[MAC] VDADecoder performance over HDMI on new Mac Mini?

I am noticing poor FPS stats and continuous dropped frames when playing back a newly created BluRay rip with VDA turned on. The FPS is half of what it is suppose to be and dropped frames make the movie unwatchable.

But when I turn the VDA off, all is fine and the only dropped frames I see is when the movie first starts. Below is some info:

This is with VDA turned off:



This is with VDA turned on:



Here is the MediaInfo for the movie:

Quote:*** MediaInfo Mac // Plain text file report
2010-07-07 14:30:44 -0500
Information for File: Righteous Kill.mkv

General / Container Stream # 1
Total Video Streams for this File -> 1
Total Audio Streams for this File -> 1
Video Codecs Used -> AVC
Audio Codecs Used -> AAC LC
File Format -> Matroska
Play Time -> 1h 40mn
Total File Size -> 7.30 GiB
Total Stream BitRate -> 10.4 Mbps
Encoded with -> HandBrake svn3424
Video Stream # 1
Codec (Human Name) -> AVC
Codec (FourCC) -> V_MPEG4/ISO/AVC
Codec Profile -> High@L4.1
Frame Width -> 1 920 pixels
Frame Height -> 816 pixels
Frame Rate -> 23.976 fps
Total Frames -> 145008
Display Aspect Ratio -> 2.35:1
Scan Type -> Progressive
Colorimetry -> 4:2:0
Codec Settings (Summary) -> CABAC / 5 Ref Frames
Codec Settings (CABAC) -> Yes
Video Stream Length -> 1h 40mn 48s 48ms
Bit Depth -> 8 bits
Video Encoder -> x264 - core 100 r1659 57b2e56
Video Encoder (Settings) -> cabac=1 / ref=6 / deblock=1:0:0 / analyse=0x3:0x133 / me=umh / subme=9 / psy=1 / psy_rd=1.00:0.00 / mixed_ref=1 / me_range=32 / chroma_me=1 / trellis=1 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=0 / chroma_qp_offset=-2 / threads=3 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / constrained_intra=0 / bframes=6 / b_pyramid=1 / b_adapt=2 / b_bias=0 / direct=3 / weightb=0 / open_gop=0 / weightp=2 / keyint=240 / keyint_min=24 / scenecut=40 / intra_refresh=0 / rc_lookahead=40 / rc=crf / mbtree=1 / crf=19.0 / qcomp=0.60 / qpmin=10 / qpmax=51 / qpstep=4 / ip_ratio=1.40 / aq=1:1.00
Audio Stream # 1
Codec -> AAC
Codec (FourCC) -> A_AAC
Audio Stream Length -> 1h 40mn 48s 42ms
Number of Audio Channels -> 2
Audio Channel's Positions -> Front: L R
Sampling Rate -> 48.0 KHz

I didn't think a log would be warranted, but can provide if needed.

let me guess, new macmini ?

noticed something strange in your MediaInfo dump..

Codec Settings (Summary) -> CABAC / 5 Ref Frames

yet

Video Encoder (Settings) -> cabac=1 / ref=6

Which is it, 5 or 6 ?

native bluray h.264 is typically 4 ref frames so I'm guessing you are transcoding.

davilla Wrote:let me guess, new macmini ?

Yes, the new MacMini

davilla Wrote:noticed something strange in your MediaInfo dump..

Codec Settings (Summary) -> CABAC / 5 Ref Frames

yet

Video Encoder (Settings) -> cabac=1 / ref=6

Which is it, 5 or 6 ?

native bluray h.264 is typically 4 ref frames so I'm guessing you are transcoding.

Shit...I forgot to change my ref statements in my preset to stay within spec for 1080p. Yes, I'm Transcoding to save space on drive. I will re-encode from source with ref=4 and see if that makes a difference.

...forgot, settings were set to ref=6 with CABAC=1.

Sorry to bang on but this looks just like what I've been trying to get attention for here: http://forum.xbmc.org/showthread.php?tid=76614. Apart from the content in the movie frame grabbed, the stats and framerate at the too look very familiar. (my own similar screenshots are posted there).

I posted the media info output on one of my problem files and it's not showing that cabac/ref discrepancy as far as I can tell, so I'd guess that might be a red herring. If you think transcoding is still an issue an obvious thing to try is to play the untranscoded source from blu ray; those are playable as mkv files out of MakeMKV so that's easy; but I'm in bed now so I'll try that tomorrow. :-)

I've also played my sourced .mkv with and without VDA. It plays fine with both, but I don't notice any CPU advantages either way, but the framerate (if I remember correctly) was fine either way. I'll test again later tonight.

Rachel Wrote:Sorry to bang on but this looks just like what I've been trying to get attention for here: http://http://forum.xbmc.org/showthread.php?tid=76614. Apart from the content in the movie frame grabbed, the stats and framerate at the too look very familiar. (my own similar screenshots are posted there).

I posted the media info output on one of my problem files and it's not showing that cabac/ref discrepancy as far as I can tell, so I'd guess that might be a red herring. If you think transcoding is still an issue an obvious thing to try is to play the untranscoded source from blu ray; those are playable as mkv files out of MakeMKV so that's easy; but I'm in bed now so I'll try that tomorrow. :-)

Please do not assume that common symptoms indicate a common problem. Video timing is a rat's nest of interdependencies.

Important points to note, dvdplayer is dropping frames like mad, vq/aq look good so it's being feed demuxer packets. vcpu is sky high and a/v is too high positive. That means dvdplayervideo is late decoding frames and is dropping them to catch up. But this timing is determined by pts timestamps.

All I see so far where is it looks like the pts timestamps are not correct or out of sync. Now other things could also cause this symptom (slow rendering for example). The 5/6 ref frames and both being greater than 4 (bluray spec for this size video) could mean that VDADecoder is internally dropping frames which means that pts timestamps that get passed through the decoder also get dropped. This can cause all hell to break lose

Once I'm done transcoding with the proper reference frames (15 hours from now), I will try again and repot back. My first time encoding 1080p content (didn't have a means for smooth playback before), so excuse my newbiness...

i tried again with HandBrake's High Profile setting, which uses 3 ref frames and cabac no change. i also tried playing the untranscoded bluray rip which does indeed use 4 ref frames, and that *still* showed the same problem.

MediaInfo output on the bluray rip:


And the High Profile transcode of chapter 1 (actually i note a discrepancy in the ref frames this time.):