Gpu performance down

Hi!

In jp45/tensorflow2.3.1, I ran Mobilenet V1 (with 6.0G swap memory
and same 5V 4a power) several times.

About 5000 images took 140 seconds at first
but after I execute jetson_clocks and some other models bigger than mobilenetv1, inference time become 900 ~ 1100 seconds

is there any way that I can check the gpu status or recover it???

I tried other board and other sd card !
Card A(in board C) is properly working and card B(in board D) is not
When I put A in board D , it works fine
but card A also went wrong…

I will try sdk manager…

Hi,

You can check GPU utilization with tegrastats.

$ sudo tegrastats

If you want to reset the power mode back to default, please run the following command or just reboot the system.

$ sudo nvpmodel -m 2

Thanks.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.