As part of my last post, I was exploring video transcoding with an Intel Arc GPU and how different levels of ICQ affect video size and quality. During which, I noticed there’s a QSV compatible filter in ffmpeg called vpp_qsv which has a detail parameter. VPP stands for Video Post Processing and the filter help text simply says “enhancement level” for detail. The potential to do some automatic picture enhancements directly on the GPU piqued my interest so I decided to have a play with it.
Transcoding with an Intel Arc GPU
I am a bit of an NVIDIA fanboy when it comes to GPUs. This is partly down to being a long time Linux Desktop user (until recently) and them having excellent drivers allowing me to pretty effortlessly get proper hardware acceleration working.
In my lab environment I have an old Nvidia Quadro P400 that’s set up for various transcoding workloads and its been working absolutely great for me despite being an older card. I have been primarily using the h264 codec but I recently moved over to h265 (hevc) mostly out of curiosity, and to see if I could get quality on-par with h264, whilst getting similar file sizes.