Any news on when hardware transcoding will be available for Linux/nvidia? I just bought a Plex pass for this exact feature and now found it doesn’t work.
Bumping this ^
Bump! I just got a used 1050 ti for my Linux Plex server and it has no problems transcoding content with my own ffmpeg binary. With Plex Media Transcoder I’m being bottlenecked by the CPU since it doesn’t do hardware accellerated decoding. It would be nice to know if this is even on the roadmap and perhaps some kind of estimate? Otherwise I’m going to have to switch to emby.
Are there any news on how this process is going?
@ChuckPa Ive seen @AnonymousRetard explain his delving into the alpha versions of the server and he seems to find a, while prolly buggy, almost fully functional hardware decode setup, minus the proper command from plex to the transcoder. Is there a reason that you could share as to why that, to my knowledge, single command hasn’t been enabled? Also potentially a timeline as to when this will be available? I was about to redo my Ubuntu server with Windows Server 2016 in order to get the ability to transcode more than 2 streams at a time when I saw his findings. THANKS!
I was just finishing up some testing and my ears were burning. Now I know why 
I just completed the build for nVidia GPU encoding on Linux.
The transcoder team is doing there thing with the actual codecs independently of me.
I just completed implementing nVidia PCI-E (GT-1060 type) GPU card on QNAP.
It works in the lab.
I’ve sent it off to QNAP for them to kick around a bit (which they will do VERY quickly).
A few of us have nVidia cards in our personal QNAP machines as well and are going to kick it.
How does this translate to what you’ll see and when? PMS 1.15.0 is on the way out the door now (final QA). I am playing with 1.15.1 on the NAS boxes. Fingers crossed and nothing else catching fire, general Linux packaging is next. I don’t have those resources in my hands (AIO machine here) so will need work that out.
Any input from anyone about paths & how to verify package names is very helpful.
I ask this because I have a very limited amount I can do in systemd startup command so I must find a way that will work regardless of where and how the nVidia drivers are installed. PMS has to remain flexible such that it will adapt with driver changes and not require coding changes on my part.
“I just completed the build for nVidia GPU encoding on Linux.”
Encoding AND decoding?
Please re-read what I wrote.
I’m confused as well.
the thread is about nvidia linux DECODING.
Wasn’t ENcoding already working?
Chuck’s post indicates ENCODING.
Did Chuck typo and actually mean DEcoding?
Or did I also miss something something somewhere?
I got a bit overexcited
. So the wait for an update on the transcoder continues
In the post above my reply, please see where I was tagged to provide feedback. I updated to where I am at in the effort.
Ok, well the post you appear to be replying about also talks about decoding.
I guess the confusion comes from, what the difference is between your work on the ENcoding portion, and what the transcoder team is ‘doing their thing’ with the codecs.
I apologize for any confusion. I think my reply was written sometime around midnight.
In the future, I will be more surgical in my reply.
If I have no “decoding” information to report, I won’t mention the other work I’ve completed which supports performing that decoding.
ok…
well I guess then suffice to say, work continues and it will be done if and when its ready to be done. 
Perhaps you can clarify how the work you’re doing on encoding supports decoding.
Afaik, encoding has worked for a very long time with Plex builds. Both ffmpeg and plex supported nvenc. As long as the Nvidia driver is installed on the host, it worked out of the box. I fail to see the connection there, maybe I’m missing something.
The work I’m doing does:
- Provides the basis for using a GPU card in a NAS box ( where CPU is a premium ) by having done QNAP first (easiest to standardize). I know that, whatever I do on the QNAP, must be portable to other Linux platforms and GPUs. This includes AMD GPUs.
- With this now in place, I can take PMS to the next step. This means it will be GPU-aware when it starts rather than waiting until the transcoder is called on. It provides Engineering that information upfront.
- My work is coordinated with other transcoder work already in progress.
- The end goal is to unify all the capable Linux machines, whether they be desktop/workstation, server, or NAS into a common ABI.
- The transcoder doesn’t have to worry about which machine it’s actually running on or its configuration. It will have all information needed when invoked. This will result in faster startup times by avoiding the need to probe all possible devices.
I feel a bit like the messenger who is getting shot because he brought news which wasn’t wanted. I thought it would be nice to know. I could just as easily wait until product launch and let it come out in the Release Announcement.
To be 100% clear: My job was to make the linkages standard. I was NOT to add new functionality
thanks for the extra insight and info.
I don’t think it is that anyone wants to shoot anyone, just more a general confusion based on the info provided and the context of the thread.
I don’t think anyone wants to discourage any amount of info provided now or the future.
Thanks for clarifying! I appreciate the work youve done and welcome any communication from you guys. Beats being in the dark by a mile.
I appreciate all of the input you give and information you provide! Thank you for the active work you provide on the forums!