Fatal Error with some HEVC HDR - VM - Hardware acceleration 0 GPU passthrough

[yannn]

hi guys,

I'm not sure where to dig more so ask for some help.

The problem: 

• On some large (80GB) 4K HEVC HDR videos I get Playback error. 
  
• Yesterday it worked, today stopped, sometimes after a reboot it works. 
  
• Some other videos work good (on some I still see some artefacts or lags, sometimes all is good): like full HD, or 20-30GB size
  

I'm trying to find is it an issue with configuration or some bottleneck of the hardware.

My setup

• Lenovo m715q, Ryzen 5 2400 GE (4 cores, 8 threads)
  
• Radeon Vega 11 built in CPU. It should support HEVC decoding (link) (look for Raven, VCN 1.0) + here is link with codecs support
  
• HDD (quite old, 5400 rpm, ST1000LM024, 1TB) used for media storage
  
• SSD (WD SN730 NVMe 256GB) to run Proxmox VM, Transcode folder is also on the same SSD but separate volume (so don't to back it up). I see Jellyfin creates files there. 
  
• Lan 1Gbps: TV and Pc are connected by Lan
  
• Proxmox with 2 VMs: 1 home assistant, 2nd - this media server
  

I think I managed to make GPU passthrough to Jellyfin (not really a pro there). Using AMD one (obvious).
How I see it that is works:

• Jellyfin allows to use hardware decode (screenshot). I tried 
  
• As it's on, used CPU usage on my VM (Ubuntu 24) is very low
  
• As I turn it off - all videos works well (still sometimes with lags), but CPU load (6 cores) is around 90%
  
• Still have doubts: whats lspci returnslspcihttps://share.cleanshot.com/XbzKbSs5
  

Logs
Attached

Question

• Any ideas what to check? Should I enable Encoding to HEVC in Jellyfin?
  
• Is it a limitation of Radeon Vega codecs? I know it's better to have Intel with QuickSync, but it's what I have now.
  

Many thanks!

[theguymadmax]

It's recommended to use VA-API on Linux, not AMF for AMD GPUs.