Has anyone used an eGPU with Qubes and done GPU passthrough? My thought would be a lower-power eGPU passed to a windows Qube for 3d modeling to avoid dual-booting. I have an Asus laptop that has an RTX4090, but I had very poor results attempting to get that working while maintaining my ability to use external displays for the non-GPU qube since the GPU effectively takes over the display port.
@RuleOfSines
Were you able to make this work?
Not the original poster, but I did manage to get an egpu working. I had to use oculink not thunderbolt though. Oculink worked no problem, first try. Thunderbolt, the pci device kept reporting an error message for it falling off the bus.
In the gaming hvm thread, it was suggested that recompiling the kernel with pci hotplug support enabled might fix it, but I did not want to go through that hassle, so I went the oculink route.
1 Like
Thank you. My laptop does not have Oculink port (this is first time I hear of such thing, lol). I’m trying to connect additional monitors and since I’m unable to use my onboard NVIDIA GPU, I was thinking maybe I could get separate standalone device, which seems to be eGPU in this case and attach monitors that way. Inability to use multi monitor setup is big deal for my productivity.
Oculink is not a common port. You would need to add it with an m.2 adapter if your laptop has a spare m.2 port.
My laptop has 2x m.2 ports, (both in use). I’ve got regular Debian on second drive in cases where I can’t accomplish something in Qubes. I probably could sacrifice the second drive if I could get multi monitor setup working. But how the port is going out of the laptop?
Its a very much diy job. Some people just take the bottom off and swap in the adapter when needed, if you have the room you could also make a hole in the frame of tour laptop and route the port to it.
1 Like