OC League Update: ESPORTS Points, WR Points Downgrade, Benchmark Points Upgrade

Following up on the recent discussions sparked from community feedback, we inform you that the following adjustments to the Overclockers Leagues will take place in the coming weeks.


1/ Introduction of the ESPORTS Points

For a long time competition points were part of the OC League equation. When OC-ESPORTS was introduced, we had to disable the points due to an unforeseen imbalance. Over the past couple of weeks, we tested re-introducing the competition points. Based on the feedback, we will proceed by pushing the points to production. To clarify the points’ origin – the overclocking competitions at OC-ESPORTS.io – they will be referred to as “ESPORTS Points”.

The new Overclockers League algorithm is now: SUM of [TOP15(GL+WR) + TOP20(HW) + TOP15(EP)], with competitions applicable for EP if closed one year or less to date.


2/ World Record Points Downgrade

The World Record Points were introduced as a patch for low-scoring world record benchmark submissions. For benchmark categories where the world record is obtained with unpopular hardware (eg. 4-Way SLI or multi-CPU systems), a total of 100 points is awarded in addition to the points generated from the Global Points algorithm. In effect, in those benchmark categories the World Record is awarded approximately as much as the most popular global ranking (usually 1xGPU).

For the benchmarks Aquamark3, 3DMark01, 3DMark05, and 3DMark06 our community noticed that the World Record Points are awarded to the most popular global rankings, resulting in an overly valued 1xGPU Global First Place. For example, the world record for Aquamark3 (1xGPU stock cooling) is worth 250 points whereas the world record for Fire Strike Extreme (4xGPU liquid nitrogen) is worth 148 points.

Based on the community feedback, we will remove the WR Points from the benchmark categories where the patch is no longer required to ensure a balanced point algorithm. Those benchmark categories are: Aquamark3, 3DMark01, 3DMark05, and 3DMark06.


3/ Benchmark Points Upgrade

Based on the information provided by the community’s benchmark beta testing and validation by our software team, we will upgrade the following benchmarks with points.

  • WR Points: Geekbench3 – Multi Core
  • GL Points: Geekbench3 – Multi Core, MaxxMem Read Bandwidth
  • HW Points: Geekbench3 – Multi Core, MaxxMem Read Bandwidth, 3DMark Ice Storm, 3DMark Ice Storm Extreme, 3DMark Ice Storm Unlimited, 3DMark Cloud Gate, 3DMark Sling Shot, 3DMark Fire Strike Ultra, Catzilla 576P, Catzilla 1080P


The transition will take a couple of weeks as we update the explanations, links, and update the point algorithms. Thank you for your understanding!


56

Argentina Alan_Alberino says:

Great news :)

TaPaKaH says:

Maxxmem is not even measuring the current system performance and you're giving it global points? Seriously? Competition points could also use some refinement (if not yet done) if winning one stage (in a multi-stage comp) can give as many points as winning a whole comp.

United Kingdom ObscureParadox says:

Agreed on Maxxmem, it's not even a memory benchmark so I'm not exactly sure why it deserves globals.....

Germany der8auer says:

I have to agree on what you say about MaxxMem. Not my favorite bench at all. It can be tricked in several ways with BCLK up- and downclocking. Not really measuring anything I think. I'm also not sure about the Geekbench.

France Strat says:

Also agree. When Single sided MFR wins you know the benchmark is clearly not reliable. :D

websmile says:

Funny that at a time when people can get 1k (temdency rising) combined at globals by benching just one single 3 months old bench in 20 categories, all hammer on 5 years old stuff because it is rewarded 1x globals^^ - apart from this, I agree that there has to come a more reliable memory bench in the future if maxmem is not improved

United States Gunslinger says:

I kind of like most of the changes, kudos.

