Low performance of AGX Drive Xavier hardware

Please provide the following info (check/uncheck the boxes after creating this topic):
Software Version
DRIVE OS Linux 5.2.6
DRIVE OS Linux 5.2.6 and DriveWorks 4.0
DRIVE OS Linux 5.2.0
DRIVE OS Linux 5.2.0 and DriveWorks 3.5
NVIDIA DRIVE™ Software 10.0 (Linux)
NVIDIA DRIVE™ Software 9.0 (Linux)
other DRIVE OS version

Target Operating System

Hardware Platform
NVIDIA DRIVE™ AGX Xavier DevKit (E3550)
NVIDIA DRIVE™ AGX Pegasus DevKit (E3550)

SDK Manager Version

Host Machine Version
native Ubuntu 18.04
Hi ,
I am using AGX Drive XAVIER platform for our autonomous project .
Currently , i am running Lidar , Radar and single camera on the target along with RVIZ .
I found performance reduced to less than 2 Fps when i run all 4 module together .
not only object detection performance but overall system runs very slow .
There is a sufficient RAM space is available on target around 18 GB and CPU also under utilized for the same .

I was trying to check PowerMizer setting and found it shows only “AC” .

Why it runs very slow despite of having enough free resource bandwidth ?

It sounds like you’re experiencing a performance bottleneck on your AGX Xavier platform when running multiple modules simultaneously. While it’s good to have enough RAM and CPU resources available, it’s also important to consider the available GPU resources for each module.

Have you checked the GPU utilization when running all four modules together? You can use the Nsight Systems tool. Nsight Systems provides a graphical interface that shows GPU utilization, memory usage, and other performance metrics.

Can you please clarify what did you expect to see when checking PowerMizer setting? Also, did you check it on the devkit? If so, how did you check it?

Trying to setup nsight tool as well. current softwares are not Compute intensive therefore i would like to know is there any way to increase performance of system . i mean like nvpmodel type of command on system.

DRIVE OS doesn’t support different power modes. In the meantime, you can monitor the resource utilization using tegrastats as you set up nsight systems.

Thanks for response
Below are the Tegrastats log
RAM 5740/27924MB (lfb 4834x4MB) CPU [95%,96%,98%,1%,0%,100%] GR3D_FREQ 63% AUX@47C CPU@49C Tdiode@49C AO@49C GPU@48.5C tj@58C
RAM 5740/27924MB (lfb 4834x4MB) CPU [91%,87%,99%,2%,0%,100%] GR3D_FREQ 12% AUX@47C CPU@49C Tdiode@49C AO@49C GPU@48C tj@57.5C
RAM 5743/27924MB (lfb 4834x4MB) CPU [95%,90%,100%,0%,1%,99%] GR3D_FREQ 2% AUX@47C CPU@48.5C Tdiode@49C AO@49C GPU@48C tj@57.5C
RAM 5742/27924MB (lfb 4834x4MB) CPU [94%,93%,100%,0%,0%,100%] GR3D_FREQ 43% AUX@47C CPU@49C Tdiode@48.75C AO@49C GPU@48C tj@57.5C

I have a few doubts regarding core uses. We have an octa-core processor, but in the logs it is only showing 6 cores. Another observation is that it always uses Core 1, 2, 3, and 6.

There are several existing forum topics that discuss this, e.g. Only 4 cores out of 6 are available on Pegasus. I recommend searching the forum using relevant keywords to find the information you are looking for. This will allow you to access detailed discussions and insights shared by the community.

I have gone through the link but am not getting a clear answer to the underutilization of the system. Can you share any official document link for the same?

Please refer to DRIVE OS Linux. Thanks.