It's Linux and Windows only due to the Intel SDK requirement but can this patch be merged into the plex encoder? I haven't done any benchmarks but I would assume it would give Intel CPUs a nice bump in transcoding performance.
I just got an inquiry from Netgear, when I began poking around about CPU temp issues whilst dong background transcoding for Sync, that perhaps PMS would be much faster (and cooler) if it used this… Add my vote.
It should be possible to have the client opt-in and download/authorise the required libraries on the server ? this is usually how OS projects bypass limitations on hosting or including licensed material i.e. putting the obligation on the individual to use experimental / alpha builds or 3rd party libraries.
As seen in the other thread, somehow it’s tripping up on libmfx,
as rcombs puts it,
“The current support for QuickSync in ffmpeg has technical issues (it only supports full-GPU pipelines when no filtering is performed) and legal ones (we can’t distribute the MFX libraries ourselves under GPL, so this would only be useful for Windows users where they’re usually preinstalled). We’ll add support on Windows when those technical issues are resolved, and we’ll add support on Linux when a different mechanism is added that doesn’t use MFX (work on this is being conducted upstream, with our input).”
i’d have to guess the “no filtering” means that you couldn’t HW transcode with subtitles or audio processing filters either, and it would also very likely also be a very rough quality HW encode / decode with no post-processing, lots of fast encoding artifacts, etc.
Interestingly Open Broadcaster Software (OBS) and StaxRip are able to utilize Intel Quick Sync Video; both of which are free ware. If Plex were to implement this I would be most pleased.
Why we don’t see any answer or feeedback from development team?
Is it a forum outside plex.tv domain where the community have questions and other guys from this community should invent hypothesis or give personal opinions because it is not a forum supported officially by plex team?
Plex doesn’t comment much about what they’re working on.
Other than the nVidia Shield TV, PMS doesn’t support any hardware-assisted decoding but would hope people recognize it’s an eventual must with 4K/HEVC entering the picture. As I see it, there’s only two choices in regard to PC hardware neither of which is perfect:
Intel QuickSync looks promising for 4K/HEVC but only on the very latest Kaby Lake processors.
We could see other solutions using hardware acceleration. So it is confuse to users.
Why they have it and here it is so difficult.
The myths about FFMPEG does not apply anymore.
What is next myth? Licensing?
Why othe solution don’t have these problems with licensing?
If it is a really problem, why we don’t do something like DivX players devices? The license is needed. The device could run it, but to run the customer should get/buy a DivX license and configure/enable it.
It is simple. Give the responsibility to pay to users.
What the next myth? Compatible processors or GPUs?
Just say: hw acceleration will only works with these models and brands. So we could buy it.
I think the money should not be a problem. Just send these problems to us and we could pay for it.
In my use case we have a condominium of houses with optical fiber. So we would like to have a central server in a pop (mini datacenter) with Plex Server to transcode to N users.
N should be about 1 to 100 simultaneous transcoding because we have 575 houses.
How we could have the compute enough using the current implementation? It is almost impossible and very expensive.
So it’s better pay for intel licenses or GPUs than buy a lot of servers, air conditioning and space.
I don’t know the performance with hw acceleration but for sure it will help a lot to implement many transcoding simultaneous.