I have run my application inside nivdia docker quite well on a 418.126 driver/ cuda 10.1 server(with T4 card). But my application runs extremely slow (about 5 seconds per frame) on another server with same hardware(also T4 GPU), the only difference is the driver version on that machine is 440.64 and cuda 10.2.
And I use command
nvidia-smi dmon -s u
to monitor the GPU usage when I run the application and I have found that it always be 0% for both nvenc and nvdec. So may someone tells me what’s wrong about that?
The Video Codec SDK version I use is 9.0.20