Server Version#: Version 1.15.1.710
Player Version#: Latest as of 2/2018
I recently upgraded to a QNAP TS-453 2018 Model. It works great except i have notice thumbnail generation is hilariously slow.
Not bad for keeping up to date, but can be days if a new show is added.
I understand that this is not a full powered PC, however it has Hardware Decoding for transcoding. This seems to work when something is being streamed with transcoding, it says (hw) on the stream.
However for thumbnails, it is extremely slow and uses 95% cpu. Does thumbnail generation not use the hardware decoding?
Is there a way to check?
If not, is that a bug or expected result?
Could it be possible to use the VAAPI MJPEG encoder combined with the qsv filters for Intel GPUs at least? NVENC also has an MJPEG encoder. Even if you had to do it in 2 passes, first pass to encode the resized and 0.5 fps mjpeg file, and then extract all frames from that to build the bif file. Is the BIF file much more than a mjpeg stream?
This works on my Xeon E3-1245V6 using Linux with ffmpeg 4.0.4 with very little CPU usage. Creates the jpegs at the same size and rate that Plex currently does.
My thumbnail generation just became super fast.
I’ve glanced at changelogs, but didn’t see anything that blatantly said HW acceleration was implemented.
I can’t track it down at the moment, but I’m pretty sure one of the devs said it was still software-based, because there were problems with graceful failure when trying to use acceleration for the task.
Yea, I don’t have any Hardware enabled, or any that currently works, and these thumbs are flyin’.
I run a LOT of 480p TV Shows and you’ll be lucky to open the Dashboard before they’re done! (in fact, I thought it was broken for a while - until I played a file and scrubbed the bar to find - they’re done…lol)