Why Xavier NX performance increase by changing power adapter

[jaylun.lee]

Hello dears,

In general, we use a 60w power adapter for the Xavier NX to test YoLov5 by
20w 6 cores and the performance is 28FPS
10w 4 cores and the performance is 25FPS

we only replace the power adapter to 120w, we see the performance increasing double to 50FPS with similar power consumption (17w)

Although both tests are going to be system throttling after few minutes, how to explain Xavier NX seems to extract out of spec power to improve performance?
Is that real power consumption in jtop what I see?
What is the real maximum power consumption of Xavier NX if the thermal and over-current issue can be managed?
Thanks for your reply.

[Trumany]

Hi, you can get the main module power data by command line “tegrastats”. The power value is calculated by the monitors in module. I don’t think changing power adapter will cause performance change if it has not been throttled before. You might need to check the real output capability of the power adapter, also need to find out if any throttling exists for thermal or power supply capability during your tests.