Using VPU's efficiently

Hi There,

I need some advise on using VPU’s for encoding.
I have a card with 3 dedicated VPU’s.
I want to create 4 channels that get’s input and encode H264 with NVENC (using the encode with CUDA flavour)

What is better:

  1. To create 4 processes each one of them handling input and passing data to NVENC…
  2. To create 1 process with 4 threads each thread handling input and passing data to NVENC…

What is the most performant way ?

Thanks,
0L

You might want to ask this in the video codec sdk forum:
https://forums.developer.nvidia.com/c/professional-graphics-and-rendering/video-technologies/video-codec-and-optical-flow-sdk/189
I guess there are more people that can answer your question than in this forum.

moved
thanks