We are currently evaluatinng our power consumption while using differents components of our software stack. We have noted that no matters if we launch our deep learning models or not the power consumption is almost the same.
Currently we are using Jetson Xavier NX on 25W mode by default. and we saw that it takes almost 25W from the Linux boot.
Is this the expcted behaviour?
Is it possible to make deactivate dynamically the GPUs when we don’t use our deep learning models to reduce the power consumption?