Intel 13gen or 14gen iGPU full passthrough to Win?
- Thread starter icehq
- Start date
-
- Tags
- igpu passthrough
If anyone else follows this thread, could they please post the exact make and model of whatever (minipc?) device that they have successfully applied iGPU passthrough to?
I've got a Minisforum HM90 (AMD) and have had no luck, trying every few months after a PVE update. I figure if "we" compare notes on various hardware combinations, then "we" might be able to cobble together one or more reliable iGPU passthrough guides.
I've got a Minisforum HM90 (AMD) and have had no luck, trying every few months after a PVE update. I figure if "we" compare notes on various hardware combinations, then "we" might be able to cobble together one or more reliable iGPU passthrough guides.
@Xclsd Thanks for the heads-up on that beast of a mobo working with Ubuntu. I actually want to use a linux desktop myself, so that is encouraging. I have been mainly testing with a Win10 image because I thought that would be the most likely to work. I will go through the procedure again with an early Ubuntu 24.04 mantic desktop on my HM90 Minisforum and see how I go, and report back.
@markc NP 
I read on another thread that the issue with the driver is also affecting other users using the UHD 770 driver: https://forum.proxmox.com/threads/i...passthrough-the-dreaded-code-43-error.113229/
I tried it on Win10 too. If I get a chance, I will try Win11, just in case it is a version specific problem.
I read on another thread that the issue with the driver is also affecting other users using the UHD 770 driver: https://forum.proxmox.com/threads/i...passthrough-the-dreaded-code-43-error.113229/I tried it on Win10 too. If I get a chance, I will try Win11, just in case it is a version specific problem.
Interesting, I've never managed to get any kind of GPU passthrough to work to even know what SRIOV has to offer. I googled it and found this page which lists a few NUC models that could be potential iGPU passthrough hardware targets... I obviously have a lot to learn.
https://projectceladon.github.io/celadon-documentation/getting-started/on-vm.html
FWIW my use case is to run my main daily driver desktop (Manjaro/KDE) in a VM on a mid to high end miniPC (hence the need for iGPU passthrough) so I can take advantage of effortless back-ups to my Proxmox Backup Server, and a trouble free bootable ZFS file system. I could handle a 5% performance hit from a VM, but SPICE gfx, for instance, is just not good enough for all day desktop usage.
https://projectceladon.github.io/celadon-documentation/getting-started/on-vm.html
FWIW my use case is to run my main daily driver desktop (Manjaro/KDE) in a VM on a mid to high end miniPC (hence the need for iGPU passthrough) so I can take advantage of effortless back-ups to my Proxmox Backup Server, and a trouble free bootable ZFS file system. I could handle a 5% performance hit from a VM, but SPICE gfx, for instance, is just not good enough for all day desktop usage.
Well, it depends on what you want to achieve: For desktop usage a VM with SPICE is enough for my need. But if you need accelerated GPUs for some reason, you might end up needing GVT-G, SRIOV or GPU passthrough. GVT-G worked somehow for me in testing scenarios.
Hello,
I'm thinking about the nuc13 (raptor lake 13th gen with iris xe). There ale a lot of posts about the poblems with the full igpu passthrough to windows VM. Two questions:
- is linux VM (e.g. ubuntu) simply working for such setup, so I did not find similar posts?
- does the full igpu passthrough give the full hdmi audio/video output beside the gpu access, and so the VM can be used as a desktop on external monitor? Maybe it's so obvious, but the posts I found cover windows VM only in relation to similar doubts.
Last edited:
It should be working. You don't need too passthrough anything to use a VM. You would just have to use a virtua GPU then.
IF the passthrough is possible (I don't know if it is): Linux will just work, if mesa / kernel is as new as needed for the GPU.
IF SR/IOV is working(I don't know if it is):: Linux will just work as well.
IF the passthrough is possible (I don't know if it is): You can use HDMI + Audio output as a usual Desktop. You'll just need to power on the device somehow via the Interface / a Script / whatever solution is possible to power on the VM.
I can confirm, AV over HDMI works well with the full gpu passtrhough for NUC13 i7 with linux VM.
