SOLVED: Slow transcodes on arc a380

[sthtyekujetyk]

It used to transcode fine but now it does so very slowly.
I am pretty sure it is not transcoding through the cpu because i don't see any spikes, it looks like it is just a very slow gpu transcoding now

I think this problem started when i updated the jellyfin image in my docker container, currently running 10.9.2
Edit:
I just tried to spin up a new container with new volumes, using version 10.8.13 and the problem is still there.
Could this be some kind of hardware failure? How do i investigate this?

Docker compose:

Code:

  jellyfin:
    image: jellyfin/jellyfin:latest
    container_name: jellyfin
    user: "1006:2002"
    group_add:
      - "105" // render group
    network_mode: 'host'
    devices:
      - "/dev/dri/renderD128:/dev/dri/renderD128"
    volumes:
      - "/mnt/media:/mnt/media"
      - "jellyfin_cache:/cache"
      - "jellyfin_config:/config"
    restart: 'unless-stopped'
    extra_hosts:
      - 'host.docker.internal:host-gateway'

Transcode log
Logs
Settings

[TheDreadPirate]

What is the underlying Linux distro and kernel?

Also, long shot, instead of passing in the specific render device, try passing in /dev/dridev/dri. It has made a difference for other users. No idea why it would matter.

Also, what CPU do you have? Is it an Intel CPU with an iGPU?

[sthtyekujetyk]

I am running proxmox with kernel 6.5.11-4-pve
On top of proxmox there is debian vm with kernel 6.6.13+bpo-amd64

The cpu is an amd ryzen 9 7900
I did try the /dev/dri trick but it made no difference
The interesting part is that this setup used to work correctly
I made this post a while ago which may contain some useful information

[sthtyekujetyk]

I just rebooted multiple times my proxmox host and finally it started transcoding again
Apparently there is an issue where occasionally it does not transcode properly after a reboot

Will dig deeper and post more info if i find anything later