Please clarify the following:
Does sys gui gpu can work with intel’s integrated gpu?
Does the GUI domain | Qubes OS updated to 4.2.3 and/or do I need to do anything beyond the salt formula listed there?
If I have another GPU(nvidia) that I can’t passthrough, can I create additional some kind of sys gui gpu that will handle the rest of the GPU intensive tasks?
1 Like
sys-gui-gpu purpose is just to remove the GPU from dom0 to reduce attack surface.
If you have an intel and nvidia GPU, if you use intel for dom0 this leaves the nvidia gpu to be used by a HVM qube. Switching to sys-gui-gpu will provide the exact same thing.
2 Likes
Got it.
The thing is I can’t passthrough my GPU physically and it’s not related directly to Qubes.
I was thinking of maybe a way to make a compromise with some kind of alternative dom0/sys gui gpu that will work as compromised solution for tasks I can take risks.
Is there anything alike this or an alternative version of it, but not installing an additional complete standalone distro alongside Qubes?
no
the Qubes OS goal is to offer a seamless GPU virtual driver to Qubes, but it’s not ready yet, there is not even a thing to test, it may take 1 or 2 years to bake. The only solution to use a GPU in a qube is to have 2 physical GPUs and go through the complex GPU passthrough setup
1 Like
Can you provide a link for the discussion regarding the seamless GPU virtual driver please
there is nothing concrete except some GitHub issues in the project
Tangentially related: I am looking forward to this issue getting a solution PCI (GPU) passthrough hardening: option ROM edition · Issue #1087 · Dasharo/dasharo-issues · GitHub
There is a third party option. Not sure if it can work with sys-gui though.