Cuda cores and ARM processors

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) AGX Xavier/Orin
• DeepStream Version DS 6.1.1
• JetPack Version (valid for Jetson only) JP 5.0.2
• TensorRT Version 11.4

Is there a way in Deepstream to choose what hardware resources the Deepstream pipeline uses between CUDA cores and the CPU ARM processor?

Thank you in advance.

If you use deepstream then all plugins use CUDA by default.

You can switch some plugins to the CPU version for example the encoders/decoders can be switch to the CPU alternatives. You can find those on the Gstreamer website.
And I don’t think any inference plugins (nvinfer, nvtracker, etc) can be used on CPU. You can use DLA cores for nvinfer if needed.

1 Like

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one.
Thanks

You can refer our guide to see the paras of each plugin. For example: nvdsosd, you can set the process-mode para to control the resources.
https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_plugin_gst-nvdsosd.html?highlight=nvdsosd

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.