Announcing release: standalone memory tester for NVIDIA GPUs

Moderators: slegrand, Site Moderators, PandeGroup

Announcing release: standalone memory tester for NVIDIA GPUs

Postby ihaque » Tue Apr 28, 2009 5:39 pm

We've just released a dedicated memory tester for NVIDIA CUDA-capable GPUs. Binaries for Windows, Linux, and OS X are all available, as is (LGPL-licensed) source code for those of you interested in doing additional development work (for example, GUI frontends).

The goods are available at https://simtk.org/home/memtest.

Update 2009-05-05: The binaries are now available at http://folding.stanford.edu/English/DownloadUtils, so you no longer need to register at SimTK to get working copies of MemtestG80. Source code is still hosted at SimTK.

Update 2009-04-29: For those of you who've seen OCCT's CUDAMemtest tool, we've found that MemtestG80 is more effective at detecting errors, especially those induced by overclocking. See viewtopic.php?f=52&t=9767#p96697 for details.


From Vijay's blog post:

MemtestG80 is a software-based tester to test for "soft errors" in GPU memory or logic for NVIDIA CUDA-enabled GPUs. It uses a variety of proven test patterns (some custom and some based on Memtest86) to verify the correct operation of GPU memory and logic. It is a useful tool to ensure that given GPUs do not produce "silent errors" which may corrupt the results of a computation without triggering an overt error.


Basically, the idea is that we wanted to put out a code to test GPU memory that's roughly equivalent to Memtest on CPUs. If you run FAH heavily on a GPU, it's a good idea to check out your GPU memory, just as one would run tests on CPU memory.

For now, this is being hosted on Simtk.org, the scientific software repository at Stanford, but we will likely move a copy to the FAH download page in time. If you're having any problems with the GPU MemtestG80 software, you can leave a bug report on the Simtk.org page.


If you find a bug in the software, please leave a bug report in the SimTK project. Thanks!
User avatar
ihaque
Pande Group Member
 
Posts: 239
Joined: Mon Dec 03, 2007 4:20 am
Location: Stanford

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby Gormar » Tue Apr 28, 2009 6:20 pm

Thx, good to know. But can someone put this program on freely available download server ?
Gormar
 
Posts: 122
Joined: Fri Apr 18, 2008 6:33 am

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby bruce » Tue Apr 28, 2009 6:47 pm

Symtk wants to count who downloaded the program by forcing you to register.

If you plan to download software, join an existing project or create a new one of your own please create a Simtk.org account. Our government sponsor NIH requires us to track usage of the software posted on Simtk.org, as one measure of its utility. We do not track usage for any other purpose. Please click the . . . link -- Our Pledge and Your Responsibility -- for more details.
bruce
 
Posts: 21276
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby werty316 » Wed Apr 29, 2009 2:52 am

Gormar wrote:Thx, good to know. But can someone put this program on freely available download server ?

Download the latest version of "OCCT" and give the "GPU: MEMTEST" tool a try. One of my video cards was producing an endless amount of EUE WUs; I tested this card using OCCT's GPU: MEMTEST tool and it found a lot of errors which were the cause of the EUE WUs.
Last edited by werty316 on Wed Apr 29, 2009 2:57 am, edited 1 time in total.
werty316
 
Posts: 150
Joined: Tue Feb 19, 2008 6:29 pm

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby AZBrandon » Wed Apr 29, 2009 2:55 am

werty316 wrote:Download the latest version of "OCCT" and give the "GPU: MEMTEST" test tool a try. One my cards produced an endless amount of EUE WUs and after running OCCT's GPU: MEMTEST tool, it found a heck of a lot of errors on the card which was the cause of the EUE WUs.


What was the solution? Did you replace the card or reduce the speed?
AZBrandon
 
Posts: 225
Joined: Sat Jan 17, 2009 1:43 am

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby werty316 » Wed Apr 29, 2009 2:58 am

Reducing the core, shader, and memory speed did not help. I had to RMA/replace the card.
werty316
 
Posts: 150
Joined: Tue Feb 19, 2008 6:29 pm

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby jaak ennuste » Wed Apr 29, 2009 5:11 am

werty316 wrote:
Gormar wrote:Thx, good to know. But can someone put this program on freely available download server ?

Download the latest version of "OCCT" and give the "GPU: MEMTEST" tool a try. One of my video cards was producing an endless amount of EUE WUs; I tested this card using OCCT's GPU: MEMTEST tool and it found a lot of errors which were the cause of the EUE WUs.


