real-life XviD 2pass speed tests

[Koke]

I knew that something fishy was going on...
Being such a pain in the ass as I can be I`ve done
some tests regarding XviD VBR 2pass speed in previous Vidomi release.


448+Nic`sNew(28.06.2002.) 9,4-10,6 FPS
448+Umaniac`s(14.06.2002.) 10,4-11,6 FPS *****
448+Koepi`s(22.05.2002.) 9,2-10,1 FPS



Testings were done on a 52sec Matrix clip with Vidomi & XviD VBR 2pass.
All XviD installations were done after complete removal of previous
XviD installation (including removing XviD settings from registry).

All Vidomi options were the same, all XviD settings were the same,
see http://www.inet.hr/~dkokeza/xvid.
AVI came out 5Mb (7 - 2 for overflow)

All clips seemed same quality to me.
That`s all folks.

Koke

[Koepi]

Well, there have been some optimizations in the core since 22.05... that could explain the speed boost. Is the color-problem already solved, do I need to build a new binary? I did one for me with some b-frame stuff which doesn't work out well - it's _very_ buggy and can just be used in 1pass quantizer mode :(

So if I'd release a new binary it won't be with b-frame stuff...

Regards,
Koepi

[gldblade]

>Well, there have been some optimizations in the core since 22.05...

Yeah, but it apparently slowed down again after 6/14 in Nic's build. Could it be explained by the use of different compilers?