Added another squeak in hopes this wheel of an issue gets some grease.
1 Like
@ChuckPa We just had a reply on the kernel GitHub. The first comment is that it’s probably a user space error followed by a kernel reset error.
That means that if there wasn’t a user space error we wouldn’t be running into the reset bug.
Please can you raise this internally? Dumps are provided on the GitHub page already and you stated you have a nuc yourself now.
Here is the reply on the linked post. “ So there is 2 issues here, one is that userspace is causing a hang in the VCS engine and the second one is that chip reset is failing.
The first is probably due to a bug in userspace, there is a lot of changes around engines instructions in GT 12/Tigerlake and newer, file a bug for Jellyfin and/or Plex and attach the GPU crash dump to help with debugging “Apr 19 10:08:27 NUC11TNHv5 kernel: [66313.338466] GPU crash dump saved to /sys/class/drm/card0/error”.
The second one is the chip reset failure, that is a i915 bug but fixing it will not fix the transcode. @sdutt can someone from your team take a look at the chip reset failure?”
1 Like
Absolutely will forward .
Please confirm whether or not the other media players can play successfully ?
Similar error was present on another media transcode similar to plex using vaapi however @geeooff did not report a crash using standard ffmpeg and vaapi.
I personally haven’t booted my nuc back up recently due to this. So can’t be of much help just highlighting responses and hoping it can be investigated.
We (users in this thread) were directed to raise a kernel issue, while they did agree the reset problem is theirs there seems to be push back on the reason for the crash in the first place. But that is my interpretation of the reply from i915 GitHub.
1 Like
Since EVERYONE knows I’m going to ask … 
May I please have DEBUG and any supporting kernel logs which show the errors as PMS tries to transcode?
Also please include version information for the ICR modules which are installed.
Thanks.
C.
PS: I’ve updated the internal issue with the feedback from the kernel/i915 team.
I’ll see what I can do to help move this along.
2 Likes
Folks,
What do I need do to recreate the failures?
I want to be certain to setup as much as possible in advance?
I did check and we have a NUC10i5FNK1
Is this going to be enough to replicate the problem?
@ChuckPa the key part of the setup is you are going to need a Nuc11. Seems it doesn’t matter which type, or what linux distro or kernel you happen to be using. I’m currently running in docker on Fedora 34, kernel 5.13.4.
I haven’t heard of anyone having the issue on Nuc versions prior to 11.
In case I need to chat with management,
Will this do the job ? (it serves as a place marker) If not , what to change ?
I’m having this problem with that exact NUC, Ubuntu 21.04, running Plex in a container (and not).
Or got the NUC11PAHi5 which is the non-pro version, if you want to save mgmt a little $$$ 
Here’s the error file I’ve generated immediately following a PMS crash from HW transcoding which was located at: /sys/class/drm/card0/error
error.txt (8.7 KB)
Progress!!! Thanks everyone who pitched in and made some noise on Gitlab!
Hi @ChuckPa the NUC11TNKi5 you linked has an i5-1145G7 and definitely has integrated graphics which should be fine for a testbed, otherwise NUC11paki5 or NUC11pahi5 would also be fine (and as @DarthBJW mentioned may save a few $$$ with the pak or pah, subject to availability in your region of course)
That i5-1145G7 processor and the i5-1135G7 among others of the same gen have been mentioned in the Gitlab ticket
Thanks to you and the team for helping progress this too, very much appreciated!
Thank you all for this info.
I’ll chat with folks on Monday and see what to do next.
In preparation –
With a NUC11 in hand,
Is there anything special I need to do recreate this?
( It might be easier to tell me which package(s) to install so I can replicate everyone’s setup ? )
Which type files? SDR ? HDR ? DoVi ?
Ive been able to reliably crash on demand transcoding a 1080p HEVC MAIN 10 file. you need to ensure hardware acceleration enabled for decoding and encoding (encoding is what causes the fault).
From speaking and listening to others I don’t think it matters what you are transcoding, that’s just my go-to testing episode.
For me it’s transcoding HEVC HDR 4K down to a lower bit rate.
HW acceleration should be enabled and tone mapping disabled (seems tone mapping isn’t working, so if this is enabled Plex goes straight to software decode and therefore doesn’t trigger the HW crash).
I’ve found anything in h265 4K will cause the crash to occur sooner, however even on h264 lower resolution and bitrate it will still happen - it will just often take longer.
Also I am running Plex media server from a docker container.
Unsure if this is what everyone else is doing. But this has been my setup across multiple distros/kernels and the issue has occurred consistently.
Which versions of packages are installed please?
I need an exact “Steps to reproduce” list
Someone may reply with steps. But honestly you don’t need any steps.
Here are some at a very high level.
Install an OS on an 11th gen nuc.
Install plex.
Ask it to transcode some media using HW transcoding. Wait for w crash.
Packages wise, just the packages you know are needed for plex as shown by the script you worked on.
Kernel wise, any you like. We have tried them all.
OS wise mainly tested on Ubuntu and each different recent version but issue also seem on other OS and other platforms.
Until you try it, you won’t know but it crashes with any example or steps we have tried for months.
Follow this thread with great interest! If the issues with Nuc11 are sorted I am hoping that desktop 11th gen issues should also be covered?
Isnt it a NUC 11 you need? 
NUC 10 should be supported?
