How to use the PVA DLA and the other co-processor

[legendflying]

hello, i do power performance test, using the yolov8 model(the video is a USB UVC camera), now the model can run and i can get the results. But the GPU take the main heavy task, and the DLA PVA are off, they are not working.
my env:

the GPU loading chart:

the jetson stats:

the DLA and PVA is OFF

my Devkit is Xavier NX, but I do the same test in AGX xavier devkit，the result is the same.

my question is how to use the DLAx PVAx to accelerate the performance? and let the DLA/PVA do more, offload the GPU task?
now we use one usb camera, if we use more cameras, the GPU’s ability is not enough.
thanks

[legendflying]

Among the computing power provided by NVIDIA, DLA PVA provides the majority of the computing power.
take the spec of orin nx as example:

[AastaLLL]

Hi,

PVA is designed for vision tasks and you can deploy jobs on it with VPI:
https://docs.nvidia.com/vpi/architecture.html#autotoc_md11

DLA can be used via TensorRT. For more information, please check below doc:

Developer Guide :: NVIDIA Deep Learning TensorRT Documentation

Thanks.

[legendflying]

copy that and thanks
use the correct parameter to enable the DLA function.
./trtexec --onnx=data/resnet50/ResNet50.onnx --useDLACore=0 --fp16 --allowGPUFallback
about PVA, we will go on learning it and the VPI.