Hello,
I am running tests on an HP E5 server with a Tesla M4 card, running Windows Server 2012 R2. I’ve noted that when I launch an NVENC session, I see a spike in the interrupts/sec rate (seen via Perfmon).
Specifically, it is the cuCtxCreate call which consistently pushes the rate to over 500k interrupts/sec. This momentarily stalls other threads running on the CPU cores which can affect the stability of our application.
I isolated this behavior by running the ‘NvEncoderCudaInterop’ sample and monitoring the Perfmon data while stepping through the initialization phase of the app.
Here is the command line I used:
NvEncoderCudaInterop.exe -i d:\media\HeavyHandIdiot.3sec.yuv -o out.h264 -size 1920 1080 -startf 0 -endf 0
I’d like to know if we can do anything to mitigate this behavior?
Thanks,
Mike