Hi all,
I've been having an issue with GPU hangs in plex. I usually use an I5-8600K, but I swapped with an I7-9700K too see if the CPU was the issue too.
Usually, I have Plex running for about a day with multiple people transcoding and I eventually get a GPU hang in the Syslog. In 6.8.3, I see the following error:
Aug 11 01:22:24 JONAD kernel: [drm] GPU HANG: ecode 9:0:0x8fcffff9, in Plex Transcoder [13784], reason: hang on rcs0, action: reset
Aug 11 01:22:24 JONAD kernel: i915 0000:00:02.0: Resetting rcs0 for hang on rcs0
Aug 11 01:22:24 JONAD kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0: reset request timeout
Aug 11 01:22:24 JONAD kernel: i915 0000:00:02.0: Resetting chip for hang on rcs0
Aug 11 01:22:24 JONAD kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0: reset request timeout
Aug 11 01:22:24 JONAD kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0: reset request timeout
Aug 11 01:22:24 JONAD kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0: reset request timeout
Aug 11 01:22:24 JONAD kernel: i915 0000:00:02.0: Failed to reset chip
Aug 11 01:22:24 JONAD kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0: reset request timeout
After this, Plex falls back to software transcoding until Unraid is rebooted.
In 6.9.2 , a GPU hang also occurs, but the chip manages to reset. This results in streams that endlessly buffer until they are manually refreshed, but stay on HW transcoding. In that case, the error resembles below:
resetting rsc0 for cs error
HW transcoding was working flawlessly for most of 2020, and at one point I added a video card for a Windows VM and messed with IOMMU groups to pass it through. I want to say I started to see problems after that, but I'm not sure. My fallback solution is to just use this Nvidia card for plex transcoding, but I rather liked using QuickSync for this purpose. Does anyone know of problems with i915 in the latest plex versions? I have tried both linuxserver and binhex repo's.
Thanks for your help!