission, 0.1 points minimum for (ranked) hw submission

It does not change anything to the ranking, but now there is a difference between inactive people (almost no submissions), and active people who are poor in overclocking.

Comment from dejo1 (OC Forums) - link
Time: Dec 29, 2009

I am one that generally uses older video cards, but cant see the benefit of the new algorithem. This will force all the extreme cooling guys to the categories that have the points and they still wont be available for people that dont have the skills or the hardware to compete on an international lever.
The guys that want the points and have the means will just go and get them, and it would prolly make it harder for the people that cant tweak or cool as well the aquire points. I say leave well enough alone.
Are we possibly at risk of loosing some of the longtime benchers that really keep the blood flowing in this arena

Comment from Thor941 (JMax-Hardware) - link
Time: Dec 29, 2009

I find the new rev really weird :/

It's gonna be hard to get used to it!

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 29, 2009

dejo1, so what's the problem, your suggestion?
If you can't tweak - learn, there are dozens of articles about this. They give you basic skills. If you don't know, how to o/c better or cool - think or ask. I know guys that got in top3 in a webmasters o/c competition that overclocked a dual Xeon system in a fridge! Don't tell me you don't have one. Or watercooling, or Peltier (nice one for grannies video).

The elite will compete in the mass products, the new GPU and CPU. My teammate just called me and said that the reward for the old stuff - the P1-P4 (478), A-XP, K7, Voodoo, GF4 and 9800 is in our hands. We can make a battle and rise the popularity and prize.

You hear that, CP? There's enough necrofils for that, be sure.
But usually it's a very complicated category, where you must learn many tweaks and hints to get to the top, so for those who like the quick&dirty boints will better bench another pack of GF8/9 or HD3000/4000. ;)

Comment from Thor941 (JMax-Hardware) - link
Time: Dec 29, 2009

Anyway there is a bug in the actual rev :)

http://hwbot.org/community/user/thor941?tab=profile

Comment from Warrior_oF_Byte (Overclocking Argentina OC Team) - link
Time: Dec 29, 2009

Thanks for sharing the testserver!!!! :D

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

Antinomy said: There's enough necrofils for that, be sure.

Haha, I like that. Next month, "HWBOT Necrofilia Cup". Benching hardware that has been claimed dead long time ago. P1/2/3 K6. :D

Comment from Turrican (Team OCX) - link
Time: Dec 29, 2009

no problem, i'm well prepared for allmost every old hardware contest. :D

Comment from thebanik (OC Alliance) - link
Time: Dec 29, 2009

Massman said:

What do you mean?

There's no point in submitting scores through the test server, the database will be reset when it goes public.

I know I know, just wanted to check how much difference 5870 runs would make to my final score and rankings which have dropped quite a bit with rev3

Comment from Sam__ (Bench Tec UK) - link
Time: Dec 29, 2009

is the test server offline? :(

btw its looking very nice and im sure it wont take people very long to get used to the new scoring and start taking advantage of it.

Comment from Gomeler (PURE) - link
Time: Dec 29, 2009

yeah it is offline.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 29, 2009

Apparently an IP change; should be fixed in ~30 minutes

Comment from Sam__ (Bench Tec UK) - link
Time: Dec 29, 2009

back up :)

was just having a browse arund and wondering why the cpu "amount" search option is only done by socket and not number of cores when the ranking is done by cores? would be extremely usefull if you could search by no of cores.

Comment from knopflerbruce (PURE) - link
Time: Dec 29, 2009

dejo1 said: I am one that generally uses older video cards, but cant see the benefit of the new algorithem. This will force all the extreme cooling guys to the categories that have the points and they still wont be available for people that dont have the skills or the hardware to compete on an international lever.
The guys that want the points and have the means will just go and get them, and it would prolly make it harder for the people that cant tweak or cool as well the aquire points. I say leave well enough alone.
Are we possibly at risk of loosing some of the longtime benchers that really keep the blood flowing in this arena

The amount of people that get boints is increased, so there will still be available spots for the "24/7-OCers". Sure they may not get as much as 10 boints, but even 2-3 will be more rewarding than 0. Plus, I don't see why it's a bad thing that the LN2 people must bench other stuff than Intel xtreme edition CPUs and top-end GPUs.

