Compute processes: not supported

Hi everybody,

I’m new here. I just installed the CUDA platform plus the latest drivers on an Ubuntu 12.04.4 64 bits. The GPU is a Quadro FX 5800. I restarted the system and when I do nvidia-smi, this is what I get:

+------------------------------------------------------+                       
| NVIDIA-SMI 5.319.37   Driver Version: 319.37         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro FX 5800      Off  | 0000:41:00.0     N/A |                  N/A |
| 35%   57C  N/A     N/A /  N/A |        3MB /  4095MB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Compute processes:                                               GPU Memory |
|  GPU       PID  Process name                                     Usage      |
|=============================================================================|
|    0            Not Supported                                               |
+-----------------------------------------------------------------------------+

I’m specially worried about the compute processes, since this is the main reason of having this GPU. Is there any way to activate this function in this GPU/system?

Thanks!

What is ‘not supported’ is the ability to see the CUDA process name(s) active on the GPU via nvidia-smi, because NVIDIA believes that to be a ‘professional’ feature and restricts it to higher end cards that are fully supported by nvidia-smi. Rest assured any CUDA code you try should run fine… it just will not show up on that list or show up as ‘Not Supported’ (without a name) and memory usage/PID

Thanks vacaloca.

Is there a way to monitor the usage of the GPU then? I’m using Munin for that machine and I can controll the CPU usage but I don’t know how to measure the GPU usage. Any application (I prefer a cli command anyway) may be useful. Thanks again.

Hi,

You can download a utility here http://cuda-z.sourceforge.net/ for free, called cuda-z. The utility tells you all about your computers graphics card capabilities, including what version of compute capability it runs. It also tells you detailed information about your graphics memory, as well as showing you detailed performance data measurements and an option to do a heavy load test. There are also two other utilities available on the site, one that tells you about your Opencl CUDA GPU and memory setting, as well as any running extensions and there is also a wrapper library: all are available for both linux and windows OS. Be sure that you have your GPU selected as a CUDA GPU in your Nvidia control panel 3d management settings.

There is a way to get nvidia-smi to report usage, it’s a bit of a hackjob, but it worked last time I tried it – you’re looking for this: GitHub - CFSworks/nvml_fix: A workaround for an annoying bug in nVidia's NVML library. Allows nvidia-smi to work once more!

In Windows, there are apps like GPU-z that will report usage via a GUI and let you export to a text file. I’m not aware of a utility that does that for Linux, basically because the libraries for reporting usage on non-professional GPUs don’t exist in Linux that I am aware of.