GPU Acceleration

[GPU]

Ive read into documentation on ways of setting up this up
Create a gaming hvm
sys-gpu

None of these have worked.
Qubes os would really boost my work flow if I be able to fully utilize my GPU

Is this at all possible?

[Syonyk]

What, specifically, are you looking to do? You can get a hardware GPU accelerated VM with separate outputs with the first technique, and the second, as I understand it, allows you to separate the GPU out from dom0 (currently, you would have some dom0 GPU acceleration, but that’s just useful for dragging windows around, as it’s not exposed to the AppVMs at all).

“General purpose GPU acceleration of AppVMs in Qubes” isn’t a thing I’m aware of.

[GPU]

Im looking to make a VM under windows 10 and linux that can fully utilize my gpu and I want to know how to configure that because all these options have not functioned for me.

Something that could be used for high quality video play back, playing video games and generally more intensive graphical applications.

[Syonyk]

The “Gaming HVM” option is the path you want to go down, then. You’ll need two GPUs (at least), and the secondary AppVM GPUs will need their own link to your display - you can’t interface them with the other windows in the “main Qubes UI” of AppVMs.

But, my normal advice is just to build two systems (or, if you really want, and are aware of the risks, dual boot). Keep all your critical data on Qubes, and play games/videos on the other system where you don’t do anything like access email, servers, etc.

[GPU]

Ive tried it and it wasnt successful for me but mabye it was I need to install drivers in the gpu accelerated VM first before anything else or no?

[apparatus]

@Syonyk

and the secondary AppVM GPUs will need their own link to your display - you can’t interface them with the other windows in the “main Qubes UI” of AppVMs.

It’s possible to use it seamlessly as well:

Seamless GPU passthrough on Qubes OS with VirtualGL

We are presenting to you a guide regarding GPU passthrough created in collaboration with peers. Qubes OS already has extensive documentation available about GPU passthrough for 3D accelerated tasks but they all require in depth configuration, extra displays, and extra input devices. With VirtualGL (https://virtualgl.org) one can take advantage of the existing Qubes OS framework for audio/display by offloading OpenGL onto the secondary graphics card directly with VirtualGL. This document will mo…

[Syonyk]

GPU, I’m sorry, I’ve not actually done it. I just use separate hardware.

@apparatus

It’s possible to use it seamlessly as well:

Interesting… thanks, I’ve not seen that before. I may have to mess with that at some point!

[GPU]

I attempt this and report back

[GPU]

Ive tried virtual gl with absolutely no success

[ddevz]

@Syonyk

“General purpose GPU acceleration of AppVMs in Qubes” isn’t a thing I’m aware of.

But it’s being worked on! See “Reasonably Secure GPU Acceleration” in:

Xen Project Summit 2024 VODs are available on YouTube

Thanks for sharing this, @renehoj! Here are the specific talks by Qubes core devs for anyone who might want direct links to them: Demi Marie Obenour: Reasonably Secure GPU Acceleration Marek Marczykowski-Górecki: Host and Guest Suspend Under Xen - S3 and S0ix Marek Marczykowski-Górecki: Linux Stubdomains Status Update