How to check performance of TensorRT/ONNX model running on Jetson NX

My question is about the performance in terms of TOPs/Power consumption/GPU-CPU usage etc.
How can we check the performance of TensorRT(FP16)/ONNX models in terms of its GPU usage,TOPs,Power consumption?

Documentation says - 21TOPS of INT8 model on NX, How can I get like these numbers for my models


You can find the GPU usage with tegrastats.

$ sudo tegrastats

To get the maximal performance, you can check our benchmark sample below:


For other metrics,like power consumption and TOPs etc , How to check ?

Please check NVIDIA Jetson Linux Developer Guide : Clock Frequency and Power Management | NVIDIA Docs