Hi.
I have an MSI x670e Tomahawk + AMD 7950x and normally use a VM (virtlib) on Debian 13 with nVidia 4080S in PCIe1 as host CPU and an Intel ARC 310 (passthrough) for the guest (Win11).
This is a good way to work well both with Debian and Win11 as the 4080S is the boot GPU in bios
The problem is when I need to have the 4080 into the guest and the Arc310 for the host ass the boot process is a mess: uefi opt ftom 4080, grub from 4080 and login and rendering in Debian with the Arc310.
Not a good situation as I’m forced to switch my display between the 2 cards if I want to control the whole boot process.
I’m then working with the 7960x’s iGPU as main GPU and 4080 or Arc310 as secondary GPU if - viceversa, of course - the Arc310 or the 4080 is passed to the VM guest.
Neither this is an optimal setup as frequently when the 4080 is assigned to the VM guest and the Arc310 as secondary to Debian and I resume from stop i get the display output moved to the Arc310 just for login and then back. I then have to switch twice the monitor input.
The question is: what if I get an Asus x670e Proart eith 2x GPU slot? Is it just a technical matter or if I pass the 4080 in slot 1 Debian will automatically switch to the Arc310 in slot 2 with no extra issues? (iGPU will be disabled).
Tnx in advance, hope I’m clear, I’m not English
1 Like
I think the architecture is not supportive to everything you want to do.
Perhaps -
iGPU (or a third, display output only card) for your Host OS;
- pass through whichever big card as required to whatever application VM you’re using.
Then you have full choice of what card you use when.
That might mean it’s also best to move some of the applications you have had out on the host, now down into VM’s of their own. That’d be a bit like a DIY Qubes OS. I can’t tell if that is worth it for you overall.
1 Like
You’re right. Is like asking the mobo to guess which card I would like to use as primary at each startup.