So, I cover in this 36 minute video what @eposvox missed in his 5 hour course on OBS.
I go into how to do this on Windows and Linux.
Unfortunately, Looking Glass did not make it into this video, as I’m waiting on “Beta 1” to make a video on Looking Glass once it’s more feature complete and backwards compatible.
Edit:
So it has come to my attention there’s confusion to dedicating a NVENC encode to a specific GPU.
On Windows, this is not all that this process is doing. By launching OBS on a secondary GPU, it’s canvas is also rendered on the secondary GPU. The canvas composites the game capture with other elements like browser sources, your webcam, and text elements. BOTH the canvas and NVENC are dedicated to a secondary GPU with this method.
The Linux GPU passthrough method also does the same thing, where the host GPU is the only one dealing with the OBS canvas and NVENC, and the capture card captures the output of the GPU being used in the GPU passthrough VM.
Hope this clears up any confusion. It’s not just NVENC people!