New site, new year... new benchmark! Introducing MaxxMem

The author and creator of MaxxPI² has been working hard to get one of his benchmark applications compatible with the HWBot site. The first graduate of this training is the MaxxMem² benchmark which tests your computer’s raw memory performance, combining copy, read, write and latency tests into one global score. From the application you can directly submit your MaxxMem² record to the HWbot website (at the HWbot landing page you’ll get a chance to attach a screenshot showing some extra details with CPU-Z).

maxxmem_pre.png

Please help us test and torture this latest benchmark, download the latest version here (MaxxPI² website).

  • - As this application is brand spanking new, a test run will be held before assigning HWboints
  • - Golden/Silver/Bronze cups and medals can be earned
  • - Please report any bugs or strange things you encounter to help us improve this benchmark where necessary
  • - You can save your score offline and submit it later, on a machine which has internet access; don’t forget to also save a valid screenshot of your result
  • - The MaxxMem² benchmark will ONE global ranking and THREE hardware rankings based on DDR/DDR2/DDR3, for time being.

31

Belgium richba5tard says:

There seems to be a bug in thread creation. Your post isn't showing. :)

Germany Don_Dan says:

Will there be different rankings for DDR,DDR2 and DDR3?

Belgium richba5tard says:

This is incorrect. It's configured as a MEM benchmark, so separated by DDR/DDR2/DDR3...

Belgium richba5tard says:

global: 1 ranking hw: split up in ddr1/ddr2/... For hwbot v3.5, we could split cpu-z, maxxmem, max reference clock up into cpu architecture.

Belgium leeghoofd says:

do we get seperate cpu rankings in the near future ? fully I7 territory at the moment :p

Germany Don_Dan says:

Yeah, for me the type of memory ( eg. DDR2 in my case ) would not sticky at first, I noticed it a few minutes ago and edited my score. Most likely the same for all the other scores.

Belgium Massman says:

jmke said: or if/when we get a memory database up and running, link the HW points to memory brand & model :D


The memory database is set up to support Manufacturer and Product Group. So no seperate products

Belgium leeghoofd says:

App doesn't support 0.5 multies for Core 2 Duo here... ( invalid integer error)

If I put CPU at 7 or 8 multi all is okay... anyone else experiencing this ?

Belgium richba5tard says:

Same error here. 10.5 is invalid integer error on a core 2 duo.

Romania tavirosu says:

MaxxMem shows error at the CPU! Im running my i7 at 4009(191x21) but in the software it shows 4180Mhz!

Belgium leeghoofd says:




Siv reads out correctly John, 8.5 x 500 here 4.25Ghz, Maxmem errors

Romania tavirosu says:

21x191=4009 Mhz but maxxmem shows 4180 Mhz! So all my results are shown as invalid. Please advise me what to do! OS: is Win7 x86

Romania tavirosu says:

I edited but the submission afterwards its shows invalid.

Belgium richba5tard says:

The checksum validation has not been activated, sorry. it will always display invalid for now.

Romania tavirosu says:

Ok thanks then! Just to let you guys know the applications seems to have a bug reading the right CPU value for the i7 920. 4025 Mhz(real value) shows at 4642 Mhz in MaxxMem.

Belgium HybridChiller says:

when does it gets global points or hardware ?

when there are enough submissions ?

just got the nr1 global award for it

Romania tavirosu says:

You sure that the software doesnt have a bugg? It shoulnd scale with cores/threads. It should only scale with the memory and yet people with better memory ram then you got lower scores.

Belgium HybridChiller says:

uploading score with online submission is always invalid,
i posted some scores (4) and all invalid,
still the run is good en succes full, en screenshot is present

Brazil WarriorSl says:

should also separate into amd and intel, since amd controller inside the cpu have some great limitations for memory speed besides intel.

Brazil Dan Jacques says:

I had the same problem that leeghoofd. Apparently the bench has a problem with multipliers x.5 (ie my case: 9.5)

Brazil Dan Jacques says:

I had the same problem that leeghoofd. Apparently the bench has a problem with multipliers x.5 (ie my case: 9.5)

Russian Federation Mefist0 says:

Very nice new design of the site! Respect for Admins, and Happy New Year!

Belgium Massman says:

tavirosu said: You sure that the software doesnt have a bugg? It shoulnd scale with cores/threads. It should only scale with the memory and yet people with better memory ram then you got lower scores.


It's the memory latency sub-test result which is strange: 11ns is 3x lower than similar clocked setups.

Chiller: did you do anything special or ran from ramdisk?

Belgium leeghoofd says:

11ns is a bit too low for latency don't ya think Johan :p

Happy New Year all

Belgium HybridChiller says:

run more then one test,
and everytime the same result 4000

keep running more test

Belgium HybridChiller says:

done more testing,
and even on less cores,
it seems that with less cores active on the cpu, your get some more score, but now for me, the scores are very high.

Belgium leeghoofd says:

Less real cores or disabling HT ? Need to fire up the I7 rig again it seems and build out the AMD

Georgia westsider says:

i ve got some bug
hm
maybe not but i still right it here

if i bench for example 2000MHz and get xx.x latency and use setfsb and go over 2020MHz and rebench it the program gives worse NS results
but if i turn off and restart the benchmark it reads the correct memory speed and gives better latency
i think somehow this program code isnt directly connected on ram performance
maybe 70% ram performance and 30% of what this software knows this ram can do
such results is it showing

and bandwidth is ok
but NS calculator is somehow not Ok
i get such feeling just for now
different results every time i rebench with same config
sometimes latency goes too much down and gives very bugged result

this software needs a lot of work to improve

but why is 75% of score from latency?
is bandwith not important?
i mean with very bad bandwidth and good latency i can get very good points
hows that?

and i7 clock detector doesnt work correctly
if i force the software that it doesnt detect my hardware i can have 500 000+ points as result

please
fix fix fix

too early for hwpoints for this bench

Belgium leeghoofd says:

Weird readout error we spotted on the LCC compo : user has a Xeon W3540 CPU, max multi of 22X, he used the Turbo mode so the board gave him the 23X multi, yet maxxmem reads out 24X... I asked the person in question to test other multies with turbo on to see if it misbehaves...


New Zealand packet says:

change your read to write times to 6 or 7 and tweak tRFC thats the important one lower the better :)

Belgium leeghoofd says:

Update on the readout bug :

"With turbo boost enabled or disabled and 21x and below (20x, 19x,18x....) maxmemm reads spot on with cpuz.

If I choose the 22x multi with turbo boost on in bios it will show that I chose that but will boot with a 23x. It will run at 23x and cpuz shows this. Maxmemm will show the cpu speed being that of 24x even though you wont be able to run 24x except in the scenarios posted above.

Now if I set 22x turbo boost off, it boots at 22x and cpuz and maxmemm report correctly with 22x speeds.

23x is only available with turbo boost on and is reported in cpuz correctly as a 23x and maxmemm reports 24x not agreeing with cpuz.

So to sum it up. Either 22x or 23x with turbo boost on will be read as 24x by maxmemm even though in both cases your running at 23x.
22x turbo boost disabled and any other lower multi with turbo boost on or off will be reported properly.

Sorry for my poor sentence structure.

I was gonna post this but figured if you want you can throw it in the thread.

Also another issue with maxmemm while I was playing around with this stuff is that once its open you can change your settings with ET6 (setfsb for gigabyte boards) and change it back and since maxmemm appears not to refresh it doesnt show the changed speeds. This wouldnt matter for overall scores except for limited competitions like this."

Thx to Madmaxx22 for going deeper into the problem... Can anyone transfer this bug to the creator plz ?

Please log in or register to comment.