16GB Zed box orin nx vs 64GB AGX Orin developer kit

Hardware ZED Box
, ,
1

WHY IS MY ZED BOX RUNNING 3 GMSL2 CAMERAS FASTER THAN MY 64 GB AGX ORIN???

Through multiple testing instances and a large data base, I have found that my 16 GB zed box constantly runs at higher frames than my 64 GB AGX. I have no idea why this is, I am using a zed box deserializer for the AGX to read my gmsl2 cameras and its still slower.

If anyone has any ideas, or thoughts please lemme know.

2

Please add more information. With no data, it’s not possible to reply to your question.

3

e.g.
I will run 3 zed x cameras via gmsl on my 16 GB Zed Box orin nx, I will get more frames than if I run the same cameras on my 64GB AGX Orin.

I don’t know what more data I can provide, I run my code on python and have based my code on Zeds SDK and have used it quite a bit. I also have depth and custom object detection enabled(my AI model uses YOLOV12 onnx), I have resolution SVGA.

Could you be more specific on what data you would like to know that could direct your answer to our question?

4
  • FPS on AGX Orin
  • FPS on Orin NX
  • Power mode on AGX Orin
  • Power mode on Orin NX
  • YOLO model
  • is the YOLO model optimized for the GPU?
  • Are you using ONNX integration?
  • Other useful information that you can provide…
5

FPS AGX: 13
FPS ORIN NX: 17
Power on both: MAXN
YOLO model: custom YOLOLIKE box objects from onnx file exported with image size = 608 and offset = 12

6

The YOLO model is not optimized for the Jetson model that you are using and it is slowing down the machine.
I recommend you follow the instructions in our documentation to convert it into ONNX format and then load it in the ZED SDK pipeline to optimize the processing.