GPU health

I was just looking for some software or some API from which I should be able to check the health of the Graphic Card.
I just wanna confirm that the my Graphic Card is OK like:
Drivers are correctly installed,
How th memory pattern change when i use some graphic application like game,video.

Anybody please help

Thanks in advance…

GPU-Z does pretty good.You calso use the save to logfile option while playing game’s or other application’s and read it later.

GPU-Z is indeed quick and easy, though the best way to really check things is to run the SDK examples.

Every new machine I configure or reconfigure, I tend to run the following quick tests (they only take a minute).

  1. SDK deviceQuery. Especially critical for multi-gpu to make sure all GPUs are enumerated for CUDA. (No SLI issues, etc)
  2. SDK bandwidthtest: Mostly to make sure I haven’t used an 8X PCIe slot by mistake
  3. SDK nbody - because it’s pretty and runs the GPU full out. Run one copy per GPU simultaneously. It’s a heat/power test too…
    4) Mem check: I’ve never found an issue on 20+ GPUs, but it’s easy to run anyway.

All four of those tests run on Windows, Linux, and OSX, so you can check any system.

There are a lot of people (especially serious folding guys) who skip all the above and just use “does folding/BOINC work” or not but I like the multiple verification.

Hi,

Where could i find these SDK’s i downloaded SDK code samples from nVidia website but i didn’t find the SDK’s you are talking about

Thanks

you must install driver,tookit and sdk before you run the example in sdk.

http://developer.nvidia.com/object/cuda_2_3_downloads.html

hi,pkhetan.
can you tell how to know what kind of PCIe you use ?
SDK bandwidthtest: Mostly to make sure I haven’t used an 8X PCIe slot by mistake

So ihave installed : Developer Drivers + CUDA Toolkit + CUDA SDK code samples

That’s right?

now how to run the 3 examples SPWorley mentioned? where are they?

Hi,

While trying to install (#make) the SDK samples, i got these messages :

make: g++: Command not found
make: *** [obj/release/bank_checker.cpp.o] Error 127

And for memtest i got these :

nvcc -c -arch sm_13 -DSM_13 -O3 -I. -I/usr/local/cuda/include -o cuda_memtest.o cuda_memtest.cu
make: nvcc: Command not found
make: *** [cuda_memtest.o] Error 127

Can any one help me to resolve these problems ?

Regards