Transcoding works at boot, then stops working entirely

[bnjhzenrzebrhnhkerkejz]

Setup is official jellyfin/jellyfin docker image running on Unraid, with an AMD GPU (RX 5600 XT) for transcoding, configured with VA-API

After a cold boot of the whole machine, everything appears to be working and transcoding is happening with radeontop on the host reporting reasonable usage (~9-10% for a 4K stream)

After a while, playing back files that need transcoding from any client will display "This client isn't compatible with the media and the server isn't sending a compatible media format."

Here's the last working transcode logs: https://sourceb.in/0MO9UTy9cb (at 19:23:10)
Later, without any change to the conf, the first errors: https://sourceb.in/uN7jR4PsZW (at 20:05:21)
For completeness, I retried playing the last working file: https://sourceb.in/21q6NJaixK (later, 22:34:54)

Relevant part from the logs probably is:
[AVHWDeviceContext @ 0x561d392856c0] No VA display found for device /dev/dri/renderD128.
Device creation failed: -22.
Failed to set value 'vaapi=va:/dev/dri/renderD128' for option 'init_hw_device': Invalid argument
Error parsing global options: Invalid argument

If I use radeontop from the host, I have:
# radeontop
Failed to open /dev/dri/renderD128: Invalid argument
Failed to open /dev/dri/card0: Invalid argument
Failed to open DRM node, no VRAM support.
Collecting data, please wait....

Do you think this is an hardware issue? Is my card dying?

[TheDreadPirate]

Does Unraid keep separate system logs? Might be worth checking those. As far as Jellyfin is concerned, the card no longer exists. Could be an Unraid software issue, could be a dying GPU. Maybe it isn't in the slot entirely. Maybe the GPU power cable(s) is/are loose.

A lot of possibilities.

[bnjhzenrzebrhnhkerkejz]

Oh you're right, there's hints in Unraid's logs:

Feb 29 19:38:10 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: free PSP TMR buffer
Feb 29 20:05:23 ServerName kernel: [drm] PCIE GART of 512M enabled (table at 0x0000008000300000).
Feb 29 20:05:23 ServerName kernel: [drm] PSP is resuming...
Feb 29 20:05:23 ServerName kernel: [drm] reserve 0x900000 from 0x817d000000 for PSP TMR
Feb 29 20:05:23 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: RAS: optional ras ta ucode is not available
Feb 29 20:05:23 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: RAP: optional rap ta ucode is not available
Feb 29 20:05:23 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: SECUREDISPLAY: securedisplay ta ucode is not available
Feb 29 20:05:23 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: SMU is resuming...
Feb 29 20:05:23 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: use vbios provided pptable
Feb 29 20:05:23 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: smc_dpm_info table revision(format.content): 4.5
Feb 29 20:05:26 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: SMU: I'm not done with your previous command: SMN_C2PMSG_66:0x0000003A SMN_C2PMSG_82:0x00000000
Feb 29 20:05:26 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: RunBtc failed!
Feb 29 20:05:26 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: Failed to setup smc hw!
Feb 29 20:05:26 ServerName kernel: [drm:amdgpu_device_ip_resume_phase2 [amdgpu]] *ERROR* resume of IP block <smu> failed -62
Feb 29 20:05:26 ServerName kernel: amdgpu 0000:03:00.0: amdgpu: amdgpu_device_ip_resume failed (-62).

I'll dig on Unraid's side to see if there are new power savings settings or something making the GPU go to sleep