M10 vGPU 8Q profile on Linux -- unable to create CUDA profile; code 999

Virtual GPU Forums More vGPU Forums Tesla Boards
1

Hi all,

Nvidia-smi on guest:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.63       Driver Version: 430.63       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GRID M10-8Q         On   | 00000000:00:0A.0 Off |                  N/A |
| N/A   N/A    P8    N/A /  N/A |    583MiB /  8192MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0       711      G   /usr/lib/xorg/Xorg                            55MiB |
+-----------------------------------------------------------------------------+

Nvidia-smi on host says:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.67       Driver Version: 430.67       CUDA Version: N/A      |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla M10           On   | 00000000:03:00.0 Off |                  N/A |
| N/A   34C    P8    10W /  53W |     13MiB /  8191MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  Tesla M10           On   | 00000000:04:00.0 Off |                  N/A |
| N/A   38C    P8    10W /  53W |     11MiB /  8191MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   2  Tesla M10           On   | 00000000:05:00.0 Off |                  N/A |
| N/A   44C    P8    10W /  53W |     11MiB /  8191MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   3  Tesla M10           On   | 00000000:06:00.0 Off |                  N/A |
| N/A   43C    P8    10W /  53W |   8141MiB /  8191MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    3     26990    C+G   vgpu                                        8128MiB |
+-----------------------------------------------------------------------------+

When I try to run anything of tests from CUDA toolkit, I got an error:

~$ ./EGLStream_CUDA_Interop
Found 1 cuda devices
Found EGL-CUDA Capable device with CUDA Device id = 0
Created EGLStream 0x55cd70ab2911
EGLStream initialized
CUDA Producer on GPU Device 0: "GRID M10-8Q" with compute capability 5.0

failed to create CUDA context (0x3E7)
&&&& EGLStream interop test FAILED

At first test run, there is a log in dmesg:

[43311.987240] nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 240

I tried more recent driver versions but no luck, so is it with P100 12GB too.

Where am I wrong? Can I somehow get more detailed debug?

Thank you in advance!

2

Hi

Has the VM picked up a license ok?

Regards

MG

3

No, it hasn’t. But is active license mandatory for CUDA to work? I could get similar configuration working in beginning of this year – since I use nvenc, I remember that just image was choppy (3fps) but there weren’t problems with CUDA. nvidia-smi dmon showed encoder usage. I just can’t figure out what changed since then :(

4

Hi

License the Software and see if that helps.

Regards

MG

5

MrGRID, you was right. I’m very very confused now about how could I see nvenc working on unlicensed card. Thank you a lot for a straight way :)

6

NVENC has nothing to do with CUDA. CUDA is not enabled without a license

7

Yes, you are right. In Gstreamer nvh264enc plugin NVENC uses CUDA and I thought that it’s the only method to use it, sorry.