Judder - non-existing SPS 1 referenced in buffering period
#31
pennant Wrote:Does the SVN have FFmpeg 0.6 yet? Maybe that will solve the problem.
Hi,

I'm having the same issues (tons of:
Code:
ERROR: ffmpeg[60FD9710]: [h264_vdpau] number of reference frames exceeds max (probably corrupt input), discarding one
in xbmc.log while watching video "pulsating" once a second) with a SVN checkout of the pvr-testing2 branch from July 2nd, which in turn also syncs trunk modifications from time to time, compiled on Gentoo x86_64 with external ffmpeg-0.6 and internal python. The issues I'm seeing are when using the VNSI addon which nicely integrates VDR as a TV backend into XBMC when watching 1080i HD-channels from DVB-S2. Interlaced SD or 720p HD channels are displayed OK, just such a pity that deinterlacing options are not remembered not even between channel switches...
Could something else than ffmpeg be the problem here?
Reply
#32
as a wild guess, I point my finger at vdpau.

is it possible to use vaapi to rule it out? or does it use vpdau indirectly as wiki suggests?
Reply
#33
have you tried this one? maybe both issues are related.

http://forum.xbmc.org/showpost.php?p=560...stcount=85
Reply
#34
tried the latest xbmcfreak livecd (based on lucid) and the issue is still there.

I'm updating to latest svn in a minute.

I'll try the live cd on a few computers to see if it's reproducible.

EDIT - I tried it on one of my computers which has a gt240 and I couldn't see any judders.

I also tried to slow down and speed up the sample. I can't see any judders if it's below 24fps. It's hard to see when the fps doesnt match the display refresh. But it seems like a performance issue.

I think this because if I change scaling type to nearest neighbour, and display size to "original" instead of "normal" the avatar intro plays nicely (doesn't help with some of the other samples). This seems odd because it's displaying a 1920x1080 video on a 1920x1080 display (1:1 mapping i would have thought requires no scaling).

Anyway. Looking forward to trying ffmpeg 0.6 when its in SVN (unless it already is, in which case it hasn't helped)
Reply
#35
Any updates regarding this issue? I've been using XBMC Windows w/ DSplayer.
Reply
#36
still getting the same judders in the latest svn. as well as the latest xbmcfreaklivecd.
Reply
#37
That sucks man Sad

I've used multiple builds of XBMCFreak Live, plus Windows 7 on the hardware in my signature and I've never had this issue on .m2ts, .mkv, .mp4, etc. With all the time you've spent on this, maybe it's time for a different solution...
http://wiki.xbmc.org/?title=HOW-TO_make_...s_a_shell)

NS
Image
Reply
#38
NotShorty, could you post your video settings? (24hz?, av sync on?, etc.)

Also, are your .m2ts and .mkv untouched blu-ray rips? And when you press 'O' to view video stats, does the error % remain at 0?
Reply
#39
My HTPC is disassembled at the moment, but I pretty much leave the video settings as default from the XBMCFreak builds. I've messed around with the sync playback to display, etc. but didn't notice a big difference...

I've tried .m2ts untouched and transcoded to 13GB or 8GB .mkv files. Sometimes I'll get a few dropped frames as the video loads, but after that the error rate is 0 if I recall correctly.

NS
Image
Reply
#40
NotShorty Wrote:That sucks man Sad

I've used multiple builds of XBMCFreak Live, plus Windows 7 on the hardware in my signature and I've never had this issue on .m2ts, .mkv, .mp4, etc. With all the time you've spent on this, maybe it's time for a different solution...
http://wiki.xbmc.org/?title=HOW-TO_make_...s_a_shell)

NS

That's how I've had it for the last month or so. I've currently got win7 and xbmcbuntu minimal karmic dualboot. I use the win7 almost exclusively now.
Reply
#41
Hi,

I'm joining this late. I've got my own thread here. I've got the same errors in my logs too.

I'm using a Revo R3610 which has the problem. I've borrowed an older R3610 and that works fine. Both using XBMC Live or XBMCFreak CD. Both 2GB RAM, same BIOS version, same processor etc. Only differences I can see now are different WLAN controller (atheros in old one ralink in new one) and different size HDD (160 in new r3610, 250 in old one), and also memtest says the memory in the older box is faster than the memory in the new one! Whether any of those things are causing it I don't know. Whether there's something else they've changed and it's either slower or causing problems with VDPAU - again, I don't know.

I've yet to look at the logs for the working revo to see if the same errors exist. And I've yet to try my non-working R3610 with Windows.

Those are my tasks for tonight and then I'll report back.
Reply
#42
Installed Windows 7 on my R3610. Installed latest nvidia drivers. Installed latest nightly xbmc win32 build.

Avatar 40GB untouched rip works absolutely fine.

So it's not a hardware fault.

It seems like it's either a linux nvidia vdpau driver issue (ie it's not happy with the hardware?) or a vdpau/hardware/xbmc incompatibility. MAybe xbmc or nvidia will address this at some point.
Reply
#43
Here's the log for the r3610 running latest xbmcfreak cd playing 32GB Avatar bluray rip perfectly...

http://pastebin.com/6VdKiLdw
Reply
#44
Hi,

I've been following your posts in the other thread.

I personally believe it's a software bug. Where the bug is is anyone's guess.

I don't have the linux knowledge to install a different player which supports vdpau. I think I might give it a try though. If anyone knows of a player that's relatively easy to setup so it's using vdpau please let me know.

regards,
Reply
#45
Might want to look at your xorg.conf settings. At 2:14 min. into Avatar when the ship in space is moving toward your screen I would notice small stops as the space ship moved forward. It took me a day to finally get the xorg.conf right for my display and now it plays perfectly. Although I am still getting the SPS 1 errors in XBMC.log but the film is perfect on screen with a total of 15 dropped frames by end of movie.
9.10
r26018
Reply

Logout Mark Read Team Forum Stats Members Help
Judder - non-existing SPS 1 referenced in buffering period0