'System Throttled due to overcurrent' while running TensorRT

Jetson Orin Nano 8GB dev kit, in 15W mode, when I run tensorrt in multiprocessing mode(python), I get an overcurrent message, although Jetson consumes 10 watts, how I can fix it???

UPD1: I noticed that this message appears when Jetson is loaded with a neural network and another process is capturing through opencv and gstreamer and outputting the image. As soon as I don’t turn on the camera everything is ok. But I need a camera, so the question is the same - how to fix this?

UPD2: without gstreamer but with other processes I get the same thing;(

1 Like

It is a misleading message, and isn’t actually an error. Jetsons have different power modes since people sometimes have limited battery life. The max power mode has its limit set due to heat generation, and if the power would cause overheat, it reduces power and says throttled due to overcurrent. There are many other modes, but throttling is to keep the consumption below some power model limit (to conserve batteries) and isn’t due to heat. They all default to the same message. Even when the message is for heat, it is only telling you that the clocks are being throttled back to maintain power limit at that model.

Temperature stays below 60 - so it’s not overheating

power is through 10A lab psu. 19v, current consumption is 0.6a at max so this must not be the cause.

so in a nutshell, we have enough power and the temperature is acceptable, the cpu load is relatively low at all times. if you say this is not an error, why does it start throttling

can I edit the power mode to lift these limitations?

I only have 7 and 15w

@linuxdev up)

Hi,

Usually throttling is triggered by heavy loading,
It will reduce the clock rate to prevent the overcurrent.

https://docs.nvidia.com/jetson/archives/r35.3.1/DeveloperGuide/text/SD/PlatformPowerAndPerformance/JetsonOrinNanoSeriesJetsonOrinNxSeriesAndJetsonAgxOrinSeries.html#jetson-orin-nx-series-and-jetson-orin-nano-series

You can customize a nvpmodel based on your use case.
Thanks.

@AastaLLL
is this considered a heavy loading?

@AastaLLL I use the 15 watt mode, the temperature is 50 degrees, the cooler is up to 50%, the load is mainly on 4 processor cores and the load on the graphics processor, why does throttling occur in this case if we have not even reached 15 watts or the critical temperature? This regulation affects the system’s FPS over time. But damn, we didn’t exceed the system throttling parameters

@AastaLLL All your advice “you just need to reduce the load” is simply ridiculous. My jetson consumes a maximum of 11 watts, not 15 watts, and at 11 watts it still throws out over-current

Keep in mind that the throttling is really relative to a power mode. If you are really in a 15W power mode and it is drawing 10 A (unless you are saying the power supply is capable of 10 A, but it sounded like the Jetson actually draws 10 A), then something else might be wrong. Perhaps the power mode itself has to be examined.

Hi,

Based on your screenshot, the total power is 9.8+4.1+2.5~16.4W.
The total power is already over the 15W limit.

The suggestion is to lower the clock to reach the 15W target rather than reducing loading.
nvpmodel is a tool to control the clock rather than system loading.

Thanks.

@AastaLLL based on this screenshot, vdd_in is the total module power so why should it be combined with 2 others to calculate total power? Aren’t vdd_cpu_gpu_cv and vdd_soc generated from vdd_in? thanks!

1 Like

Hi,

Sorry, you are right. The VDD_IN is the total power for Nano.
Just mixed with other devices.

There are three types of OC events.
Could you help to check which event triggers the OC on your device for us?

https://docs.nvidia.com/jetson/archives/r35.4.1/DeveloperGuide/text/SD/PlatformPowerAndPerformance/JetsonOrinNanoSeriesJetsonOrinNxSeriesAndJetsonAgxOrinSeries.html#overcurrent-event-status

Thanks.

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