Server Version#: 1.14.1.5488
Player Version#: Plex for iOS 5.11.1
I recently upgraded my server with a Quadro P2000. After enabling hardware transcoding, two of my users reported audio drift over time. When I disable hardware transcoding again, the drift seems to disappear. One reported the issue appearing after 20 minutes of playback, while the other reports it after just 5.
Neither the GPU nor CPUs in my server are particularly strained when I get reports of the audio issues, and the user’s buffers are usually fairly far ahead of where they are in playback. Both of these users are on Plex for iOS, one using an iPad and one using an iPhone X.
I have been unable to reproduce the issue on any device I own, although I do not have an iOS device to test on so I cannot be completely thorough. I’ve played back video both on my network and on external networks, with no issues. I’m working with the users to try to do some testing on their end as well.
I would be tempted to place the issue on the iOS client, except for the fact that it apparently goes away when I turn off hardware transcoding.
Other potentially relevant information:
I am on the West coast of the US while the users I am referring to are on the East coast
Plex is installed in an LXC container, with the GPU passed directly through to the container.
The video is being hardware transcoded from 1080p h264 -> 1080p h264, and the audio is being transcoded from EAC3 5.1 -> AAC Stereo
I’ve searched around for a while and haven’t found much, pretty much all of the similar threads I’ve found have generally seen the issue on all clients or mention nothing about hardware transcoding. If this is something known, I’d appreciate being pointed in the right direction. Otherwise, I’d be glad to provide any information needed that I might have missed.
More information will be needed. The metadata of the item(s) being played to see if there is an issue with a particular codec or a common PMS issue as it might pertain to the Remux after transcoding and/or subtitle processing.
Logs are definitely of help but player logs will definitely be needed to correlate the information between the two. ( A controlled recreation of the failure).
Also, As a test, what happens when you downgrade PMS version to public ?
Thank you for your response. At this point I don’t have too many different examples just due to normal viewing patterns, but here’s some metadata from a couple of files I remember specifically: https://drive.google.com/open?id=1yh7HX4yOYBJM2FCH9ohN88ZQx2Lb2Y6d. When I have more examples I can update that file.
My server update channel is currently set to Public, so I’m not quite sure what I would downgrade to? I also noticed that the current version on https://www.plex.tv/media-server-downloads/ is 1.15.1.780-a27ffa5be, which is above what I’m using. I can update to that version and do some testing, if that works. (quick edit - the current non-plex-pass download is version 1.14.1.5488-cc260c476, which is either exactly what I’m using or almost exactly.)
So far, the one who’s responded says he doesn’t have it enabled.
My gut was telling me that it wasn’t the client because it only happens with hardware transcoding enabled, but I now also have a report from someone else with around the same network latency who hasn’t been having audio issues. He isn’t using iOS.
This makes me wonder how much difference there is in the output depending on the type of transcoding. Maybe there’s some weird edge case here?
