(2015-02-11, 22:08)smallint Wrote: Is your sample a raw dump of the blu-ray or did you convert it?
It's a raw dump I extracted with eac3to. So all Blu-ray stream limitations should apply which are more restrictive than H.264 High Profile the chipset should be able to decode.
The only thing I changed is the audio track: DTS-MA -> Flac.
Checked mkvmerge v7.0.0 and mkvmerge v7.5.0, so multiplexing should't be the issue.
Maybe subtitle-overlay is to blame (meaning, taking too long)?
One thing is odd (29.9
40 fps):
Media-info:
Code:
Frame rate : 29.940 fps
Original frame rate : 29.970 fps
...
Scan type : Interlaced
Scan order : Top Field First
My old Windows-ffmpeg is confirming the uncommon frame-rate:
Code:
Seems stream 0 codec frame rate differs from container frame rate: 59.94 (60000/1001) -> 59.92 (719/12)
But the frame-rate doesn't seem to matter, because after several remuxing and frame-rate changes, the problem remains.
E.g. one remux just with audio, no subtitles. Now MediaInfo shows this:
Code:
Frame rate : 30.000 fps
Original frame rate : 29.970 fps
=> problem remains. Even when I try to mux with a PCM-audio track.
Afterwords I muxed a mkv with the video only. And the problem remains also, because the people still seem to walk too slow (I compared it to a PC's output). The video-decoding itself just seems to be too slow. This is weird, too slow for 29.97fps?
Btw, disabling de-interlacing slows down video decoding even more!
EDIT:
After some researching I recognize that this sample is PAFF rather that MBAFF. x264 just does MBAFF-encoding, so this may be encoded by mainconcept or another encoder.
MBAFF does interlacing on macroblock level, PAFF on (whole) picture level.
So I think the issue is:
MBAFF 60i is more like 30p decoding
PAFF 60i is more like 60p decoding with seperate fields (= 2 times 1920x540).
Is it possible that the hardware decoder needs
two passes for
one PAFF frame, doubling the decoding time?
EDIT2:
To be more precise, I should state that above i60 means 29.97fps fully interlaced. But after deinterlacing we should have 29.97p which should render in time? Nope.
Just for some comparison: An old and trusty ION1-chipset (2009) was totally unimpressed by Blu-ray data-streams, also - of course - every Blu-ray player available (since 2006?) is capable of up to 1280x720p60 and 1920x1080i60, called 30i here:
http://en.wikipedia.org/wiki/Blu-ray_Disc#Video - Erm, what year do we have right now?
EDIT3:
Something comes to my mind, because I also recoded the sample to MBAFF using x264 several times (what didn't help).
Recoding didn't help, because I got flickering. This means that the stream probably has got some tff/bff changes. Unfortunately I don't have the software to analyze a .h264 stream, but changing tff/bff-flags could trigger additional decoding delay, e.g. tff->bff: one additional field time-frame needed, what would be ~17ms. I don't know, what ffmpeg, or the hw-decoding threads do, but perhaps tff/bff-changes aren't handled well.
Field-changes could also explain the uncommon/inconsistent mkv-container frame-rate of 29.9
4