MaxxMem.....not so much. :confused:

zeropluszero says:

"community feedback"

Belgium Massman says:

Guys, If you have problems with the announcement feel free to argue why. "I don't like this benchmark", "it's not a memory benchmark", "single mfr wins" and so on, are not good arguments to not bump it. Historically, we know that a benchmark is put through its paces when points are enabled and beta-testing it doesn't always reveal all bugs. So it's possible we will have this issue with a benchmark in the future. MaxxMem Read Bandwidth has been used in competitions and I haven't seen much complaints about it so far. So if there's anything that makes this an unreliable benchmark, please post below and we can still scratch if off the list.

zeropluszero says:

I dont understand why you'd give hw points to all different flavours of 3dmark13, they're basically all the same. why not give points to all the different presets of the previous versions while you're at it? just give wr, gbl and hw points to FS extreme -or- ultra and be done with it. same goes for points in cinebench11 and 15, they're essentially the same benchmark.

Belgium Massman says:

Why wouldn't we give points to those benchmarks? Some people enjoy running them, they're stable, and the points are a nice reward for the effort. It opens up points for people who run unusual benchmarks.

United Kingdom ObscureParadox says:

zeropluszero said: I dont understand why you'd give hw points to all different flavours of 3dmark13, they're basically all the same.
why not give points to all the different presets of the previous versions while you're at it?

just give wr, gbl and hw points to FS extreme -or- ultra and be done with it.

same goes for points in cinebench11 and 15, they're essentially the same benchmark.


So's wPrime 32/1024, then super pi 1M and 32M but nobody complains about that.

Same benchmark argument isn't really valid here.

For Maxxmem I'd like to see a benchmark where it really is the memory and IMC being tested and not..... Well whatever the hell it actually does I have no idea, but it isn't memory performance :P

I don't really have a solution to offer for that since I don't really know many memory benchmarks personally, but I'm sure a user on here can find one to suggest for beta testing.

For example, Lams result for the G.SKill comp was 25.5K with a G3240 and ram @1652MHz, I get around 23K running my 24/7 system with quad channel ram and no tweaks. This ram is clocked at 2400MHz too so it's no slouch. We all know that there is no way the dual channel system really has as much bandwidth as a high clocked quad channel system ;)

Belgium Massman says:

There aren't that many alternatives for memory benchmarks. Maybe AIDA64 is a decent alternative. Regarding your example: I believe this is because MaxxMem is a single threaded benchmark application. As in: it doesn't use all cores to measure the memory bandwidth. As a result, quad channel will not be as effective. AIDA64 used to be single threaded too and the read bandwidth figures were very close to MaxxMem's. A couple of years ago, they changed the way to measure memory bandwidth to multi-threaded, hence why it's showing massive bandwidth. Source: http://forums.aida64.com/topic/1326-new-cache-and-memory-benchmarks-in-aida64-v300/

But more importantly, we've replaced the outdated set of memory benchmarks with brand new ones. The new bandwidth (read, write copy) benchmarks now use multiple threads to squeeze out every last bit of performance from the caches and the memory modules. On modern multi-core processors, using the old single-threaded benchmarks you couldn't see the actual memory bandwidth, but only the memory bandwidth available for single-threaded applications. With the new benchmarks you will however get considerably higher scores, much closer to the theoretical memory bandwidth available. It is especially true for 3-channel and 4-channel memory configurations, such as Intel X58 and X79 based high-end desktop systems; and also for NUMA-enabled multi-socket systems, such as 2- and 4-way AMD Opteron and Intel Xeon based servers and workstations. For example: Core i7-3960X with X79 chipset and 4-channel DDR3-1600: - AIDA64 v2.85 Memory Read: 16825 MB/s [ old ] - AIDA64 v3.00 Memory Read: 45640 MB/s [ new ]

United States rtsurfer says:

zeropluszero said: I dont understand why you'd give hw points to all different flavours of 3dmark13, they're basically all the same.
why not give points to all the different presets of the previous versions while you're at it?

just give wr, gbl and hw points to FS extreme -or- ultra and be done with it.

same goes for points in cinebench11 and 15, they're essentially the same benchmark.


Agreed totally. No need to enable points on all different 3D Marks. We will have a new one coming in a few months after Windows 10/DX12 launch anyways.

Personally I'd like to see points removed from regular FS & added to Ultra. Lot of people might hate me for saying this.

