iGPU passthrough help please

Hello,

I’m struggling to passthrough my iGPU (intel 580) to a VM.

I’m following the OVMF guide on the Arch wiki along with numerous guides on the net, doing my best to try to understand what is going on, but I’m hitting a wall.

I’ve got VT-d on in the bios. I’ve enabled IOMMU.

I can’t use gpu-passthrough manager as I’m on systemd-boot (and this renders almost all guides useless as I don’t know where to plug in the IOMMU and PT commands)

I’m unsure where to set IOMMU on in Linux under system d… I think…
It’s in /etc/kernel/cmdline?

I think my IOMMU groups are valid…

I haven’t got a clue how to successfully isolate my iGPU from host. No matter what I do, I can still see it in system settings and use it with the host on another monitor. My understanding is the host should NOT be able to see it, or use it…

The arch guide seems a bit unclear on the modprobe section.

It says I can use modprobe or adding to INTRAMFS… trying one, the other, or both doesn’t seem to successfully isolate my iGPU.

Most guides seem to do both, but I can’t do anything with intramsfs due to endeavour using dracut?

I’m trying to plug the relevent settings into dractut, then doing a rebuild but again, nothing seems to isolate the iGPU from the host.

Seems like all I’ve achieved is lengthening the boot time, and again, I don’t understand if I should be using both options or just one of the two.

This is as far as I can get at the moment. I nuked my system and had to reinstall after trying hikari knights tool… I’m assuming that’s setup for grub and doesn’t work with system d as after trying it I didn’t have bootloader anymore :frowning:

Before I nuked the system, I was able to passthrough the iGPU into QEMU build but I couldn’t actually load a driver for it or use it for anything. Windows 10 didn’t know what to do with it.

Any help would be appreciated.

My goal is to be able to run Windows in a VM with best possible performance and be able to delete my dual boot and just use Linux going forward. This passthrough problem is the only thing preventing my from doing that at the moment.

Thanks. Any help / guidance is much appreciated.

I just use Windows on another drive. Even though it’s dual boot it doesn’t affect linux in any way. Dual booting doesn’t bother me and it’s way less effort than trying to use qemu. But it’s your hardware and your time. :man_shrugging:

Need regular access to helix edit, which is windows/Mac only (USB controller for Helix Multi effects processor) and the VM feels frustrating slow without hardware acceleration

1 Like

Have you tried vmware pro?

No idea about intel iGPUs, but people have been successfull with AMD iGPUs. I was looking into this when I first got my R7 5700G.

In my experience, make sure these things are correct:

  • BIOS settings related to IOMMU and graphics
  • IOMMU groups enable wanted function of passing iGPU, might still be possible with patch
  • guest settings are correct
  • GPU driver is properly unloaded on the host and there is no conflicts

Try virtual machine manager. A Qemmu/kvm GUI. It’s faster than the other virtualisation technologies. It’s actually usable.

1 Like