I have a Tegra X1 with compute capability 5, and to my knowledge, this means that system profiling is not supported when using nvprof! This is verified by running nvprof on my application with --print-gpu-trace --system-profiling on
How can I measure the power utilization of this device?