If I wanted to Bench a GPU Bench heavily dependent on CPU, I'd start benching legacy Benchs.

Just my 2 Cents. Peace.

United Kingdom ObscureParadox says:

Massman said: There aren't that many alternatives for memory benchmarks. Maybe AIDA64 is a decent alternative.

Regarding your example: I believe this is because MaxxMem is a single threaded benchmark application. As in: it doesn't use all cores to measure the memory bandwidth. As a result, quad channel will not be as effective.

AIDA64 used to be single threaded too and the read bandwidth figures were very close to MaxxMem's. A couple of years ago, they changed the way to measure memory bandwidth to multi-threaded, hence why it's showing massive bandwidth.

Source: http://forums.aida64.com/topic/1326-new-cache-and-memory-benchmarks-in-aida64-v300/


AIDA64 was the only other one that came to mind for me also but isn't it a paid benchmark?

So that's a little disappointing for finding a new benchmark.

Unless one of the clever people on the forum want to try and make a new one ;)

United States steponz says:

Hmm.. not sure about this.. Fs,fse,fsu should all be global and Wr.. they are different. ultra needs to be there... it's becoming more relevant every hardware refresh. .. Why not add Vantage extreme and 3d11 extreme.. that would be innteresting.. I honestly think the whole points system should change.. it should be done by how hard it is to achieve.. ALl 3d should be higher than cpu points.. maybe do it by how many pots are needed.. or components.. cpu and mem would be higher than cpu. Cpu and gpu would be very high and multi gpu would be high.. just not as much as Single.. Anyways... figure something out.. it just seems way out of wack right now..

United Kingdom ObscureParadox says:

steponz said: Hmm.. not sure about this..

Fs,fse,fsu should all be global and Wr.. they are different.
ultra needs to be there... it's becoming more relevant every hardware refresh. ..

Why not add Vantage extreme and 3d11 extreme.. that would be innteresting..

I honestly think the whole points system should change..
it should be done by how hard it is to achieve..

ALl 3d should be higher than cpu points.. maybe do it by how many pots are needed.. or components.. cpu and mem would be higher than cpu. Cpu and gpu would be very high and multi gpu would be high.. just not as much as Single..

Anyways... figure something out.. it just seems way out of wack right now..


It's impossible to go by difficulty rating because what may be hard for you might be simple to me and vice versa.

Has to be a universal system where the rules are consistent and I think it currently is, more submissions in "x" benchmark = more points given, so you kind of already have a difficulty rating since the hardest ones are generally worth 49.9 HW points

zeropluszero says:

ObscureParadox said: So's wPrime 32/1024, then super pi 1M and 32M but nobody complains about that.

Same benchmark argument isn't really valid here.

For Maxxmem I'd like to see a benchmark where it really is the memory and IMC being tested and not..... Well whatever the hell it actually does I have no idea, but it isn't memory performance :P

I don't really have a solution to offer for that since I don't really know many memory benchmarks personally, but I'm sure a user on here can find one to suggest for beta testing.

For example, Lams result for the G.SKill comp was 25.5K with a G3240 and ram @1652MHz, I get around 23K running my 24/7 system with quad channel ram and no tweaks. This ram is clocked at 2400MHz too so it's no slouch. We all know that there is no way the dual channel system really has as much bandwidth as a high clocked quad channel system ;)

I'm all for removing wPrime 32 and superpi 1m as well.
lams a known cheater so...
rtsurfer said: Agreed totally. No need to enable points on all different 3D Marks. We will have a new one coming in a few months after Windows 10/DX12 launch anyways.
Personally I'd like to see points removed from regular FS & added to Ultra. Lot of people might hate me for saying this.
If I wanted to Bench a GPU Bench heavily dependent on CPU, I'd start benching legacy Benchs.

