I have an Intel 13th 1335u (Irix xe GPU) as bare metal on which i have installed an configured Proxmox. I enabled GPU accelration using vGPU implementation (i can share among VMs up to 7 vGPU). On a windows 11 VM, i succeded to PCIe passthrought one of those GPU and everything works fine there. This information should be helpful, as it is a proof the at least proxmox and win11 are able to use vGPU.
Then i have installed another VM with TrueNAS SCALE Fangtooth 25.04 and ad APP Immich v1.9.6 .
Immich APP is configured at the moment to use TrueNAS configuration GUI. GPU passthorugh flag is enabled as well as OpenVINO
TrueNAS scale, under System Advanced Isoleted GPU devices is able to see mi Intel Iris Xe GPU (of course i have not enabled isolation as this is for VM under TrueNAS not APP (containers)).
truenas_admin@truenas[~]$ sudo dmesg | grep 0000:02:00.0
[ 0.476681] pci 0000:02:00.0: [8086:a7a1] type 00 class 0x030000 PCIe Endpoint
[ 0.483486] pci 0000:02:00.0: BAR 0 [mem 0x80000000-0x80ffffff 64bit]
[ 0.491478] pci 0000:02:00.0: BAR 2 [mem 0x7000000000-0x701fffffff 64bit pref]
[ 0.516129] pci 0000:02:00.0: DMAR: Skip IOMMU disabling for graphics
[ 0.705403] pci 0000:02:00.0: vgaarb: setting as boot VGA device
[ 0.705403] pci 0000:02:00.0: vgaarb: bridge control possible
[ 0.705403] pci 0000:02:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none
[ 12.996170] i915 0000:02:00.0: [drm] Found ALDERLAKE_P/RPL-U (device ID a7a1) display version 13.00 stepping E0
[ 14.998184] i915 0000:02:00.0: [drm] ERROR Device is non-operational; MMIO access returns 0xFFFFFFFF!
[ 14.998963] i915 0000:02:00.0: [drm] ERROR Device initialization failed (-5)
[ 14.998970] i915 0000:02:00.0: probe with driver i915 failed with error -5
[ 15.241385] xe 0000:02:00.0: Your graphics device a7a1 is not officially supported by xe driver in this kernel version. To force Xe probe, use xe.force_probe=‘a7a1’ and i915.force_probe=‘!a7a1’
truenas_admin@truenas[~]$ midclt call app.gpu_choices | jq
{
“0000:02:00.0”: {
“vendor”: “INTEL”,
“description”: “Intel Corporation Raptor Lake-P [Iris Xe Graphics]”,
“error”: null,
“vendor_specific_config”: {},
“gpu_details”: {
“addr”: {
“pci_slot”: “0000:02:00.0”,
“domain”: “0000”,
“bus”: “02”,
“slot”: “00”
},
“description”: “Intel Corporation Raptor Lake-P [Iris Xe Graphics]”,
“devices”: [
{
“pci_id”: “8086:A7A1”,
“pci_slot”: “0000:02:00.0”,
“vm_pci_slot”: “pci_0000_02_00_0”
}
],
“vendor”: “INTEL”,
“uses_system_critical_devices”: false,
“critical_reason”: null,
“available_to_host”: true
},
“pci_slot”: “0000:02:00.0”
}
}
Now two things happens:
-
if I configure proxmox to pass a Intel Iris xe vGPU + standard VGA, Immich APP will start but under Immich container dev/dri does not include any /dev/dri/renderDxxx. Running a Machine Learning process will trigger a Machine Learning container LOG “No OpenVINO GPU Found” (I think because it is trying to use the standard VGA, that does not have kind of acceleration)
-
if I configure proxmox to pass just a Intel Iris xe vGPU, Immich APP will crash at start , and inside TrueNAS log I found
“error gathering device information while adding custom device “/dev/dri”: no such file or directory\n”
Under TrueNAS it seems that kenel module is loaded but driver is missing
lspci -v
02:00.0 VGA compatible controller: Intel Corporation Raptor Lake-P [Iris Xe Graphics] (rev 04) (prog-if 00 [VGA controller])
Subsystem: Intel Corporation Raptor Lake-P [Iris Xe Graphics]
Physical Slot: 0-2
Flags: fast devsel
Memory at 80000000 (64-bit, non-prefetchable) [size=16M]
Memory at 7000000000 (64-bit, prefetchable) [size=512M]
Capabilities:
Kernel modules: i915, xe
Kernel driver in use:
is there a way to fix this ?