NVIDIA Jetson Xavier NX OC Alarm

Hi Sir,

I’m using an NVIDIA Jetson Xavier NX running AI software, but it shows the OC alarm (OC1).
My image is JP 4.6.0
Power adapter: 12V / 60W

Power Mode: 15W / 6 core
System log: syslog

Also, I have changed the “current_limit” from 5000mA to 5500mA.
But still no work.

But if I changed the Power Mode from 15 W/6 core to 10W / Desktop, this issue was gone.

Also, I realized that.
The NVP of the NX is 15W and 20W, but the VDD_IN instantaneous power is 25W?

Does this mean that for 15W or 20W its instantaneous power will be 25W, and this causes OC1 alarm?

Is the value of VDD_IN in the following picture normal?

Thank you for your help and reply.

Best Regards,
Neptune Liu

Hi neptuneliu,

Are you using the devkit or custom board for Xavier NX?

Could you also verify with the latest JP4.6.4(L4T R32.7.4)?

It might be. Please share the result of the following command on your board.

# grep "" /sys/class/hwmon/hwmon*/oc*

You may need to create custom nvpmodel configuration for your use case to prevent OC event.

Hi Sir,

1.Sorry, end users cannot upgrade to JP4.6.4 (L4T R32.7.4) due to AI software.
Why do you think JP4.6.4 or higher would be helpful?

~$ grep “” /sys/class/hwmon/hwmon*/oc*
/sys/class/hwmon/hwmon3/oc1_cpu_throttle_ctrl:0
/sys/class/hwmon/hwmon3/oc1_gpu_throttle_ctrl:1073741824
/sys/class/hwmon/hwmon3/oc1_irq_cnt:13
/sys/class/hwmon/hwmon3/oc1_priority:0
/sys/class/hwmon/hwmon3/oc2_cpu_throttle_ctrl:2147485455
/sys/class/hwmon/hwmon3/oc2_gpu_throttle_ctrl:3221684224
/sys/class/hwmon/hwmon3/oc2_irq_cnt:0
/sys/class/hwmon/hwmon3/oc2_priority:100
/sys/class/hwmon/hwmon3/oc3_cpu_throttle_ctrl:2147483919
/sys/class/hwmon/hwmon3/oc3_gpu_throttle_ctrl:3221291008
/sys/class/hwmon/hwmon3/oc3_irq_cnt:0
/sys/class/hwmon/hwmon3/oc3_priority:200
/sys/class/hwmon/hwmon3/oc4_cpu_throttle_ctrl:2147485455
/sys/class/hwmon/hwmon3/oc4_gpu_throttle_ctrl:3221684224
/sys/class/hwmon/hwmon3/oc4_irq_cnt:0
/sys/class/hwmon/hwmon3/oc4_priority:255
/sys/class/hwmon/hwmon3/oc5_cpu_throttle_ctrl:2147483663
/sys/class/hwmon/hwmon3/oc5_gpu_throttle_ctrl:3221291008
/sys/class/hwmon/hwmon3/oc5_irq_cnt:0
/sys/class/hwmon/hwmon3/oc5_priority:100
/sys/class/hwmon/hwmon3/oc6_cpu_throttle_ctrl:0
/sys/class/hwmon/hwmon3/oc6_gpu_throttle_ctrl:1073741824
/sys/class/hwmon/hwmon3/oc6_irq_cnt:0
/sys/class/hwmon/hwmon3/oc6_priority:0

Hi Kevin,

In addition, I am curious why in other NX systems with the same configuration, some QC Alarm times are 0, some are only hundreds, and some are tens of thousands? (Under the same test time)

Thank you for your reply and suggestion.

Best Regards,
Neptune Liu

Hi Kevin,

Do you have any updates or suggestions for this issue?

Thank you for your help.

Best Regards,
Neptune Liu

We would fix the known issue in the latest release and we also always verified and check the issue on this release locally.

It seems you hit OC1 in your case, which means the instantaneous power is higher than 25W.
You can find more details in Software-Based Power Consumption Modeling.

Please create custom nvpmodel configuration to meet your requirement(in your use case) and prevent the OC throttling occurs.

Hi Kevin,

Why do some systems (with the same configuration) not have this problem, while others do?

Can you provide an explanation for this question?

Thank you for your help.

Best Regards,
Neptune Liu

Hi Kevin,

May I change the “current_limit” from 5000mA to 5500mA?
According to our hardware design, we can support up to 12A, but we don’t know how about the SOM is, is there a risk?

Thank you for your support.

Best Regards,
Neptune Liu

What do you mean about “some systems”? Do you mean different platform or release or module?
It may depend on the use case and the loading (through Jetson Power GUI or tegrastats).

We don’t suggest modifying the current_limit directly.
We would suggest creating the custom nvpmodel configuration for your use case.

Hi Kevin,

These systems have the same SOM, carrier board, and image, but some systems have this issue and some do not.

Thank you for your help.

Best Regards,
Neptune Liu

You can check the power log for detailed power usage on different board with the same SOM+Board+Image.
Maybe the failed board just near to the threshold of the OC event.

Hi Kevin,

How to check the power log?
Could you provide the command?

Thank you.

Best Regards,
Neptune

You can dump it from Jetson Power GUI.
Or just run sudo tegrastats to show the power usage.

Hi Kevin,

The problem we are having is that the instantaneous currents are too high for jtop or tegrastats to see.

Best Regards,
Neptune Liu

Some instantaneous current cannot be monitored through jtop or tegrastats since it occurs too short. You can check the OC count and also analyze your power usage through power log and create the custom nvpmodel configuration according to your use case to prevent the instantaneous high current issue.

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