Comment from TiN (XtremeLabs.org) - link
Time: Dec 29, 2009

Btw, maybe better change

Popularity: 17183 submissions were made using this processor.

to

Popularity: 17183 submissions were made using this processor, xxx users..

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 29, 2009

Test server is going down, need to do some fixes. It will be back online tomorrow morning.

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Dec 29, 2009

jmke said: do check out why; in order to get rewarded you have to bench popular hardware (can be new or old), your team takes a "hit" because you had a lot of results in unpopular categories

jmke, I totally understand why we take a hit, but that does not change the fact that we're not being rewarded for any OC'ing skill even though it's on what is deemed "unpopular" hardware. Just as much time and effort is put into benchmarking that hardware as any brand new hardware. It's a moot argument anyway. HWBot will continue to do revisions as they see fit, which is their right, and the demise of team CP has already begun. Most of my team has taken the stance that they were discriminated against for racking up 3200 plus points on old hardware in less than a year. Thus pissing off a few teams behind us that use extreme cooling and spend more dough on hardware. Anyhow, I doubt many on my team will continue to bench here for the above reason. As for me, I reserve my opinion simply because I like to bench, and will continue to do so here. It's a hobby for me, and I'm not willing to give that up for something as silly as some points that mean nothing anywhere else but here.
Best Regards
~Scotty :)

Comment from VasGTO (PURE) - link
Time: Dec 29, 2009

Mr.Scott said: jmke, I totally understand why we take a hit, but that does not change the fact that we're not being rewarded for any OC'ing skill even though it's on what is deemed "unpopular" hardware. Just as much time and effort is put into benchmarking that hardware as any brand new hardware. It's a moot argument anyway. HWBot will continue to do revisions as they see fit, which is their right, and the demise of team CP has already begun. Most of my team has taken the stance that they were discriminated against for racking up 3200 plus points on old hardware in less than a year. Thus pissing off a few teams behind us that use extreme cooling and spend more dough on hardware. Anyhow, I doubt many on my team will continue to bench here for the above reason. As for me, I reserve my opinion simply because I like to bench, and will continue to do so here. It's a hobby for me, and I'm not willing to give that up for something as silly as some points that mean nothing anywhere else but here.
Best Regards
~Scotty :)

I see both points of view, as an owner of a large variety of hardware (some of which I haven't even tested yet, and now there is no need to) I can understand objective of the new revision. At the same time, I can tell you there is obviously no need to test older unpopular hardware unless first place is worth 5+ points.

I know the revision isn't completed yet, but here is a good eample... with my 7950GT in an SLI configuration, in 9th place (out of 18 different users) with the current revision it is worth 3.9pts. With the new revision it is only worth .1 pts :( .... I'm hoping it a bug.

The new system will obviously keep the "older popular" hardware .... well popular. This is the only hardware I will seek to bench. I can't see the reason to ever bench any of my other older hardware (50 + processors) unless its ... LOL popular :) So if everyone has this same mentality, older hardware simply wont get benched anymore unless you might have a 1st place "gem" that might get you 1 point :)

Again, I recognize the importance of the new revision. Now I can shop for cheaper "popular" hardware on Ebay. By then we will all be outbidding each other and that old processor isn't so cheap any more...

/end rant ;)

Comment from Bones (Classicplatforms.com) - link
Time: Dec 29, 2009

My feelings are about the same as Mr. Scott's on perhaps why we took such a hit. Racking up a good deal of points without the benefit of newer hardware and without using extreme cooling methods isn't easy and really does speak for the skills within our team.
It probrably did get under the skin of a few teams/members behind us but all I can say to that is simply "Get over it".
We didn't get what we have by lack of skill, rather we got our points as a result of it along with alot of hard work on our part - Nuff said.

