**• Hardware Platform (Jetson / GPU) Jetson orin nx
**• DeepStream Version: 6.2
**• JetPack Version (valid for Jetson only) 5.1.1
**• TensorRT Version8.2.2
Hello， I run my yolov5n model reference https://github.com/NVIDIA-AI-IOT/yolov5_gpu_optimization，
If I don’t use DLA, it works just fine， the fps can reach 40+(8batch_size)。
But when I use DLA, the fps is only reach 6.
Please refer to Matrix Multiply on DLA and checking DLA usage - Jetson & Embedded Systems / Jetson AGX Xavier - NVIDIA Developer Forums and Why run slower when use DLA and GPU together , even if the DLA model was transfromed all in DLA? - Jetson & Embedded Systems / Jetson Xavier NX - NVIDIA Developer Forums
Please use “trtexec” to test the model performance with and without DLA.
Jetson DLA does not support FP16 acceleration. And the INT8 performance is lower than GPU. We are developing a INT8 yolov5 DLA optimization sample, you may try the sample after we release the sample.
OK， Ths
Please notice me when you release new sample, thanks again
