Server Version#: 1.14.0.5470
Player Version#: < latest version as of 12-15-2018 for fire-tv & browser-client >
I have been trying to figure out why my 4K content buffers on playback and I was hoping someone here could bring a fresh perspective.
Server hardware (HW acceleration enabled):
Intel i7-4930k @ 3.40GHz (6 core, 12 threads)
32GB DDR3 RAM
GeForce GTX 1050
I have tried to play several 4k movies with varying bitrates from 33KB/s to 56KB/s from several devices including a wired-LAN connected Fire-TV(v2) and a high powered 4k laptop also with sufficient network hardware.
The behavior is an occasional pause of 2-6sec to buffer and then it resumes – otherwise it starts relatively quickly and working normally.
I have observed the server CPU usage to be minimal and the GPU usage to spike to near 100% in the first split second of the video playing and then will linger between 8% - 34%, where spikes toward higher values occur when the GPU is transcoding to re-fill the buffer (I assume). Otherwise everything looks normal, and I confirmed that HW transcoding was enabled for both encode & decode using the server ‘now playing’ status page where it showed HEVC (hw) => H.264 (hw).
According to this page describing transcoding compatibility with NVIDIA cards, this one should be able to handle a 4k stream (especially when there is just the 1 stream).
My CPU passmark’s around ~13k so I know it can’t handle a 4k transcode by itself, but should still be powerful enough to assist in any processing that may need to be done.
The attached logs were taken a few minutes after trying to stream a 4k movie while just the 1 stream existed. The (plex-web) client was a non-4k capable desktop with an i5-3570k and 16GB RAM
Plex Media Server Logs_2018-12-16_09-34-27.zip (7.1 MB)
Edit: For clarity, I know not all my clients can display 4k video, but transcoding to any 1080p format seems to suffer from the same buffer problem. Overall 2 of 3 clients tested have 4k-support, and 1 does not.
Can someone take a look at the logs and help me figure out where the bottleneck is?
I’d really appreciate it.
Thanks