I also have a much better understanding of the new points system the way it's listed now and thanks for making it easier to see how it works.

I also will continue to bench and post up to further improve my skills as well.
I'm always looking to push the envelope and better myself. I will say this site has given me the chance to show what I can do along with what I need to do to overcome whatever shortcomings I have.

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

vasgto said:
The new system will obviously keep the "older popular" hardware .... well popular. This is the only hardware I will seek to bench. I can't see the reason to ever bench any of my other older hardware (50 + processors) unless its ... LOL popular :) So if everyone has this same mentality, older hardware simply wont get benched anymore unless you might have a 1st place "gem" that might get you 1 point :)

Again, I recognize the importance of the new revision. Now I can shop for cheaper "popular" hardware on Ebay. By then we will all be outbidding each other and that old processor isn't so cheap any more...

/end rant ;)

We are hoping collecting cups is also an incentive to get a top stop with old, unpopular hardware. There are quite some achievements based on collecting a certain amount of cups, and you've got to admit a profile with lot's of silverware looks attractive, even if he got few points. :cool:

Comment from VasGTO (PURE) - link
Time: Dec 30, 2009

richba5tard said: We are hoping collecting cups is also an incentive to get a top stop with old, unpopular hardware. There are quite some achievements based on collecting a certain amount of cups, and you've got to admit a profile with lot's of silverware looks attractive, even if he got few points. :cool:

Very True, however I'd prefer to advance myself or my team with points. I would bench pmark 2004, pc vantage and sisoftsandra if I wanted cups ;)

Comment from jmax_oc (JMax-Hardware) - link
Time: Dec 30, 2009

I just notice a 'huge' problem. I don't know if someone write it in this topic, so I will post.

Lots of Athlon X2 and more generally AMD's recent CPU can give additionnal cores by unlocking them. That's why lots of X2 with 4 effective cores compete with other regular dual core CPU in WPrime for example. It's not really fair to overclock an Athlon X2 to 5GHz and to be defeated by a stock Athlon X2@X4 at 3GHz.

I also notice a problem with skull and bi cpu platforms. There is a big difference if a member write 2x QX9775 (8 cores) or just QX9775 (4 cores).

I hope you will fix these problems because WPrime WR are a little funny to see.

Good luck for rev3 ;)

Comment from Mr.Scott (Classicplatforms.com) - link
Time: Dec 30, 2009

vasgto said: Very True, however I'd prefer to advance myself or my team with points. I would bench pmark 2004, pc vantage and sisoftsandra if I wanted cups ;)

Right on.:)

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

jmax_oc said: I just notice a 'huge' problem. I don't know if someone write it in this topic, so I will post.

Lots of Athlon X2 and more generally AMD's recent CPU can give additionnal cores by unlocking them. That's why lots of X2 with 4 effective cores compete with other regular dual core CPU in WPrime for example. It's not really fair to overclock an Athlon X2 to 5GHz and to be defeated by a stock Athlon X2@X4 at 3GHz.

I also notice a problem with skull and bi cpu platforms. There is a big difference if a member write 2x QX9775 (8 cores) or just QX9775 (4 cores).

I hope you will fix these problems because WPrime WR are a little funny to see.

Good luck for rev3 ;)

We are aware of this. :) If you have an unlockable cpu, you have to select the amount of active cores. This is not yet implemented, it needs to be rushed in by tomorrow.

Comment from komadyret (411 Overkill) - link
Time: Dec 30, 2009

I'm feeling provoked.

"Why test older unpopular hardware unless first place is worth 5+ points"
Easy; because if just a few persons actually DO bench it, it WILL be worth 5+ points! Just wait and see what happens in the GF4 classes next month as the competition gets rolling. IF hwbot users have some pride in them, they will see that winning a monthly competition will give quite a recognition.

