Log in

View Full Version : Glitch on Lagarith output


Lyris
22nd August 2012, 06:53
Anyone else get this? Anyone know what causes it?


This pops up from time to time with me - and has done across various different builds and systems. The glitch always occupies the last 1/4th or so of the frame.


http://img855.imageshack.us/img855/1406/whatcauses.jpg (http://imageshack.us/photo/my-images/855/whatcauses.jpg/)

Uploaded with ImageShack.us (http://imageshack.us)

Taurus
22nd August 2012, 07:26
My guess:
You are using an old AMD cpu without sse2 instruction set.
Nothing you can do about this.
An old version of lagarith might work.
Or buy an uptodate processor/mainboard later intel pentium4 :p;).

Lyris
22nd August 2012, 19:44
Thankfully not. Here's my system:

Intel Core i7 950 (3.07ghz overclocked to 3.8ghz) - perhaps the OC is the problem - but I've had this on older CPUs that weren't, so...
22gb of RAM (16gb usable in Win7!)
Win7 x64

Dark Shikari
22nd August 2012, 20:00
Lagarith relies on floating point math in the arithmetic coder, which nearly guarantees that errors like this will occur eventually, depending on the phase of the moon and so forth. The format should generally be avoided if possible.

Taurus
22nd August 2012, 20:41
@Lyris:
Are you using the 64bit version?
The changelog is mentioning 64bit trouble.
Why you dont try other lossless codecs?
May I suggest UTVideo or FFV1(ffdshow).
There was a comparision somewhere here on the board
a few days ago.

Lyris
22nd August 2012, 23:31
DS - interesting. The programming behind this is a mystery to me, so thanks for the info.

UTVideo seems like a good alternative to try out :)

Richard1485
23rd August 2012, 00:36
The format should generally be avoided if possible.

Which lossless codec would you recommend instead?

Dark Shikari
23rd August 2012, 01:08
Literally anything else. Floating point in any video format is very bad, doubly so in a lossless format, and putting it in the entropy coder means that any rounding error means not merely "not lossless", but "complete loss of video until the end of the frame".

Useful formats might be Huffyuv, Utvideo, FFV1, lossless H.264 (x264), or others depending on your use-case.

manono
23rd August 2012, 10:54
If I've used Lagarith once I've used it a thousand times and never yet seen anything remotely resembling that picture (not that it couldn't happen, I guess). However, once when I had a stick of memory go bad I would get a bunch of frames like that during an encode. My suggestion would be to run Memtest86 or some other memory testing program.

Atak_Snajpera
23rd August 2012, 11:16
also just in case check cpu stability with linx (linpack gui)

Richard1485
26th August 2012, 00:36
Useful formats might be Huffyuv, Utvideo, FFV1, lossless H.264 (x264), or others depending on your use-case.

Does Huffyuv support YV12 though?

manono
26th August 2012, 02:24
Does Huffyuv support YV12 though?
There are several lossless YV12 codecs:
VBLE Codec (by MarcFD): A huffyuv based encoder discussion.
LocoCodec (by TheRealMoh): see also here.
ffvfw codec - has various modes, in particular HuffYUV yv12.
Lagarith codec (by Ben Greenwood) - better compression than Huffyuv but slower
http://avisynth.org/mediawiki/FAQ_YV12

The ffvfw codec (which includes HuffYUV YV12) can be found here:

http://www.free-codecs.com/ffvfw_download.htm

Richard1485
26th August 2012, 22:06
Thanks. :)