[Milhouse]

That's a useful piece of information. Can you confirm that top of master is still in the "worse" state?

Not sure if it's good news or not, but top of master is still in the "worse" state, testing with the following settings:

Code:

<network>
    <cachemembuffersize>20971520</cachemembuffersize>
    <readbufferfactor>4.0</readbufferfactor>
<network>

When resuming Blade Runner Final Cut (DD5.1 audio selected), what follows is the codec info for the movie as playback resumes:



I may be interpreting this incorrectly, but my understanding is that "B 0%" represents how full (or empty) the buffer is, with aq and vq similarly representing the audio and video queues. I have no idea what the "vb:" number represents, it tends to be all over the place, anywhere between 0 and 9830400.

Although the movie begins playing OK, after 10-20 seconds it stalls while the buffer fills (which is taking place in the stalled screenshot below):



However, if the movie is started from the beginning (and not resumed) I see quite different results - the buffer will begin to fill from the very beginning which doesn't happen when resuming:



and eventually the buffer will reach 100% (99% in the screenshot below):



So there seems to be a basic difference in buffering behaviour when starting from the beginning and when resuming a movie (some movies, or just this movie) - is this important/relevant?

However, as the bit rate picks up in the "started from beginning" movie, the buffer will eventually run dry and playback stall while the buffer is refilled, so the readbufferfactor doesn't seem to be as effective as hoped, or isn't working for me as the network read rate is pretty much static with no obvious peaking, not even when the movie is stalled and the buffer is being filled. I would have expected to see more network reads as the buffer needs to be filled, and before it runs dry.

In the Jul 21 build, the buffer doesn't run dry and peaking of network traffic is evident (and with no stalling). I'll try and work back through master to see if I can narrow down the relevant commit(s) - it's gonna take a while...