And for the CP guys; obviously there aren't many enough of you benching that classic stuff to rack up the points? You need to fire up some internal competitions to get more users on a particular piece of hardware. If you got 40 benchers and all of then grab a Pentium3 733MHz, the points for that old chip will ramp up quite good!
Or to see it another way; the 300 point hw limit is still in place, it just got a little harder to get there, but unlike earlier you're bound to wind up with a few global points along the way, so your theoretical maximum of points are increased.

And like a few others here have stated; I'm still gonna beat on my old athlons and voodoos and whatever when I have the time to spare, weather it rewards me points, silverware or just the excitement of letting the smoke out of an ancient piece of hardware!

Now, stop seeing the disadvantages (hey, I look at loosing half my points myself, and plummeting from ~500 to ~1000 in the global rankings) and look at the possibilities that opens up. If you bench a good result, the points for that score will increase over time, as you taunt the competition into trying to beat you! lol

rev3 hurts, but it's temporarily, and I see a fine hour for older hardware to gain some ground. There's not need for many results to beat those 2.5 points a #1 submission was worth earlier, and points ramp up faster with # of users. So it's better to get to work, lobby the hwbot crew into putting up a monthly challenge in "your favorite hardware/benchmark" and have some great fun! :D

Comment from VasGTO (PURE) - link
Time: Dec 30, 2009

komadyret said: I'm feeling provoked.

Didn't mean to provoke you... :) You bring up some excellent points, should some competition with older hardware develop, of course I will probably bench it. However, if I am on a "mission" such as hardware masters ranking... Then I see no need to bench something that will only gain .1 pts. I will channel my energy somewhere else.

When I bench a piece of hardware, I look it up in the database. I basically aim for the middle of the pack. Amazingly enough, the middle of the pack seemed to get you about 1 point. If I did better than the middle... Hooray.... I obviously have more points :)

Now it's kinda sad to see that the middle of the pack will probably always be .1pts unless of course the hardware is popular :) I'm not to sure where you draw the line of "popular" but even a category with submissions from 18 different users yielded me .1pts and I was, you guessed it, in the middle of the pack (9th).

Then again, we haven't seen the final revision yet so let's just wait and see. If you haven't noticed, I obviously like to beat the crap out of my hardware and have acquired bunch of it. I will still continue to do so, but the chances of me benching old hardware are very slim. Yes, I will scout the popular old hardware and focus my attention there, but I'm going to need a very good reason to submit an score that's only worth .1pts, heck I was reluctant to submit a score that was worth less than .5 pts :D

Comment from dinos22 (i4memory.com) - link
Time: Dec 30, 2009

LN2 suppliers and pot builders will be happy i think hehehehe
site went down for me now but i noticed Sepron 140 wprime score getting 170+ points lol! Is that because of the popularity of the category or some other reason or some mistake?!?

Comment from VasGTO (PURE) - link
Time: Dec 30, 2009

dinos22 said: LN2 suppliers and pot builders will be happy i think hehehehe
site went down for me now but i noticed Sepron 140 wprime score getting 170+ points lol! Is that because of the popularity of the category or some other reason or some mistake?!?

Cool, I think Fry's still sells those for about 35-40 bucks... I'm pretty sure stage5 of the HWBOT country cup is responsible for 170+ points :)

Comment from Splave (OC Alliance) - link
Time: Dec 30, 2009

is it offline again?

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 30, 2009

It's the global rankings being separated by #CPU_cores responsible for lots of points. The 140 is the best single core CPU, so it's global benefit from all of the global results. Which are mostly, done on single cores :)

Comment from jmax_oc (JMax-Hardware) - link
Time: Dec 30, 2009

dinos22 said: site went down for me now but i noticed Sepron 140 wprime score getting 170+ points lol! Is that because of the popularity of the category or some other reason or some mistake?!?

Yes, because lots of Sempron 140's owners add an extra core to ther CPU. So WPrime is biased by it since they have 2 cores and they compete in single core category (easy to win).

The same problem occurs with X2@X4.

