[SOLVED] One GPU multiple VMs acceleration and Plex encoding

[MJFlanigan]

I have dell r720 that I'm running proxmox on. I'd like to buy one big graphics card to be able to run my windows 10 pro vm and ubuntu 20.04 with some machine learning tasks off one gpu.

I was looking at buying something like a rtx a5000 or rtx a6000 to accomplish running both and using a little of it for Plex encoding running on truenas core in a vm.

Or should I just buy two or three gpu to make this happen?

 

[Pedulla]

It's possible, but you have to buy a license from Nvidia to do this with a Quadro.
Nvidia Docs, sect 1.1.3

 

[jasonsansone]

You can easily share an nVidia card to multiple LXC containers, but VM’s requires a Grid license as mentioned above. I have Tesla cards in each node that are shared between Plex containers, machine learning, transcoding, etc etc. It all works perfect and can even still support HA failover between nodes when configured correctly. If you want to run a GUI OS like Windows I would just use multiple GPU’s with passthrough.

 

[Pedulla]

You can easily share an nVidia card to multiple LXC containers,

I like this. I'm gonna give it a shot.
Just so I understand, you install the hVidia drivers on the proxmox host, right?
How did you setup/configure the LXC containers?
Is the nVidia card also the video card for the host or container?

 

[Pedulla]

You can easily share an nVidia card to multiple LXC containers,

Thanks for the direction here.
I'm having great success with this: https://forum.proxmox.com/threads/gpu-general-concept-questions.97947/post-450823

Your application is all headless, do you have any experience with GUI performance in those LXC's? (See question in above post)