After updating to version 1.01 the bandwith issue seems to be resolved but now the player keeps on buffering while there is plenty of bandwith available. It seems related to high resolution 24bit flac files.
Is there anyone else experiencing the same issue?
How can i enable/gather debug logging?
I seem to be getting this on a remote Windows client with 1.0.1. Plenty of available bandwidth from the client end (business-class 200/200Mbps fibre), plenty of available bandwidth from the server end (11Mbps+ upload as measured by speedtest-cli), but still getting buffering on some files. No issues playing the same files in Plex Web with direct play enabled. Internet upload speed is configured at 9Mbps on the server, with a remote stream bitrate limit of 8Mbps (1080p).
Just did a fresh start of Plexamp and observed 700KB/s upstream for a few seconds upon initial playback for a 924KB/s FLAC - 44.1KHz, 16bit - no buffering issues. Periodic spikes up to 300KB/s upstream during playback from the server.
A few tracks into playback, the buffering issue occurred on a track. Total upstream seems capped at 107KB/s, as it stays pinned there while the track plays in small sections. Peak bitrate on the track is 931Kbps, as reported by ffprobe. In quiet/slow sections, with lower bitrate requirements, playback continues without skipping. Once any track starts skipping, the buffering issue seems persistent, as if some bandwidth negotiation kicks in to cap the maximum transfer rate.
Just had a confirmed case of buffering on an MP3 file - much less noticeable than with FLAC, due to the lower bitrate, but the total upstream on the server is showing as 40.7 - 41KB/s, which is only just above the 320Kbps of the file in question. Restarted Plexamp, and after quite a few seconds of startup buffering without playback (as if the client is consuming the entire file up until the resume point, about half-way through), playback resumes normally without skipping or stuttering. Post-restart, transfer rates of up to 117KB/s.
There’s something off about buffering, we’ll see what we can do about improving it. It’s a bit of a fine line, because in 1.0 we had an 18MB read-ahead buffer, but people reported it taking up too much bandwidth.