Look my previous post here with RB's answer : http://hwbot.org/forum/showpost.php?p=43359&postcount=491

Comment from matose (lab501.ro) - link
Time: Dec 30, 2009

Is the point calculation working properly now on the test server? Because if I sum all my global points it is well over de 98 that shows in my profile.... :)

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 30, 2009

Still fixing bugs

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

Test server is down to load a fresh dump of hwbot database. It will take a few hours to get back online.

Comment from thebanik (OC Alliance) - link
Time: Dec 30, 2009

So rev3 is still going online from Jan 1?? or it will be delayed if all the bugs are not fixed???

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

It will be delayed if critical bugs are not solved, obviously. Currently still on schedule.

RANKINGS ON DEV SERVER CURRENTLY DOWN, DO NOT CHECK & COMPLAIN ;)

Comment from K404 - link
Time: Dec 30, 2009

Im feeling especially tactless.

@ the moaning ClassicPlatforms guys: if your points are getting slashed, the category is not popular. Not popular= not many results. Whos yardstick of OC skill are you using if theres not many other results to compare it to? Why should you be rewarded if theres little competition? I win the race to the train station every morning. Oh wait... im the only guy taking part! If the top-flight moved away from 32nm + CF-X for a few weeks, and moved to the same platforms, where will your results be?

IMHO you can be kings of nothing with no points or quite possibly drop down the rankings of a popular category, because eventually, no matter what platform you are on, someone comes along who does it better or gets a better chip.

Stop complaining. If you want points, compete for them. If you want the best results in the categories that interest you, you're already there and congratulations on that.

Comment from Monstru (lab501.ro) - link
Time: Dec 30, 2009

Very well said, but to whom??? REV 3 is work in progress, and since the link is public, people are helping to beta-test and solve issues. If you tell people "stop complaining", they will not provide feedback, and that is crucial right now, for the fine tuning of Rev 3.

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

We've had some valuable feedback here, some people found bugs and some pushed us to balance the algoritm better. I'm used to complaints, I understand why. Keep em comming. :D

Comment from Oj0 (PURE) - link
Time: Dec 30, 2009

Just an idea for fun, how about giving the person who comes last in a category 0.1boints? Or add another reward for someone who comes last in 50 benches or something?

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

We already changed the algorithm to give at least 0.1pts to your best hw submission. It's not possible to grind with such a low reward, but it's still better than nothing. :)

Comment from Oj0 (PURE) - link
Time: Dec 30, 2009

Why can't you have one for Least Promising Overclocker? :D Maybe the person with the lowest number of points (with more than 0.0)? Or a certain number of results that don't give points? You have some that are stupidly easy (make your first submission) and some that are ridiculously difficult (1000 gold cups), some things for added entertainment in the middle ground could be fun.

Comment from knopflerbruce (PURE) - link
Time: Dec 30, 2009

Sounds like a decent idea, in some ways:p A bit like the slideshow achievement. However, the reward should be an achievement, not extra boints.

Comment from Oj0 (PURE) - link
Time: Dec 30, 2009

When I said reward I meant achievement :)

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

Like "make 100 submissions without reaching 10 points"?

- edit -
Darn, I would win that in rev3. :D

Comment from K404 - link
Time: Dec 30, 2009

If so, and if you introduce that award, that would be classed as abuse of power. :p "Lets hand out awards that only I can win!!" :p

Comment from Oj0 (PURE) - link
Time: Dec 30, 2009

Make more :D I was honestly hating the move to Rev3 until I saw the new features, now I feel emo when the server goes down and I have to roll back to Rev2 :P

Comment from K404 - link
Time: Dec 30, 2009

I know some current Hardware Masters wont class as such in Rev3. It would be nice if there was some way of highlighting that they were in the HML in Rev2. Some guys will lose their awards with no "memory" of it. :(

Comment from knopflerbruce (PURE) - link
Time: Dec 30, 2009

HWBot is turning into Modern Warfare 2 (achievement part:D)

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

Achievements can be added as we go, they are not tied to an engine upgrade. So for now, we are focussing on ironing out the bugs before we add more achievements. :)

Comment from Oj0 (PURE) - link
Time: Dec 30, 2009

Awesome :)

Comment from Quentj (JMax-Hardware) - link
Time: Dec 30, 2009

vasgto said: Didn't mean to provoke you... :) You bring up some excellent points, should some competition with older hardware develop, of course I will probably bench it. However, if I am on a "mission" such as hardware masters ranking... Then I see no need to bench something that will only gain .1 pts. I will channel my energy somewhere else.

