Log in

View Full Version : 3 fast passes or 2 slow ones?


Marsu42
9th October 2005, 22:37
Dear Dr. H264/AVC,

while x264 encoding dvd material two questions occurred to me:

1. Afaik, the only difference between last pass (--pass 2) and n-th pass (--pass 3) ist that the latter does update the stats file. I am wondering what performance penalty this has, if it is small I'd just always use --pass 3 as second stage, have a look at the encoded file and then decide whether to run another --pass 2 on it.

2. Which should usually give more quality when encoding your average hollywood movie, 3 passes w/ a fast first pass using reduced motion search and partitions followed by a second and third medium speed pass (e.g. --me umh), or 2 passes w/ the same high quality settings (e.g. --me esa)? The total processing time would be equal in both cases.

...to me as a long-time XviD user the underlying question is: In which cases does a 3rd pass do some good and what does it do anyway?

Concerning Dr. Doom9, please note that I am not asking which is better, but which is more likely to result in a higher quality encode.

TIA for your input :-)

Sirber
9th October 2005, 22:41
In any case, best not using "--me esa".

Marsu42
9th October 2005, 22:45
In any case, best not using "--me esa".

... this was just an example for different encoding speed/quality settings, I surely wouldn't use --me esa in reality since is too slow. Usually, my choice is either --me hex or --me umh.

Sirber
9th October 2005, 22:48
1) You could try. I don't think it would hurt the quality though.

2) 2 pass with same quality would be faster, but 3 pass with 1 fast would be better. You might only see a difference if your bitrate is low though.

(All the above strictly IMO and is purely theorical ;))

foxyshadis
9th October 2005, 23:24
It could be useful in really low bitrate encodes, but usually it's most useful in really high bitrate encodes, where you have quite a bit to throw around for fine-tuning. Otherwise 3rd pass doesn't make noticeable differences (in my tests, your circumstances may show different). It can't hurt to update the stats file, though, especially if you are going to 2nd pass encode at several different qualities.

After the third or fourth pass it can cause problems though, by fine-tuning too far and messing things up.