Hey,
Hooow? I've been scratching my head for a few days trying to get HDMI to my Debian VM. SR-IOV or full GPU passthrough, I haven't gotten it working.
Got GPU passthrough for use with plex and whatnot, yeah. But HDMI for my VM seem impossible...
You have a guide or something?
Sure, a few days to dig my config out.
These are my configs, the other commands are added manually (all this comes from that other thread).
You must run with the hdmi already connected. That simply works, both audio and video from VM.
I've just noticed I have duplicates in cmdline and in blacklist.
These are my configs, the other commands are added manually (all this comes from that other thread).
You must run with the hdmi already connected. That simply works, both audio and video from VM.
I've just noticed I have duplicates in cmdline and in blacklist.
Code:
root@pve:~# grep _CMDLINE_LINUX_ /etc/default/grub
#GRUB_CMDLINE_LINUX_DEFAULT="quiet"
GRUB_CMDLINE_LINUX_DEFAULT="intel_iommu=on iommu=pt module_blacklist=i915,snd_hda_intel,snd_hda_codec_hdmi,snd_sof_pci_intel_tg1"
root@pve:~# update-grub
root@pve:~# grep vfio /etc/modules
vfio
vfio_iommu_type1
vfio_pci
root@pve:~# cat /etc/modprobe.d/blacklist.conf
blacklist i915
blacklist snd_hda_intel
blacklist snd_hda_codec_hdmi
blacklist snd_sof_pci_intel_tg1
root@pve:~# cat /etc/modprobe.d/vfio.conf
options vfio-pci ids=8086:a7a0,8086:51ca
options vfio_iommu_type1 allow_unsafe_interrupts=1
root@pve:~# update-initramfs -u -k all
root@pve:~# systemctl reboot
root@pve:~# mkdir /usr/share/kvm/nuc13
root@pve:~# wget -P /usr/share/kvm/nuc13/ https://github.com/gangqizai/igd/raw/refs/heads/main/gen12_gop.rom
root@pve:~# wget -P /usr/share/kvm/nuc13/ https://github.com/gangqizai/igd/raw/refs/heads/main/gen12_igd.rom
root@pve:~# md5sum /usr/share/kvm/nuc13/*
d4a9ba02ddd8134be5abafa6dfd699cf /usr/share/kvm/nuc13/gen12_gop.rom
42abaea033982f2b2d1b616ae6062936 /usr/share/kvm/nuc13/gen12_igd.rom
root@pve:~# sed 's/[-0-9a-fA-F:]\{12,\}/xxx/' /etc/pve/local/qemu-server/100.conf
agent: 1
args: -set device.hostpci0.addr=02.0 -set device.hostpci0.x-igd-gms=0x2 -set device.hostpci0.x-igd-opregion=on -debugcon file:/root/igd_debug.log -global isa-debugcon.iobase=0x402
bios: ovmf
boot: order=scsi0
cores: 4
cpu: x86-64-v2-AES
efidisk0: local-btrfs:100/vm-100-disk-0.raw,efitype=4m,pre-enrolled-keys=1,size=528K
hostpci0: 0000:00:02.0,romfile=nuc13/gen12_igd.rom
hostpci1: 0000:00:1f.3,romfile=nuc13/gen12_gop.rom
ide2: local-btrfs:iso/Fedora-Workstation-Live-x86_64-39-1.5.iso,media=cdrom,size=2079836K
machine: q35
memory: 8192
meta: creation-qemu=8.1.2,ctime=1705861785
name: fedora
net0: virtio=xxx,bridge=vmbr0,firewall=1
numa: 1
ostype: l26
scsi0: local-btrfs:100/vm-100-disk-1.raw,discard=on,iothread=1,size=20G,ssd=1
scsihw: virtio-scsi-single
smbios1: uuid=xxx
sockets: 2
usb0: host=1-7
usb1: host=1-1
vga: none
vmgenid: xxx
Last edited:
M8 you saved me really. With exactly your settings with minor differences i managed to passthrough from my Intel Ultra 245K the igpu to a WIndows 11 VM that i use as workstation.
I just need now to find a way to passthrough the NPU of the 245k where anything i tried was unsuccessful
Thank you again.