When I bench a piece of hardware, I look it up in the database. I basically aim for the middle of the pack. Amazingly enough, the middle of the pack seemed to get you about 1 point. If I did better than the middle... Hooray.... I obviously have more points :)

Now it's kinda sad to see that the middle of the pack will probably always be .1pts unless of course the hardware is popular :) I'm not to sure where you draw the line of "popular" but even a category with submissions from 18 different users yielded me .1pts and I was, you guessed it, in the middle of the pack (9th).

Then again, we haven't seen the final revision yet so let's just wait and see. If you haven't noticed, I obviously like to beat the crap out of my hardware and have acquired bunch of it. I will still continue to do so, but the chances of me benching old hardware are very slim. Yes, I will scout the popular old hardware and focus my attention there, but I'm going to need a very good reason to submit an score that's only worth .1pts, heck I was reluctant to submit a score that was worth less than .5 pts :D

+1

The fall of points between ranks in unpopular categories is too fast (I don't speak of the points of first place). If you are not in the top5, for many categories, you have 0.1pts, just like the guy who just run the bench, even if your component is overclocked a lot.

Why isn't the fall a little slower ?

And without this point, I think the rev is a good job.

Comment from knopflerbruce (PURE) - link
Time: Dec 30, 2009

I think about 50% of the results will eb awarded with boints, which means that if 5th place is without it, then you need 8 results in total. Normally you don't need to REALLY push to get 4th out of 8 results, but with such a small number of scores you will find a few exceptions;)

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

Interesting stuff for those who want to take a peek under the hood of the hwboint algorithm. You can see the score maxes out at +-1000 participants (weight = 3):


