About OrinNX Performance (GPU)

Robotics & Edge Computing Jetson & Embedded Systems Jetson Orin NX
,
1

Hi,
I am looking at the following Jetson Orin NX Series Power and Performance document.
https://docs.nvidia.com/jetson/archives/r35.3.1/DeveloperGuide/text/SD/PlatformPowerAndPerformance/JetsonOrinNanoSeriesJetsonOrinNxSeriesAndJetsonAgxOrinSeries.html#supported-modes-and-power-efficiency.

In the Jetson Orin NX 16GB,

  • At 15W, GPU maximal frequency (MHz) is 612
  • At 25W, the GPU maximal frequency (MHz) is 408
    for 25W, the GPU maximal frequency (MHz) is 408.

Why is the GPU value lower at 25W than at 15W?
Is there a reason why the GPU value should not be higher than the 15W GPU?

2

Hello,

Thanks for visiting the NVIDIA Developer forums! Your topic will be best served in the Jetson category.

I will move this post over for visibility.

Cheers,
Tom

4

Thanks for moving the category.
Looking forward to your answers.

6

Hi,

In 25W, 4 Texture Processing Clusters (TPC) are enabled. One TPC has 256 CUDA cores.
But in 15W, only two TPCs are enabled.

So the frequency of the 25W mode is reduced to meet the limited power budget.
Thanks.

7

Thanks for the answer.

I am not familiar with the details and do not understand.
In the case of 25W, it uses 4 “GPU TPCs”.
15W uses two “GPU TPCs”.
In the case of 15W, even at 612MHz, two TPCs are used to meet the power limit,
In the case of 25W, if I use 612MHz, four TPCs will not meet the power limit, so I have to lower the frequency to 408MHz?

8

Hi,

Yes, more TPCs indicate more CUDA cores which means more computational resources (you can see it as threads on GPU).
Since more computation resources are available, the power consumption will increase, and the clock rate needs to be reduced to meet the 25W budget.

Thanks.

9

Hi,

Thank you for the clear explanation.
I understand now.

11

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