PCI at NVIDA Tesla P 100 in shared pass through mode is disabled

I have successfully completed NVIDIA Tesla P100 Grid Setup on the vSphere Host Server with VMware ESXI 6.7.While trying to add PCI devices in the VM, the option to choose PCI devices is shown as in the “Add Other Hardware” setting in “Virtual Hardware". Please refer the screenshot below.

enter image description here

I can add PCI devices when NVIDIA is used as Direct I/O passthrough, but when I use NVIDIA shared passthrough, PCI device is shown as disabled. As, I need to share the GPU, I’m forced to use NVIDIA shared passthrough.

nvidia-smi is as follows

 NVIDIA-SMI 410.91       Driver Version: 410.91       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 P100-PCIE...  On   | 00000000:3B:00.0 Off |                  Off |
| N/A   31C    P0    26W / 250W |     31MiB / 12287MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

The result of #esxcli graphics host get is as follows.

 Default Graphics Type: SharedPassthru
   Shared Passthru Assignment Policy: Performance

Kindly help me to enable the PCI at NVIDA Tesla P 100 in shared pass through mode.

Answer

i know this question is old, but i just ran into the same “issue”. you cannot use vGPU without vCenter. you must install a vCenter to get the add shared pci ressource option for a VM.

Attribution
Source : Link , Question Author : Sarath Zacharia , Answer Author : kolbicz

Leave a Comment