snazz

I'm thrilled to report HDR Tone Mapping with the iGPU is working with Plex version 1.29.1.6241 with Unraid 6.11.0 on my i5 12600k system.

It's not working for me with 6.11 with Plex Pass (currently 1.29.0.6244). @flyize must be running a beta that isn't GA yet. I know he's been "on the wall" (get it?😁) in the Plex forums watching for development updates on this.

Woo! Getting closer! New Plex version dropped this morning, version 1.28.1.6018 (beta update channel.) With this Plex version + Unraid 6.11.0-rc2, I'm now able to enable both "Use hardware acceleration when available" and "Use hardware-accelerated video encoding" without crashes or any errors in the Unraid or docker logs. CPU usage is minimal if not transcoding audio. With everything set as above...If I enable HDR tone mapping, the CPU usage is pegged as if it's software transcoding and it can't keep up with transcoding a single 4K HDR10 stream (with 4 E cores only) for more than a few seconds at a time. This combo used to not work at all for me...had to disable HDR tone mapping for playback to even begin.

Disable "Use hardware-accelerated video encoding" and it should work. It's not ideal, but the best we can do for now. Edit: I did some transcode performance testing using a very high bitrate 4K HDR10 file with only "Use hardware acceleration when available" enabled. The results were disappointing. Using Tautulli to monitor the transcode speed, it was always under 1.0 (roughly .5 to .7) and the very limited stream buffer ran out after only a minute or two of playback. Performance was much better without HA enabled, even with only the 4 "E-cores" of my i5-12600k available to Plex....granted they were at 95% utilization due to the load. (The transcode speed was usually around 1.5 with this test file.) Obviously HDR tone mapping works with non-HA transcoding....so bottom-line recommendation: don't bother with enabling the HA setting until encoding is fixed (and probably wait for HDR tone mapping to be working for those of us that need it.)

It sounds odd that you don't see any CPU activity while transcoding with only the HA setting enabled. I see CPU activity on my system in the same circumstances. If not 100% sure this is accurate, but I assume transcoding video includes decoding of the original stream and encoding it in a different combination of format / bitrate / resolution, so i would expect to see CPU activity if the software encoding is occurring. Maybe you're not actually transcoding in your testing? Is it direct streaming perhaps where only the container is different?

@feraay Do you have both "Use hardware-accelerated video encoding" enabled and "Use hardware acceleration when available" enabled in Plex, or just the latter? Plex crashes due to GPU hang within about 5 minutes when transcoding with HA video encoding enabled along with "Use HA", but doesn't with only HA acceleration enabled.

Yeah, I get how it's confusing. I don't fully understand all this ICR stuff. Bottom line is that it's being worked on and it will eventually work. As far as your situation is concerned...if you still won't need HW transcoding after migrating to Alder Lake, then I'd say go for it. It will work fine without HW transcoding and would be a huge step up in performance from your Ivy Lake system.

Saw 6.11.0-rc1 was out early this morning. HDR Tone Mapping still hangs trying to start playback. Disabling that but leaving HW transcoding enabled allows playback. Too early to comment about stability. Edit: Just tried this from the plex docker's console: apt-get install ocl-icd-libopencl1 beignet-opencl-icd ...and got: beignet-opencl-icd is already the newest version (1.3.2-7build1). ocl-icd-libopencl1 is already the newest version (2.2.11-1ubuntu1).

Which Dell dock do you have? I might have the same one at home and could try passing it through to get audio. And yeah, I have a 3080 passed through to the VM. Performance is great and the VM & host OS are stable, although I haven't done much to stress both at the same time.

I've got an ASRock z690 and the 12600k. Running a 6 P core W11 VM with no issues performance wise. I couldn't pass through the onboard audio though due to too many things in the IOMMU grouping, so YMMV. Luckily I have sound via HDMI out.

Unraid's kernel version has lagged behind a bit for various reasons, but it gets updated with each release pretty much. You can use the 3080 for gaming in a VM, and you can use it for transcoding with Plex...just not at the same time. You'd have to shut down the VM for the 3080 to become available to the docker...and likely have to restart the Plex docker once that has occurred. If you expect to have more than a couple simultaneous transcodes, consider getting a 2nd cheap nvidia GPU for transcodes until the iGPU is fully supported. I have a GTX 1050 3GB card in my main server...it works great, even for a few simultaneous 4K HDR H.265 transcodes.

You have that mostly right. Once Unraid is updated to run on a Linux kernel that fully supports the Intel 12th gen iGPU, Plex (in a docker) will crash when it tries to do hardware transcoding of HDR media. You can use a Nvidia GPU passed through to the docker container for hardware transcoding in the meantime. I'm not sure if the iGPU can be successfully passed through to a VM yet...cause that could allow Plex to use it to transcode so long as the guest VM was Windows or Linux 5.16/5.17 kernel-based. BTW, a 3080ti is waaaay overkill for Plex transcoding. Also, be aware that you can't have the GPU assigned to a running VM and a running docker at the same time.

Can you be more specific about what didn't work? I have a 12600k too (I exchanged the 12900k) and have CPU cores 0-11 isolated for a Win11 VM which is now my daily driver. Unraid seems to be fine with just the E-Cores in my case.

