ffdshow's DXVA video decoder problem

Thanks, I will test it. ;)

Tested ; seems to work but the CPU works as much as with FFVideoSource() while it should use less CPU if GPU is used (I chose well DXVA-Copyback in LAV's options).:confused: Actually the CPU works even a little bit more with DirectShowSource() and LAV (1 or 2 %) than with FFVideoSource(). I also verified with AVSMeter, same result. Everything was disabled in ffdshow and the command DirectShowSource() didn't work before to install LAV, so LAV was clearly used thus I don't understand this result.:confused:

I just tried with Nvidia 9500GT and CPU usage is 0-1% with DXVA Copyback. Are you sure that LAV Video decoder is used and that it uses UVD chip? Download and start GPU-Z to check GPU and UVD usage. Also, open LAV decoder during test and see if it uses DXVA-Copyback or fails to avcodec.

What OS do you have? LAV Video needs Windows 7/8 or at least Windows Vista. It only supports DXVA2. If you open LAV Video decoder during script playback you can see what mode is used by LAV video. If you see dxva2cb that DXVA-Copyback is used. If you see avcodec that LAV Video decoder falls back to software decoding.

AFAIK, nev does not plan to add support for DXVA1 decoding as official support for Windows XP ends in spring 2014. So, your only options for hardware accelerated video decoding while using Avisynth are: - buy Nvidia card and use CoreAVC with CUDA or - upgrade OS to Windows 7 or Windows 8 and use LAV Video decoder in Copyback mod.