Catalina i.MX95 NPU

The GW9200 and other i.MX95 Catalina SBCs have the following built in NPU for machine learning (ML):

• eIQ® (Edge Intelligence) Neutron N3-1024S NPU (Neural Processing Unit)
  • Approximately 2 TOPs
  • Designed for high-efficiency and low latency inferences
  • Software stack from NXP (Use the NXP Yocto software image)
  • Benchmarks can be 2-3x of the i.MX8M Plus NPU, even at the same TOPs
  • Support for Modern AI/ML Workloads
    • CNN, MLP, RNN, LSTM, TCN, and more
  • Award Winning eIQ Development Environment
    • Enables TensorFlow, Pytorch, Caffe, ONNX, etc.

For more serious NPU / TPU calculations, consider a Mini-PCIe or M.2 accelerator that can be added to one of the slots on the GW9200.

Cards:

• ​https://trac.gateworks.com/wiki/hailoai
• ​https://trac.gateworks.com/wiki/TPU

NXP has a lot of great resources which should be used to learn more about the NPU and how to use it. See links below.

See Also

• Venice NPU Examples (very similar)
• ​https://www.nxp.com/design/design-center/training/TIP-DEEP-DIVE-INTO-IMX-95-NPU-AND-VISION-PIPELINE
• ​https://www.nxp.com/applications/technologies/ai-and-machine-learning/eiq-neutron-npu
• ​https://www.nxp.com/design/design-center/software/eiq-ai-development-environment:EIQ
• ​https://www.nxp.com/webapp/Download?colCode=TP-DEEP-DIVE-INTO-IMX-95-NPU-AND-VISION-PIPELINE