@Achilles - Relevant modification below that caused the strange behavior.
...MinutesAllowedPaused="60" ButlerTaskRefreshEpgGuides="1" ButlerEndHour="4" ButlerStartHour="0" VaapiDriver="i965"/>
PMS: 1.25.5.5492
Do your server log files show that its initialization is hung up by using i965?
I had checked the âPlex Media Server.logâ after I noticed the web UI wouldnât connect, but didnât see anything that appeared relevant while glancing through it.
If you made an editing error in the Preferences.xml then PMS will erase it and start over. Erasing and starting over causes a new server ID to be created (as if new)
If you can open your browser to 127.0.0.1/IP.addr.of.host and the Plex/Web browser loads, then PMS itself is running
To test this, I also added VaapiDriver="i965" at the end of my Preferences.xml for PlexMediaServer-1.25.5.5492-12f6b8c83
There were no issues.
@Achilles @ChuckPa - Iâm thinking it may have been due to me forgetting to stop the plexmediaserver service prior to updating Preferences.xml. Because I just repeated this process, this time ensuring to stop PMS first, make the edit, then start PMS and everything is fine now.
Now to test a VC-1 movie to see if I get the annoying pixelation.
On a side note, do either of you know if this regression bug is tracked somewhere so that those of us who have implemented this workaround know when it is fixed and we can revert the change?
It took years to get Intel to fix the issues with the intel-vaapi-driver. A 3rd party GitHub person figured out the issue and presented the solution to Intel. There is a GitHub issue for the same issue with the Intel-Media-Driver (iHD) for almost 3 years now. I doubt they resolve this anytime soon. There is no need to use iHD with Coffee Lake CPUs and older. iHD is required by Comet Lake and newer CPUs.
Thank you!
Any artifacts?
Half-way through The Bourne Identity and nothing yet. I had seen it on all VC-1 movies recently; The Matrix Trilogy (especially during the rainy sky fight scenes between Mr. Smith and Neo), The Thing (especially during scenes where fire was involved), etc⊠Iâll post an update shortly once I get through a couple; but looking good so far.
I am quite sure you wonât encounter any more macroblocking artifacts with i965. As a ninja, I personally monitor and test rigorously for proper HW VC-1 decoding against new PMS versions. I currently have both a Hades Canyon NUC (Kaby Lake G) and a Quartz Canyon NUC (Coffee Lake) configured with VaapiDriver="i965" that is used to test for any regressions.
This is the end of the line for me using Intel QSV. Moving forward with any new HW, I will be using AMD Ryzen and Nvidia P2200. I prefer AMD Ryzen over Intel CPUs these days and with QSV/iHD/VC-1 regression there is no reason for me to pursue the new Intel CPU offerings. Nvidia also does a much better job with supporting and fixing issues with video codecs.
No artifacts. Thanks again for chiming into this thread with the workaround. This seems like something that perhaps Plex should include in the support article for hardware acceleration as a known issue and workaround. Even though it isnât a Plex issue, it would still be helpful knowledge for Plex users to have the best experience.
I hear you on QSV. Intel is pretty clear on their site that they donât officially maintain the Intel graphics drivers for Linux (per 01.org).
My dedicated PMS is a Supermicro SYS-E300-9C with several customizations for networking, storage, and transcoding optimization; along with ultra-quiet fan replacements since the factory ones are insanely loud. It has served me very well over the years.
Sometimes a little push helps, so thank you for the link!
If you find this post through google in a month or so and still nothing happend: Feel free to post a comment on github.
I also bumped the GHI. Its now assigned.