Thanks.
steponz said: Hmm.. not sure about this..
Fs,fse,fsu should all be global and Wr.. they are different.
ultra needs to be there... it's becoming more relevant every hardware refresh. ..
Why not add Vantage extreme and 3d11 extreme.. that would be innteresting..
I honestly think the whole points system should change..
it should be done by how hard it is to achieve..
ALl 3d should be higher than cpu points.. maybe do it by how many pots are needed.. or components.. cpu and mem would be higher than cpu. Cpu and gpu would be very high and multi gpu would be high.. just not as much as Single..
Anyways... figure something out.. it just seems way out of wack right now..

sure, they're all different, but how different are we talking?
I'm cool with adding one FS/FSE/FSU, but not all three surely.

Netherlands rsnubje says:

ObscureParadox said: AIDA64 was the only other one that came to mind for me also but isn't it a paid benchmark?

So that's a little disappointing for finding a new benchmark.

Unless one of the clever people on the forum want to try and make a new one ;)


You can do the memory benchmark in the free version too, except there will be some scores greyed out. Only the read bandwidth is shown.


TaPaKaH says:

Massman said: If you have problems with the announcement feel free to argue why.
...
MaxxMem Read Bandwidth has been used in competitions and I haven't seen much complaints about it so far. So if there's anything that makes this an unreliable benchmark, please post below and we can still scratch if off the list.
I don't mind having more benchmarks being awarded global / hardware points, but Maxxmem can't even be called a system benchmark. As Roman already mentioned, it doesn't react to the frequency/setting changes you perform in the OS. This leaves a whole lot of "efficiency" loopholes, which people will eventually explore now that this "benchmark" earns globals (we've seen this happen many times in the past, haven't we?).

Another thing that bothers me is seeing lots of hwbot staff members pop up in this thread and disagree with (some of) the changes, which kind of suggests that things were not discussed internally before getting actually implemented.

zeropluszero says:

I'd be surprised if anything was discussed internally.

France Taloken says:

I tested Geekbench yesterday evening, well, a bench you have to pay to launch it in 64 bits -> without me. About MaxxMem Read, it fill a bit the lack of Memory-category bench, but it's not stressful, doesn't push the tech in his limits. In a 4~5G spi32m, ram skill is über-important, and it really require a strong comprehension of every settings and timings. For 3DMark, well not necessary, keep the light ones like Cloud Gate, Ice Storm and so for challenges. It's not revelant to use dozens of liters of LN2 for these.

Belgium Massman says:

Received some information ... let's put a hold on MaxxMem for now.

United States steponz says:

So Geekbench on 32 bit only? or can 64bit be used? Same rules apply with Cores? No windows 8 and up right?

SkaL- says:

Geekbench is not so bad :)

United States steponz says:

It's a pretty boring benchmark.. and it's not very clear.. so we have to have Internet access to submit?

GENiEBEN says:

steponz said: It's a pretty boring benchmark.. and it's not very clear.. so we have to have Internet access to submit?
You need internet access if you're not willing to pay for the application so it displays the score offline, just like with the 3dmark's.
steponz said: So Geekbench on 32 bit only? or can 64bit be used? Same rules apply with Cores? No windows 8 and up right?
Any. Hold on for confirmation on Win8, it was ok last time I checked it.

United States steponz says:

Can you run both 32 bit and 64 bit foe scores? Seems you can't do 64 bit with free.. likely be slower ..?? ..no???

Belgium Massman says:

Since some people are confused about the downgrade of the WR points for the legacy benchmarks, here's a comparison of before and after of the amount of points available for the global benchmarks.

For 1xGPU and 1xCPU




For 1xGPU and 4xGPU




Full List of GL+WR benchmarks

(click for full list)


Belgium Massman says:

3DMark06 1xGPU was 100pts more than 3DMark03 1xGPU

United States steponz says:

Way too many points for 2d... 3d should be up there more..

United States steponz says:

Still waiting to know if you 4 bit can used for geekbench..

Belgium Massman says:

Yup, 64bit allowed :)

United States rtsurfer says:

Does the Class Column mean # of Cores of Cpu or # of GPUs.?

United States steponz says:

The xtu ones are so out of whack... compared to the others... people complain about the legacies and thats cool?
Honestly I'm not even sure what to bench anymore to be in these rankings...