INFO  RankingFacadeImpl - Recalculating global ranks for 159 applications.
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type 3Dmark 2003 ranking 1, weight: 2.7206833, active participating us            ers: 838
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type 3Dmark 2003 ranking 1, weight: 2.7206833, active participating             users: 838
INFO  RankingFacadeImpl - Recalculating 1000 results (2000>3000) of type 3Dmark 2003 ranking 1, weight: 2.7206833, active participating             users: 838
INFO  RankingFacadeImpl - Recalculating 719 results (3000>4000) of type 3Dmark 2003 ranking 1, weight: 2.7206833, active participating users: 838
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type SuperPi, weight: 2.998939, active participating users: 1680
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type SuperPi, weight: 2.998939, active participating users: 1680
INFO  RankingFacadeImpl - Recalculating 1000 results (2000>3000) of type SuperPi, weight: 2.998939, active participating users: 1680
INFO  RankingFacadeImpl - Recalculating 1000 results (3000>4000) of type SuperPi, weight: 2.998939, active participating users: 1680
INFO  RankingFacadeImpl - Recalculating 1000 results (4000>5000) of type SuperPi, weight: 2.998939, active participating users: 1680
INFO  RankingFacadeImpl - Recalculating 1000 results (5000>6000) of type SuperPi, weight: 2.998939, active participating users: 1680
INFO  RankingFacadeImpl - Recalculating 646 results (6000>7000) of type SuperPi, weight: 2.998939, active participating users: 1680
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type 3Dmark 2005 ranking 1, weight: 2.6529157, active participating users: 802
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type 3Dmark 2005 ranking 1, weight: 2.6529157, active participating users: 802
INFO  RankingFacadeImpl - Recalculating 1000 results (2000>3000) of type 3Dmark 2005 ranking 1, weight: 2.6529157, active participating users: 802
INFO  RankingFacadeImpl - Recalculating 791 results (3000>4000) of type 3Dmark 2005 ranking 1, weight: 2.6529157, active participating users: 802
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type PiFast, weight: 2.8063583, active participating users: 897
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type PiFast, weight: 2.8063583, active participating users: 897
INFO  RankingFacadeImpl - Recalculating 905 results (2000>3000) of type PiFast, weight: 2.8063583, active participating users: 897
INFO  RankingFacadeImpl - Recalculating 785 results (0>1000) of type PCMark 2005 ranking 1, weight: 0.27931672, active participating users: 162
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type CPU-Z, weight: 2.9990914, active participating users: 1703
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type CPU-Z, weight: 2.9990914, active participating users: 1703
INFO  RankingFacadeImpl - Recalculating 1000 results (2000>3000) of type CPU-Z, weight: 2.9990914, active participating users: 1703
INFO  RankingFacadeImpl - Recalculating 1000 results (3000>4000) of type CPU-Z, weight: 2.9990914, active participating users: 1703
INFO  RankingFacadeImpl - Recalculating 1000 results (4000>5000) of type CPU-Z, weight: 2.9990914, active participating users: 1703
INFO  RankingFacadeImpl - Recalculating 1000 results (5000>6000) of type CPU-Z, weight: 2.9990914, active participating users: 1703
INFO  RankingFacadeImpl - Recalculating 863 results (6000>7000) of type CPU-Z, weight: 2.9990914, active participating users: 1703
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type wPrime 32m ranking 1, weight: 1.1864797, active participating users: 437
INFO  RankingFacadeImpl - Recalculating 811 results (1000>2000) of type wPrime 32m ranking 1, weight: 1.1864797, active participating users: 437
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type wPrime 32m ranking 2, weight: 1.7989993, active participating users: 560
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type wPrime 32m ranking 2, weight: 1.7989993, active participating users: 560
INFO  RankingFacadeImpl - Recalculating 942 results (2000>3000) of type wPrime 32m ranking 2, weight: 1.7989993, active participating users: 560
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type wPrime 32m ranking 4, weight: 2.613941, active participating users: 784
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type wPrime 32m ranking 4, weight: 2.613941, active participating users: 784
INFO  RankingFacadeImpl - Recalculating 185 results (2000>3000) of type wPrime 32m ranking 4, weight: 2.613941, active participating users: 784
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type wPrime 1024m ranking 1, weight: 0.6257682, active participating users: 302
INFO  RankingFacadeImpl - Recalculating 288 results (1000>2000) of type wPrime 1024m ranking 1, weight: 0.6257682, active participating users: 302
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type wPrime 1024m ranking 2, weight: 1.1816528, active participating users: 436
INFO  RankingFacadeImpl - Recalculating 935 results (1000>2000) of type wPrime 1024m ranking 2, weight: 1.1816528, active participating users: 436
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type 3Dmark 2001 ranking 1, weight: 2.699528, active participating users: 826
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type 3Dmark 2001 ranking 1, weight: 2.699528, active participating users: 826
INFO  RankingFacadeImpl - Recalculating 1000 results (2000>3000) of type 3Dmark 2001 ranking 1, weight: 2.699528, active participating users: 826
INFO  RankingFacadeImpl - Recalculating 843 results (3000>4000) of type 3Dmark 2001 ranking 1, weight: 2.699528, active participating users: 826
INFO  RankingFacadeImpl - Recalculating 939 results (0>1000) of type 3DMark Vantage - Performance ranking 2, weight: 1.4394217, active participating users: 488
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type 3Dmark 2001 ranking 2, weight: 0.69149905, active participating users: 321
INFO  RankingFacadeImpl - Recalculating 310 results (1000>2000) of type 3Dmark 2001 ranking 2, weight: 0.69149905, active participating users: 321
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type 3Dmark 2003 ranking 2, weight: 1.294217, active participating users: 459
INFO  RankingFacadeImpl - Recalculating 542 results (1000>2000) of type 3Dmark 2003 ranking 2, weight: 1.294217, active participating users: 459
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type 3Dmark 2005 ranking 2, weight: 1.4041626, active participating users: 481
INFO  RankingFacadeImpl - Recalculating 598 results (1000>2000) of type 3Dmark 2005 ranking 2, weight: 1.4041626, active participating users: 481
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type SuperPi 32m, weight: 2.900649, active participating users: 1001
INFO  RankingFacadeImpl - Recalculating 1000 results (1000>2000) of type SuperPi 32m, weight: 2.900649, active participating users: 1001
INFO  RankingFacadeImpl - Recalculating 1000 results (2000>3000) of type SuperPi 32m, weight: 2.900649, active participating users: 1001
INFO  RankingFacadeImpl - Recalculating 242 results (3000>4000) of type SuperPi 32m, weight: 2.900649, active participating users: 1001
INFO  RankingFacadeImpl - Recalculating 746 results (0>1000) of type PCMark 2004, weight: 0.15909067, active participating users: 72
INFO  RankingFacadeImpl - Recalculating 1000 results (0>1000) of type Aquamark ranking 1, weight: 2.340225, active participating users: 688

