About "System throttled due to over-current"

When I run my app on Orin Nano with custom board, the system throws “System throttled due to over-current”.
So I’m trying to find out is the problem with the app or with the custom board.
I followe this link to stress test on custom board and Orin Nano devkit respectively.
After I run matrixMulCUBLAS, both custom board and devkit throws “System throttled due to over-current”.
I use the following command to check OC events

orinnano@ubuntu:/usr/local/cuda-11.4/samples/0_Simple/matrixMulCUBLAS$ grep “” /sys/class/hwmon/hwmon*/oc*_event_cnt

Is this “Average Overcurrent”?
Is it normal for running Nvidia’s recommended GPU stress test in devkit?
Are there any negative effects of the device being in the above state for long time?
I guess the GPU is overloaded, if there are negative effects, how can I avoid them?
By the way, jetpack is both 5.1.3, and the devkit is flashed via SDK Manager.

Which Orin nano is in use? 8GB or 4GB? Is the oc even on both devkit + custom board or only custom board?

Orin Nano 8GB

both devkit + custom board

These are from devkit

I set “Warn Current” to 3984mA with the following command

echo 3984 > /sys/bus/i2c/drivers/ina3221/1-0040/hwmon/hwmon4/curr1_max

The warnings are gone. Is that okay? I mean I pull up the current limit wouldn’t damage the device?

It will damage the board… please do not do that…

please refer to document for how this mechanism works here.


In brief, Orin Nano and Orin NX has its own power limit that some high loading workload may trigger throttling on it. It is normal behavior and expected.

You mean there’s no need to worry about this warning?
Well, how do I disable this warning message.

Also, I’m confused about running matrixMulCUBLAS on devkit causing “System throttled due to over-current”.
Testing GPU stress with it will keep the device in “over-current”. This is to test if the device can hold a long term “over-current”?

Hi @EEEXcalibur

May I ask a question about what is your stress test target here? Why do you run this stress test?
For thermal design? or other purpose?

Sure, @WayneWWW
The environment in which the device runs might be accompanied with hours of high temperatures, and our application will keep the CPU and GPU close to full usage. It is essential to perform some stress tests.


Double checked with our internal team, the clock spec is designed for common usecase and not for stress test.
If you want to test stress, then it will easily hit oc throttling. And cannot be prevented.

Thanks, I get it.
What I want to know is if the module is kept in OC throttled state for a long time will it be damaged, such as decreasing the life of the module, etc,.
If yes, this will prompt us to upgrade to higher performance jetson modules to support our application.
If not, we can just disable the “OC throttled” warning on the desktop.

It is basically a question that what you want to do here?

Throttling will make the application into lower performance and it is a mechanism to protect board. It won’t hurt the board but the performance of your application may not meet your expectation.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.