I've been using UnRaid for a few years now. I'm in the process of setting up my second server based around the Minisforum MS01. I have the internal Pci-e port consumed by a sas expansion card so I can't just throw the GPU in there. Anyways... I have this (TH3P4G3 85W PD Charging For Thunderbolt) gpu dock from Ali-express which at first glance seems to work fine, but when I pass through to my viseron docker (via extra params: "--device=/dev/dri:/dev/dri") ffmpeg can't use the hardware for video decode. The card doesn't seem to be present in /dev/dri . Is is accessible through a different location?

ls -al /dev/dri/by-path/ <--shows only the pcie address of the integrated gpu
dmesg | grep -i drm <--Shows a device is non-operational
uname -r 6.12.24-Unraid

Any troubleshooting tools, tips, or tricks would be appreciated. :)


Here's some base info:

lspci | grep -i vga

00:02.0 VGA compatible controller: Intel Corporation Alder Lake-P Integrated Graphics Controller (rev 0c)

39:00.0 VGA compatible controller: Intel Corporation DG2 [Arc A310] (rev 05)

root@BreauxMeauxSVR:~# ls /dev/dri

by-path/ card0 renderD128

root@BreauxMeauxSVR:~# ls /dev/dri/by-path/

pci-0000:00:02.0-card@ pci-0000:00:02.0-render@

root@BreauxMeauxSVR:~# dmesg | grep drm

[ 53.989791] ACPI: bus type drm_connector registered

[ 54.339427] i915 0000:00:02.0: [drm] Found ALDERLAKE_P (device ID 4626) display version 13.00 stepping D0

[ 54.340199] i915 0000:00:02.0: [drm] VT-d active for gfx access

[ 54.340392] i915 0000:00:02.0: [drm] Using Transparent Hugepages

[ 54.346068] i915 0000:00:02.0: [drm] Finished loading DMC firmware i915/adlp_dmc.bin (v2.20)

[ 54.366970] i915 0000:00:02.0: [drm] GT0: GuC firmware i915/adlp_guc_70.bin version 70.44.1

[ 54.367122] i915 0000:00:02.0: [drm] GT0: HuC firmware i915/tgl_huc.bin version 7.9.3

[ 54.382658] i915 0000:00:02.0: [drm] GT0: HuC: authenticated for all workloads

[ 54.383744] i915 0000:00:02.0: [drm] GT0: GUC: submission enabled

[ 54.383889] i915 0000:00:02.0: [drm] GT0: GUC: SLPC enabled

[ 54.384550] i915 0000:00:02.0: [drm] GT0: GUC: RC enabled

[ 54.385791] i915 0000:00:02.0: [drm] Protected Xe Path (PXP) protected content support initialized

[ 54.387442] [drm] Initialized i915 1.6.0 for 0000:00:02.0 on minor 0

[ 54.393763] i915 0000:00:02.0: [drm] Cannot find any crtc or sizes

[ 54.394398] i915 0000:39:00.0: [drm] Found DG2/G11 (device ID 56a6) display version 13.00 stepping C0

[ 54.403641] i915 0000:00:02.0: [drm] Cannot find any crtc or sizes

[ 56.396609] i915 0000:39:00.0: [drm] ERROR Device is non-operational; MMIO access returns 0xFFFFFFFF!

[ 56.401229] i915 0000:39:00.0: [drm] ERROR Device initialization failed (-5)

root@BreauxMeauxSVR:~# lsmod | grep -i xe

xe 2732032 0

drm_gpuvm 40960 1 xe

drm_exec 12288 2 drm_gpuvm,xe

gpu_sched 57344 1 xe

drm_ttm_helper 16384 1 xe

drm_suballoc_helper 12288 1 xe

drm_buddy 20480 2 xe,i915

ttm 94208 3 drm_ttm_helper,xe,i915

your trying to pass a gpu via a thoundbolt pcie adapter?

Please post a diag file.

whats the output of lspci and lsusb

it could be a usb thunderbolt deivce thus the /dev/drive for the gpu is not shown...

the gpu is also a bit weird to use on unraid due to kerneal driver:

https://www.reddit.com/r/unRAID/comments/1kwp2ac/intel_arc_a310_eco_support/

https://forums.unraid.net/topic/187952-need-help-with-intel-arc-a310-and-unraid/

do you have intell top installed?

Sorry for the extreme delay, I do have 'intel_gpu_top' plugin installed, but it is showing the integrated gpu.
Here's the Diagnostic .zip generated from the gui. I have selected the anonymize option.unraid_ms01_thunderbolt-gpu-diagnostics-20251116-1807.zip

something to note, I run another unraid server on a dell power-edge r510 and the gpu works great directly installed into a pci-e slot. I also have a buddy who has one working flawlessly on an amd am4 platform server, also with a direct pci-e slot.

23 hours ago, bcm3152 said:

Sorry for the extreme delay, I do have 'intel_gpu_top' plugin installed, but it is showing the integrated gpu.
Here's the Diagnostic .zip generated from the gui. I have selected the anonymize option.unraid_ms01_thunderbolt-gpu-diagnostics-20251116-1807.zip

something to note, I run another unraid server on a dell power-edge r510 and the gpu works great directly installed into a pci-e slot. I also have a buddy who has one working flawlessly on an amd am4 platform server, also with a direct pci-e slot.

the arc gpus reuse the i915 onbard graphics drivers. its either onboard or pcie slot.

so we may need to blacklist flush the onboard to get the other g card working...

the issue is duplicate kernel mod driver and werid connection for the g card via pcie thunder port.

I tried to test that theory just now. Not sure if this is a valid test but I think it disproves the two girls one driver theory. I ripped a 1080ti out of my livingroom PC and put in on the thunderbolt adapter. Installed NVIDIA driver plugin and same result. It recognizes the card in lspci, but the noeveau driver showed not in use and the nvidia-smi tool couldn't connect to the card. I even tried to assign the card to vfio-pci at boot but when I went to assign to a vm it showed no pci devices available for pass-trough.
Thinking it's more of a thunderbolt issue. Not sure of any tools useful for troubleshooting this.