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

dev server online with recent database dump...

Comment from PeterStoba (Bench Tec UK) - link
Time: Dec 30, 2009

I've gained 0.8 boints from what I'd previously dropped to. Has there been a change, or is it just because more results have been submitted? :)

Comment from TiN (XtremeLabs.org) - link
Time: Dec 30, 2009

Most of ppl lost points quite noticeably with fresh db :) Some bug?

Comment from Gomeler (PURE) - link
Time: Dec 30, 2009

Server crashed.

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 30, 2009

Server has been brought down, need time to develop.

Comment from Splave (OC Alliance) - link
Time: Dec 30, 2009

goes live jan 1?

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 30, 2009

We're aiming for tonight

Comment from Dualist (Bench Tec UK) - link
Time: Dec 30, 2009

woo hoo..!! :woot:

Comment from Brian y. (PURE) - link
Time: Dec 31, 2009

and so it begins.....

Comment from Turrican (Team OCX) - link
Time: Dec 31, 2009

yep :D

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 31, 2009

I think I wrote the longest update script ever. Already crunching for 4 hours and still not ready. :D I think ± 2 or 3 hours to go.

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 31, 2009

I'm off making a couple liters of Coffee ...

Comment from richba5tard (hwbot.org benching team) - link
Time: Dec 31, 2009

I think you better go to sleep and wake up at 6am. :)

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 31, 2009

If there's one day I should be active at hwbot ... it's probably today LOL

Comment from Antinomy (Hardware Hackers) - link
Time: Dec 31, 2009

Yeah, then you'll go to bed and wake up next year :D

And I with my teammates agree about the boints fall after first place too quick. If you look in the CPU categories, it's very sad to be even in top10. There's only about 10-15 exclusions to this rule in the CPU's. And yes, the answer "no problem, be the first" makes sense, but it's not always a "no problem" one ;)

Comment from Turrican (Team OCX) - link
Time: Dec 31, 2009

argh.
damn, soon i'll need a new F5 button. :D

Comment from TiN (XtremeLabs.org) - link
Time: Dec 31, 2009

Turrican

:eek::D

You are not alone here

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 31, 2009

Still bugfixing, but looking better and better :-)

Comment from Massman (Madshrimps Belgium OC Team) - link
Time: Dec 31, 2009

Closing thread.

Bugfixes => http://hwbot.org/forum/showthread.php?t=4465

Please log in to comment.

Navigation

Competitions

Identification

Advertisement

Quick search

Compare processors

Compare videocards

HWBOT benchmarks

Most popular

Overclockers League

Advertisement

Hardware masters

Community rankings

Country rankings

HWBOT records

Important! Opinions regarding HWBoints algorithm revision

Comments