Just build an Unraid server with AMD 9950x and Intel B580 for transcoding. I have Intel GPU TOP installed with GPU Statistics. Emby will not detect B580 for transcoding only the 9950x AMD APU. Is the B580 currently not supported in unraid 7.1.4?

Did you get a reply for this? I just purchased the same card and did the research afterwards. Apparently, the Unmanic support does not extend to the B580 for AVI transcoding. I understand Plex does not work atm either

i think your gpu is loaded with xe driver which is not managed by plex or gpustat.

you should use i915 driver (don't know if it takes care of b580 though)

Ok, so I spent a bit of time with google AI and seeeeeem to have my B580 working on my 9950x platform. This is Unraid version 7.2.2 if it matters. Havent fully tested yet, wish GPU Statistics was working...

Trick is to pass "--device /dev/dri/card1:/dev/dri/card0 --device /dev/dri/renderD129:/dev/dri/renderD128" to all your dockers that you want to use in 'extra parameters'.

I found other tips to use "--device=/dev/dri", but this passes through both GPU's, and I really only want the actual discrete card passing through so I can use the igpu for a VM or something later.
BUT FIRST you have to make sure that card0 and renderD128 are the right ones. I have my igpu enabled and that takes up card1/D129 on my system.

Steps: Get plugins "Intel GPU TOP", "NVTOP" (I use this for the top in CLI, you may not need?), "GPU Statistics" (this one is bugged until the kernel updates and supports ARC and then Unraid adopts that, but at least its a quick check to see if it even recognizes my card)

In unraid CLI, use these commands:

ls -l /dev/dri This should print out your card0, card1, and renderD128/D129 and whatever else.

readlink /sys/class/drm/card0 This will print out some stuff, what you want is right before /drm/card0

readlink /sys/class/drm/card1 This will print out some stuff, what you want is right before /drm/card1

lspci -nnk | grep -i vga -A3 This will let you match your actual video card to your card1/renderD129 or whathaveyou, for the above "--device /dev/dri/card1:/dev/dri/card0 --device /dev/dri/renderD129:/dev/dri/renderD128"
Now, in my case I want to use card1 which is my B580, and not card0 which is my igpu. What this actually means, I have no clue. Actually, scratch that (just asked AI, editing post). You want card0 and D128 to be on the RIGHT side of the above command, whatever your left sides are. Just so happened that mine were this way. Ask AI why, it can explain somewhat easily.

Pic example:

I do the extra CPU resources and RAM settings for each docker, doesnt matter for this guide.

Hope it helps!

Edited December 14, 2025Dec 14 by RaidUnnewb

Okay, so I was still having transcoding errors but that was all fixed in my flows and docker extra parameters now. Successful transcodes with the below settings!

I did these 2 commands in Unraid CLI per AI instructions, dont know if I messed anything up:

chmod 777 /dev/dri/*

setfacl -m u:99:rw /dev/dri/*

In the custom arguments section of Tdarr, i had to do some inputs and outputs on suggestions with AI. I did some extra stuff but that wouldnt work without running the docker in privleged mode, which I didnt want to do past testing.
Here is how I got the B580 working without having a privileged docker:
Input: -hwaccel qsv -hwaccel_output_format qsv -qsv_device /dev/dri/renderD129

Output: -max_muxing_queue_size 1024 -global_quality {{{args.variables.user.qual}}} -extbrc 1 -async_depth 4 -look_ahead 1 -look_ahead_depth 30 -extra_hw_frames 30 -b_strategy 1 -adaptive_i 1 -adaptive_b 1

Some of that output above is needed, some isnt. I don't know which is required due to the B580 and what is needed per the flow instructions. The RenderD129 part is again, because 128 is my AMD igpu, 129 is the Intel Arc GPU.

Also in the Video Encoder options, set your codec, in my case av1, and AI says its very important to use "veryfast" due to specifically the Arc or B580 compared to how igpu's handle this. Also qsv as hardware type is important.

In the Docker settings (for both Jellyfin and Tdarr), I got rid of all the extra stuff, and now just have:
--memory=16g --cpu-shares=512 --device /dev/dri:/dev/dri

The latter part is what matters, and matches what others have said. I was just missing the stuff inside the docker itself.

For Jellyfin, I also made a specific device mapping just in case: Not 100% sure its needed. I don't know how to actually test Jellyfin transcoding properly since gpu tops arent working in unraid right now.

In Jellyfin I specified QSV and typed the D129 device path.

Well, had to restart Unraid and the render devices swapped... Now my B580 is RenderD128 instead of D129.
Since the physical slot is still /dev/dri/by-path/pci-0000:03:00.0-render the AI is telling me to change the Extra Parameters in the Tdarr docker.

Since colons cant be used without errors, a hardlink needs to be made:

Type this in unraid CLI: ln -s /dev/dri/by-path/pci-0000:03:00.0-render /dev/dri/intel_b580

Now, we need to make that a script so it runs every reboot: Do this in User Scripts, running every first array start.

Finally: Change the extra parameters to: --device=/dev/dri/intel_b580:/dev/dri/renderD128

Now that were back to using D128, have to go back into the Flows Custom Arguments and change it to D128 from D129:

And it finally works again... Moral of the story: Just buy Nvidia since thats the default people code stuff for...

Don't forget to swap the render node for Jellyfin/Plex and any other containers needing the GPU..

Whether this is working, I still dont know, havent had time/ability to test.