nvidia-smi reports 0% utilization

Issues related to applications and software problems
Post Reply
Elliott
Posts: 44
Joined: 2018/09/14 01:30:14

nvidia-smi reports 0% utilization

Post by Elliott » 2018/11/07 19:21:44

Hi everyone. I'm running CentOS 7.2 because of some older software compatibility. I have two GPUs in the machine and nvidia-smi always shows 0% utilization, even when I know the GPUs are working. I saw the problem with Nvidia driver 384, so I updated to 390 and the problem persists. Any idea what is wrong?

P.S. I don't know why X is running on both cards, since I only have one display connected.

Code: Select all

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 390.87                 Driver Version: 390.87                    |
|-------------------------------+----------------------+----------------------+
| 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 P6000        Off  | 00000000:04:00.0  On |                  Off |
| 31%   64C    P0    75W / 250W |    540MiB / 24449MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  TITAN Xp            Off  | 00000000:84:00.0 Off |                  N/A |
| 47%   77C    P2   176W / 250W |   5643MiB / 12196MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      7332      G   /bin/X                                       105MiB |
|    0     21878      G   /opt/Autodesk/flame_2019.1/bin/flame         405MiB |
|    1      7332      G   /bin/X                                        56MiB |
|    1     21878      G   /opt/Autodesk/flame_2019.1/bin/flame           3MiB |
|    1     23303      C   ...k/wiretapgateway/current/wiretapgateway  5565MiB |
+-----------------------------------------------------------------------------+

chemal
Posts: 776
Joined: 2013/12/08 19:44:49

Re: nvidia-smi reports 0% utilization

Post by chemal » 2018/11/07 23:54:49

Your're talking about this?

https://knowledge.autodesk.com/support/ ... sreqs.html

From looking at this DKU tarball, it seems pretty much obvious they are doing their own thing which seems to be either plain CentOS 7.2 or CentOS 7.4 without any updates + the contents of this tarball. You should ask their support, if anything isn't working as expected.

User avatar
TrevorH
Site Admin
Posts: 33202
Joined: 2009/09/24 10:40:56
Location: Brighton, UK

Re: nvidia-smi reports 0% utilization

Post by TrevorH » 2018/11/08 00:29:54

You should also be aware that CentOS 7.2 went out of support with the release of 7.3 in Nov 2016. Assuming that you have never patched it since installing it then you currently have somewhere around 68 "Critical" security vulnerabilities potentially present on your system and many more that are "Important". You should ask Autodesk if they are willing to reimburse you the costs of any security breach that you might suffer by running unpatched.
The future appears to be RHEL or Debian. I think I'm going Debian.
Info for USB installs on http://wiki.centos.org/HowTos/InstallFromUSBkey
CentOS 5 and 6 are deadest, do not use them.
Use the FAQ Luke

Elliott
Posts: 44
Joined: 2018/09/14 01:30:14

Re: nvidia-smi reports 0% utilization

Post by Elliott » 2018/11/13 17:45:37

Trevor you are right, and I advocate for security patches all the time. But I have to test and support what my customers demand, which sometimes includes 7.2. Autodesk has already been qualifying CentOS 7.4 for nearly two years; they usually skip the odd numbers for some reason. Unfortunately many shops would prefer to run old software than to spend the engineering resources to upgrade and test everything, so they don't update until absolutely necessary. Autodesk does not make it easy to upgrade the OS, you have to wipe the hard drive and reinstall.

I find the same problem with nvidia-smi using 7.4. I did file a bug report with Autodesk, but they told me this issue happens in other CentOS systems too, not specific to their image.

Elliott
Posts: 44
Joined: 2018/09/14 01:30:14

Re: nvidia-smi reports 0% utilization

Post by Elliott » 2018/11/16 17:44:13

I found that this problem is not specific to Autodesk. It happens any time you configure PowerMizer using RegistryDwords. If anyone is interested the bug is described here https://devtalk.nvidia.com/default/topi ... ilization/

Post Reply