https://bugs.kde.org/show_bug.cgi?id=430193
--- Comment #3 from Jason Straight <j.straight-...@straights.net> --- (In reply to Vincent PINON from comment #2) > There is a misunderstanding: real_time is for *processing* threads (MLT > calculating frames/slices in parallel) while encodethreads is for *encoding* > (typically libx264 or libvpx compressing the calculated video). > > Increasing encodethreads works, but beyond 4 it doesn't speed up much. > But MLT parallel processing doesn't always work well (with several frei0r > effects/transitions). > > It is difficult to find a balance filling the cores for render, as either > processing or encoding can be bottleneck depending on video > characteristics... > > For previews, the codec used doesn't use frame-to-frame compression (MJPEG, > DNxHD...), so encoding almost does not come into equation. Yeah. I should have elaborated. I'm using GPU for encoding. Nvenc. Increasing the processing threads from 4 to 12 really sped up my render times for final rendering and timeline preview. -- You are receiving this mail because: You are watching all bug changes.
