GPU sharing (extendend) setup help

[Venson]

Hey,

I am currently planning to create a small server with my NUC11 and a 3080. I aim for 2 people to work of it and have some docker containers running of it. But I also want the containers to have GPU accelleration (jellyfin) and need to have low latency from both windows VMs for Parsec use.

 

I tried two solutions:

A: HyperV

Host: Windows 11 Pro

Virtualization: Hyper-V

VMs:

1. WSL2 - for docker container use

2. Windows - Games

3. Windows - Development 

Premise: The GPU-P feature from jamesstringerparsec/Easy-GPU-PV: A Project dedicated to making GPU Partitioning on Windows easier! (github.com) is working great, but although there are claims that GPU-P should also work with Linux guests, I could not find any way to make it work. 

 

B: Proxmox

Host: Linux Ubuntu

Virtualization: Proxmox (KVM)

VMs:

1. Linux Ubuntu

Premise: There are some scripts, also very good tutorial from Craft-Computing regarding this, however the essential DualCoder/vgpu_unlock: Unlock vGPU functionality for consumer grade GPUs. (github.com) script does not support Ampere and reading into it, never will. Same goes for the IGpu in the Nuc, it seems like all versions right up to 11th gen support IGpu partitioning but not 11th gen.

 

I would preferer to go with a Linux host and have 1 Linux guest and 2 Windows guests but if anyone has some ideas how to continue I would greatly appreciate it.