Guess Ill stick to single....

Belgium Massman says:

rtsurfer said: Does the Class Column mean # of Cores of Cpu or # of GPUs.?


For 3D benchmarks it's amount of GPU cores, for 2D benchmarks it's amount of CPU cores

steponz said: The xtu ones are so out of whack... compared to the others... people complain about the legacies and thats cool?
Honestly I'm not even sure what to bench anymore to be in these rankings...

Guess Ill stick to single....


Define "out of whack".

The algorithm to calculate the points of a submission is based on two objective parameters:

1) the position of that submission within the ranking (eg. 1st, 2nd)
2) the amount of competitors in the ranking (eg. 5ppl, 1000ppl)

The higher a result ranks in a category with more competitors, the more points the submission will get.

This is the way we currently quantify the "value" of a submission. We're always open to suggestions on how to do it better, but keep in mind that we have thousands of hardware items in our database. The method of calculating the points must be not only good for the latest generation hardware, but also for the Pentium 3, Athlon 64, etc categories.

Germany der8auer says:

Massman said: Since some people are confused about the downgrade of the WR points for the legacy benchmarks, here's a comparison of before and after of the amount of points available for the global benchmarks.

pictures


Looks much better and balanced now. I like it :)

Poland Nifir says:

I don’t understand this point calculation system. If you calculate sum of my 15GL points and 20HW you get ~500points, but where are the CP? And which points should be included in the League points:
“Points gained for your rank in official competitions” or “Points gained by participating in official competitions”?

France Taloken says:

Occasion to ask for the third time finally expecting a answer : What is the difference between OC-Esports Points and Competition Points ? The second hadn't to be replace by the first ? double usage here.

United States Splave says:

can I be first for 1 day? thanks :)

websmile says:

No :)

United Kingdom 8 Pack says:

Legacy are tough benches to get working well so should be as is.

Geek bench and maxxmem oh dear such guff.

United States Planet says:

Splave said: can I be first for 1 day? thanks :)


Sure you just need to pay off PJ

United Kingdom ObscureParadox says:

Cpt.Planet said: Sure you just need to suck off PJ


Fixed that for you ;)

Belgium Massman says:

Wow, that's pretty offensive language ...

United States jpmboy says:

Changes make sense. Good decision, crisp implementation Massman.

United States rtsurfer says:

Well I just noticed they added GeekBench which is limited to Win7 & below. That isn't a good decision IMO.

France Taloken says:

And 64 bits version is ofc better than 32 bits. "Achievement Get : Introduce a Pay2Win benchmark"

Germany der8auer says:

I don't get this. People have thousands of euros to spend on hardware but complain about 9 euro for a benchmark. Also you don't have to use this bench if you don't like it. Only 15 GL + 20 HW count to the profile total and there is a big load of other benchmarks to fill the profile. Nobody forces you to buy the benchmark.

TaPaKaH says:

How many people actually have thousands to spend on hardware, again? Also, if you spend X euros on hardware, you can get that money back by reselling. With benchmarks it is not the case.

You have to draw a line somewhere. Before you know it, this attitude might evolve in people being forced to spend hundreds/thousands on software, which is completely useless outside of this website.

United States Mr.Scott says:

Charge for bench? How'd that work out for Catzilla again?

zeropluszero says:

How did it work out for Futuremark again?
oh yeh, really well.

Belgium Massman says:

Guys, If you don't like the benchmark, don't pay. There are plenty of rankings available with free benchmarks. If you do like the benchmark, or you want to compete for the points, then the cost really is not that high. Really, no one is forcing you to do anyway. Enjoy overclocking like you want to enjoy it ...

Australia newlife says:

Whatever happened to the ESport points?

Australia JunkDogg says:

newlife said: Whatever happened to the ESport points?


No reply to this??

Or this...

http://forum.hwbot.org/showthread.php?t=139777

????

So what happened?

Not enough 'community feedback'?

Belgium Massman says:

It's all tested on UAT (test environment), but I need the table on the User > Points page to display the competitions that are contributing to the OC League. Which currently is not the case.

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)