2019-08-22, 03:24
The player (madVR) is using the GPU shaders for rendering, so all VRAM usage is dictated by the GPU drivers. DXVA2 (copy-back) uses some system RAM, but you can't move the video processing from the GPU to the CPU.
You can move the video decoding to the CPU by choosing software decoding in LAV Video. You can also reduce VRAM usage by madVR by reducing the size of the present frames in advance number and the GPU queue size.
You can move the video decoding to the CPU by choosing software decoding in LAV Video. You can also reduce VRAM usage by madVR by reducing the size of the present frames in advance number and the GPU queue size.