Hi everyone,
I’m working with the Orin NX 16GB and running YOLOv8 for object detection tasks. Initially, after flashing a fresh system, I ran some benchmarks and got the somewhat the expected performance. However, after installing Redis and RabbitMQ, I noticed a significant drop in performance. I conducted another round of benchmarks after downloading these services, and the performance had worsened drastically.
Here’s what I’ve tried so far:
- Disabled both Redis and RabbitMQ, but this hasn’t improved the performance.
- Verified that no other intensive processes are running.
- Fresh installation of Jetson software
I have followed the following tutorials:
NVIDIA Jetson - Ultralytics YOLO Docs
Deploy YOLOv8 with TensorRT | Seeed Studio Wiki
Has anyone else experienced similar issues with Redis/RabbitMQ or other services impacting the Orin NX performance? Any suggestions for troubleshooting or fixing this problem would be greatly appreciated!
Thanks in advance for your help.
Best regards,
Koen
The data:
| Format | Status | Size (MB) | mAP50-95(B) | Inference Time (ms/im) | FPS |
|---|---|---|---|---|---|
| PyTorch | ✅ | 6.2 | 0.6381 | 107.54 | 9.30 |
| TorchScript | ✅ | 12.4 | 0.6117 | 14.26 | 70.10 |
| ONNX | ✅ | 12.2 | 0.6117 | 21.74 | 45.99 |
| OpenVINO | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorRT | ✅ | 16.1 | 0.6092 | 7.99 | 125.15 |
| CoreML | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow SavedModel | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow GraphDef | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow Lite | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow Edge TPU | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow.js | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| PaddlePaddle | ✅ | 24.4 | 0.6092 | 611.20 | 1.64 |
| NCNN | ✅ | 12.2 | 0.6034 | 47.95 | 20.85 |
and after installation of redis and rabbitmq
| Format | Status | Size (MB) | mAP50-95(B) | Inference Time (ms/im) | FPS |
|---|---|---|---|---|---|
| PyTorch | ✅ | 6.2 | 0.6381 | 106.69 | 9.37 |
| TorchScript | ✅ | 12.4 | 0.6117 | 14.24 | 70.21 |
| ONNX | ✅ | 12.2 | 0.6117 | 24.29 | 41.18 |
| OpenVINO | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorRT | ✅ | 15.8 | 0.6091 | 14.21 | 70.38 |
| CoreML | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow SavedModel | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow GraphDef | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow Lite | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow Edge TPU | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| TensorFlow.js | ❌ | 0.0 | 0.0000 | 0.00 | 0.00 |
| PaddlePaddle | ✅ | 24.4 | 0.6092 | 595.86 | 1.68 |
| NCNN | ✅ | 12.2 | 0.6034 | 35.49 | 28.18 |