DXVA hardware accelerated decoding, AMD


After experimenting with this great toolkit for a while, we are aiming to use it for a proyect that uses video playback, and remote control.

My problem is how to hardware accelerate video decoding inside VVVV, as I manage to do it with MPC, VLC and other players. This is my current setup:

  • AMD embedded discrete GPU E8860 with UVD (Unified Video Decoder 4.0)
  • AMD R464L APU with integrated HD7000 graphics (disabled)
  • Win 7 PRO x64 up to date
  • Latest drivers of everything installed
  • VVVV x64 beta 33.3 with addonpack.
  • LAV filters 0.63, latest matroska splitter (Haali)
  • 30 FPS h264 videos in mp4 container (as AMD’s UVD only decodes h264 and vc-1) pLaying in 60 hz screens

My current GraphStudioNext graph is HAALI MEDIA SPLITTER --> LAV VIDEO DECODER --> VIDEO MIXING RENDERER… and DXVA works fine in other programs, as said before but not in VVVV. I installed K-lite codec pack, only using LAV for everything, disabled Windows Media foundation, configured LAV for N12 color output and DXVA in COPYBACK mode, as I understand that NATIVE mode needs direct rendering after decoding and it’s not applicable to VVVV.

The nodes I use for up to 2K resolution testing are FILESTREAM2 -> VIDEOTEXTURE (wait for 2th) -> QUAD (scaled to fullscreen, x & y scale to 2) -> EX9 RENDERER (all default).

We really need hardware acceleration… as you can see, the hardware is low power/industrial aimed (but pretty much highend in that market) and we are trying to get a VVVV based OEM product.

The question is: is there a better way to configure VVVV and windows filters to get the most out of this hardware?

I will also accept professional (VAT invoiced) collaboration to get this proyect working.