I installed Intel GPU TOP and added "blacklist i915" to i915.conf in /boot/config/modprobe.d. Syslinux config hasn't been modified other than isolating all 8 (16 w/ hyperthreading) P cores for a couple VMs. (i9 12900k)

Based on the mostly positive feedback here, I went ahead and built an i9-12900k Unraid server last weekend. Performance and stability have exceeded my expectations so far on 6.10.0-rc2. Schnazzy12's Completed Build - Core i9-12900K 3.2 GHz 16-Core, GeForce RTX 3080 10GB 10 GB Founders Edition, Meshify 2 ATX Mid Tower - PCPartPicker

I built an Intel 12th gen (Alder Lake) Unraid server over the weekend and have been struggling to get Unmanic working with Intel QuickSync. When I try to convert a video to H.265 with the hevc_qsv plugin, it immediately throws the following errors in the Unmanic log: [AVHWDeviceContext @ 0x55eee97f8dc0] Error setting child device handle: -17 free(): double free detected in tcache 2 I have the Intel GPU TOP Plugin installed and --device=/dev/dri in Extra Parameters. Hardware Acceleration with the iGPU is working in Plex, so the issue appears to be isolated to Unmanic. Any ideas for troubleshooting further are greatly appreciated. Thanks!

Just came across this theming stuff today. Apparently, it's not working under 6.10.0-rc2 either...or at least I can't get it to work.

I'm brand new to UUD. I'm following the guide and get to the point where Grafana should start (Influxdb and Telegraf are started) but it won't. Seeing the following in the Grafana log: GF_PATHS_DATA='/var/lib/grafana' is not writable. You may have issues with file permissions, more information here: http://docs.grafana.org/installation/docker/#migrate-to-v51-or-later mkdir: can't create directory '/var/lib/grafana/plugins': Permission denied GF_PATHS_DATA='/var/lib/grafana' is not writable. You may have issues with file permissions, more information here: http://docs.grafana.org/installation/docker/#migrate-to-v51-or-later mkdir: can't create directory '/var/lib/grafana/plugins': Permission denied I've tried adding :latest to the repository line but it doesn't make a difference. I appreciate any suggestions. Thanks!

@dborisAre you passing through a hardware NIC? I'm getting an error "Generating NiceHash Miner rig identifier failed!" followed by it not connecting on the next screen. My network config for the VM: Edit: Nevermind, got it working. Not sure what it was exactly. I created a new VM with a passed-though USB NIC and it worked. I removed the NIC passthrough and it still worked, so I guess something got messed up with the original VM.

Just wanted to +1 this recent development and add a bit of insight that may help others. I have a Plex setup that does not have surround sound capability but can handle H.265 video. When I stream a H.265 video on that device that does not have a stereo track, Plex transcodes the video stream to H.264 while it downmixes the multi-channel audio to stereo. This transcoded stream now requires more bandwidth simply because it's H.264. This change to Unmanic is awesome in that it has automatically gone through my entire library adding a stereo audio track to videos that didn't already have one, and now the client I'm referring to will directplay the H.265 videos because of the presence of the 2 channel audio track. Thanks @Josh.5 !

Unraid 6.8.3 SickChill Info: Branch: master Commit: 99336df163c7e776c7931d864e84d1f0f83afe4d Database Version: 44.2 Python Version: 3.8.5 (default, Jul 27 2020, 08:42:51) [GCC 10.1.0] SSL Version: OpenSSL 1.1.1g 21 Apr 2020 OS: Linux-4.19.107-Unraid-x86_64-with-glibc2.2.5 Locale: en_GB.UTF-8 Sickchill hasn't been able to update for the past several days. I've tried removing the container & image and pulling the latest, but the update fails in the same spot even before I've restored a config from backup. Any help offered is appreciated. Thanks! Here's the latest related entry in the log w/ debugging enabled: 2020-09-10 07:21:13 DEBUG :: WEBSERVER-HOME_0 :: cur_commit = 99336df163c7e776c7931d864e84d1f0f83afe4d, newest_commit = 7435bdf348cf507cc0d026769ea2dc49869667ce, num_commits_behind = 305 2020-09-10 07:21:13 INFO :: WEBSERVER-HOME_0 :: Clearing out update folder /config/sr-update before extracting 2020-09-10 07:21:15 INFO :: WEBSERVER-HOME_0 :: Creating update folder /config/sr-update before extracting 2020-09-10 07:21:15 INFO :: WEBSERVER-HOME_0 :: Downloading update from https://github.com/SickChill/SickChill/tarball/master 2020-09-10 07:21:19 INFO :: WEBSERVER-HOME_0 :: Extracting file /config/sr-update/sr-update.tar 2020-09-10 07:21:28 INFO :: WEBSERVER-HOME_0 :: Deleting file /config/sr-update/sr-update.tar 2020-09-10 07:21:28 INFO :: WEBSERVER-HOME_0 :: Moving files from /config/sr-update/SickChill-SickChill-7435bdf to /opt/sickchill 2020-09-10 07:21:28 DEBUG :: WEBSERVER-HOME_0 :: Traceback: Traceback (most recent call last): File "/opt/sickchill/sickchill/update_manager/source.py", line 190, in update os.renames(old_path, new_path) File "/usr/lib/python3.8/os.py", line 270, in renames rename(old, new) OSError: [Errno 18] Invalid cross-device link: '/config/sr-update/SickChill-SickChill-7435bdf/.appveyor.yml' -> '/opt/sickchill/.appveyor.yml'

The update resolved the issue for me. Thanks!

Same issue here.

Does the "assigning cores in pairs due to hyperthreading" recommendation apply to AMD Ryzen CPUs which don't have "Hyperthreading", per se? I have a Windows 10 VM assigned 2 "Hyperthreaded" / SMT cores, and about a dozen docker containers spread out and pinned (in pairs) to various other cores, and I'm seeing a high context switching value (12000 - 15000) as measured by the Glances docker. In general performance seems to be decent in the VM and the docker apps (Plex, Ombi, Sab, Sick, Couch, PlexPy...), but I'm new to AMD CPUs and I'm not sure if this is the optimal approach for my setup. Any ideas on how to minimize context switching, or should I not worry about it? Thx!