• Hardware Platform (Jetson / GPU) Jetson Orin NX 16GB
• DeepStream Version Deepstream-6.2
• JetPack Version (valid for Jetson only) Jetpack 5.1
• TensorRT Version 5.1
• OpenCV without CUDA
I’m having trouble dissipating heat in my Nvidia Jetson Orin NX in my environment. If the Jetson starts to overheat I would like to reduce the power. Can I change the NVP model during operation to reduce the performance of the Jetson without rebooting?
I use the deepstream app with a camera as a source. There are long periods in which the deepstream app is running even though it is not needed. Once the Deepstream app is needed, I don’t have time to wait for the application to load (which can take half a minute). That’s why I leave it running permanently so that I can use it quickly.
My idea was to install a delay in the main loop so that the FPS is reduced and the GPU is not used so heavily when idle. As soon as I need the application, I deactivate the delay again. This should reduce the heat generated when idling. However, I don’t like the solution. Is there a way in the deepsteam app to temporarily switch off the inference processing of the neural network? Maybe by temporarily not calling a function so that the GPU is no longer used? The deepstream app would then run at the maximum FPS that the camera can provide. This should drastically reduce the heat output.