I was not able to test with OCCT, having 4 cards in one rig. Most probably You have to isolate faulty card to another rig ant dest it there. If You test double card, You have to internally SLI it and run as 1 GPU entity. Am I right???

Jaak
building 32 GPU folding rig: 16 x NVIDIA GeForce GTX 295 cards; dual PSU solution; 4 nodes.
Website: Estonia Donates, ambitious 400 PPD supercomputer project
Sponsored by AdlerFlow ceramic coated Porsche exhausts
User avatar
jaak ennuste
 
Posts: 332
Joined: Thu Jan 08, 2009 11:30 am
Location: Tallinn, Estonia, EU

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby X1900AIW » Wed Apr 29, 2009 8:21 am

On Topic:
Downloaded memtestg80 for windows, copied cudaart.dll into the directory, it works (default run: 128 MG & 50 iterations) and uploaded the results.

But afterwards the cmd-windows was closed and I can´t found a txt-file or something to see the results. There is no other command line parameter to produce an output in any kind. Where can I see or find my results ?

One question: when prompted to GPU and RAM MHz I put in "666" and "1150", because some software tools demand DDR-specifications: 1150 or (2x) 2300 MHz ?
X1900AIW
 
Posts: 164
Joined: Thu Apr 17, 2008 8:50 am
Location: Germany

cudart.dll missing, when running memtestg80.exe

Postby poyaochuang » Wed Apr 29, 2009 11:44 am

cudart.dll missing, when running memtestg80.exe
poyaochuang
 
Posts: 37
Joined: Sun Apr 12, 2009 3:26 am

solved

Postby poyaochuang » Wed Apr 29, 2009 11:58 am

found it on appdata roaming folder
copy and paste onto memtestg80 folder
thank you
poyaochuang
 
Posts: 37
Joined: Sun Apr 12, 2009 3:26 am

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby toTOW » Wed Apr 29, 2009 2:10 pm

X1900AIW wrote:But afterwards the cmd-windows was closed and I can´t found a txt-file or something to see the results. There is no other command line parameter to produce an output in any kind. Where can I see or find my results ?


Do not double click on the executable ... run it from the command line prompt.
Folding@Home beta tester since 2002. Folding Forum moderator since July 2008.

FAH-Addict : latest news, tests and reviews about Folding@Home project.

Image
User avatar
toTOW
Site Moderator
 
Posts: 8914
Joined: Sun Dec 02, 2007 10:38 am
Location: Bordeaux, France

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby ChelseaOilman » Wed Apr 29, 2009 2:14 pm

X1900AIW wrote:But afterwards the cmd-windows was closed and I can´t found a txt-file or something to see the results.

You have to open a command prompt window and change directory to where you put the executable, then run it. That way the command prompt window stays open when finished. If your using Vista, you may need to open the command prompt by right clicking and choosing to run as administrator.

X1900AIW wrote:One question: when prompted to GPU and RAM MHz I put in "666" and "1150", because some software tools demand DDR-specifications: 1150 or (2x) 2300 MHz ?

You need to enter the graphics card memory speed and GPU core speed. If you don't know what they are you can get and run GPU-Z to find out.

Edited to correct wrong statement.
User avatar
ChelseaOilman
 
Posts: 1742
Joined: Sun Dec 02, 2007 3:47 pm
Location: Colorado @ 10,000 feet

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby Gormar » Wed Apr 29, 2009 3:25 pm

ChelseaOilman wrote:You need to enter the graphics card memory speed and shader speed, not the GPU core speed.

I'm sorry but you are wrong. It is GPU core speed.
From README file:
"If transmitting data back to Stanford, the memory and core (non-shader) clock
speeds of the card are very useful data."
Gormar
 
Posts: 122
Joined: Fri Apr 18, 2008 6:33 am

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby ChelseaOilman » Wed Apr 29, 2009 5:55 pm

Gormar wrote:I'm sorry but you are wrong. It is GPU core speed.

My bad. :oops:
User avatar
ChelseaOilman
 
Posts: 1742
Joined: Sun Dec 02, 2007 3:47 pm
Location: Colorado @ 10,000 feet

Re: Announcing release: standalone memory tester for NVIDIA GPUs

Postby Distraught » Wed Apr 29, 2009 6:35 pm

Are there any plans to make a bootable version?
Distraught
 
Posts: 2
Joined: Wed Apr 29, 2009 5:07 pm

Next

Return to NVIDIA specific issues

Who is online

Users browsing this forum: No registered users and 2 guests

cron