AI pulled me through this one. For those of you who hate it (I almost do hate it!) Here's what I did to make it start working.

1. I was coming from an nvidia gpu. When I pulled the card, Plex docker (lscr.io/linuxserver/plex) stopped working. I had to remove all the custom entries I'd put in years ago for that nvidia card - including a --runtime=nvidia thing that was behind the Advanced toggle.

2. AI wanted me to change my Network from Host to Bridge. That broke port forwarding and I couldn't connect to my Plex server, so I changed that back to Host.

3. I had to add two Devices to the docker template.

   1. Add another Path, Port, Variable, Label or Device

   2. Select Device

   3. Name: /dev/dri/card0

   4. Value: /dev/dri/card0

   5. If you want to make sure these things are present, like I did, you can run:

root@mayorgoodway:~# ls -la /dev/dri
total 0
drwxrwxrwx  3 root root       100 Oct  1 03:42 ./
drwxr-xr-x 17 root root      4080 Oct  1 10:46 ../
drwxrwxrwx  2 root root        80 Oct  1 03:42 by-path/
crw-rw----  1 root video 226,   0 Oct  1 10:44 card0
crwxrwxrwx  1 root video 226, 128 Oct  1 03:42 renderD128
root@mayorgoodway:~#

1. (formatting blah)

2. (formatting blah)

3. (formatting blah)

4. I did add this into Extra-parameters "--group-add 44". This is what AI said about it:

   1. C. Add the video Group Access

      Docker containers often need to be part of the correct host group to access the GPU device nodes. Since your ls -la output shows the devices belong to the video group, you must grant the Plex container access to this group.

   2. 

5. AI asked me to double check I had some boxes ticked in the Plex settings. They were already checked because of the previous nvidia stuff I had in there. They were:

   1. Go to Settings (wrench icon) → Transcoder.

   2. Check the box for "Use hardware acceleration when available".

   3. Check the box for "Use hardware-accelerated video encoding".

   4. Click Save Changes.

6. I put on a South Park, forced it to transcode, and badda bing: (HW) showed up and I could see the card doing it's thing.

7. Doing it's thing. I optionally installed a Community Apps > "Intel-GPU-TOP; ich777"

1. This added a command line interface program called intel_gpu_top. That was slick for watching the card do it's thing.

2. it cal also give you a list of your available devices. Here's mine

root@mayorgoodway:~# intel_gpu_top -L
card0                    Intel Dg2 (Gen12)                 pci:vendor=8086,device=56A6,card=0
└─renderD128            

root@mayorgoodway:~#

So anyhow, no really biggie.

Ah another slightly unrelated thing. I have a PiKVM v3 Pre-assembled to help manage this headless server. Because my CPU has no onboard GPU, the HDMI on my motherboard is dead. When I put the new Sparkle Intel Arc A310 ECO in there, and it detected there was no monitor attached: no POST/boot. I had to swipe the monitor mouse and keyboard from my gaming rig to get the unraid server to BOOT. I ordered one of those dummy HDMI things. I also posted over at pishop.com forums about WTH. I'll probably end up buying the V4 of the PiKVM so I can still remote access this puppy.

Edited October 1, 2025Oct 1 by rutherford

too bad my A310 wont even bind a driver to the card so /dev/dri isnt even a thing