View Single Post
Old 12th November 2009, 22:25   #214  |  Link
turbojet
Registered User
 
Join Date: May 2008
Posts: 1,840
If you look through the benchmarks in this thread you'll see 1 and 2 MB buffer are more often faster than 4, 0 and 8 are some times faster as well, although about 5% max gain. I think it should an option, after all what use is a benchmark to find the fastest encode if you can't all the benchmark settings?

A very significant speed up is using x86 with first pass and veryfast, ultrafast presets, you can gain 20% fps in these situations and in my tests 2 pass is always significantly slower with x64 then it is x86 because you lose 10-15% fps on the first pass.

I've coded both of these in and reduced min x264 version to 1300 because 1332 is much slower, weightp breaks coreavc and is forced unless mbtree, psy-rd is disabled or interlaced is enabled. Also as it's a new feature without any public testing bugs can be expected. The only thing is I get a statement was expected but found a procedure error when compiling and I can't figure out what is causing it. Source here if anyone wants to look at it.

EDIT: also couldn't figure out how to set buffer size with a combobox.

Last edited by turbojet; 12th November 2009 at 22:43.
turbojet is offline   Reply With Quote