A story with two images. On the left, my background service is encoding 1 x stream at 60 fps, then another run with 2 x streams at 60 fps, Windows 11 Hardware-accelerated GPU scheduling is on (a 1650 graphics card). On the right a 1 x stream encode, then another run with a 2 x streams, scheduling is off.
The former gives me huge latency spikes with 2 streams, 100 sometimes 200 or more ms on some frames (isolated it to nvEncEncodePicture) causing regular serious glitches in playback. GPU spikes around. On the right, no spikes at all.
Can someone explain this behaviour? I’d like to understand it before I issue advice about it. I’ve found the reg key to detect hardware scheduling, so I’m able to provide the setting in diagnostics when we get the inevitable customer calls.
(apologies for incorrect title in initial post!)