Skip to main content
Datasheets.com
Application NoteNxp

i.MX RT700 eIQ Neutron NPU Enablement and Performance Application Note

Technical guide on using the eIQ Neutron N3-64 NPU with the NXP i.MX RT700 MCU for high-performance, low-power machine learning inference and neural network acceleration.

View application note

Overview

This application note details the implementation and performance of the eIQ Neutron N3-64 Neural Processing Unit (NPU) integrated into the NXP i.MX RT700 microcontroller family. It explains the software enablement path using the eIQ toolkit and LiteRT for microcontrollers (formerly TensorFlow Lite). The document provides benchmarking data demonstrating up to 169x performance gains and 83x energy efficiency improvements for neural network models compared to standard Cortex-M33 execution. Key technical sections cover the 41.6 GOPS architecture, power management using the PCA9422UK PMIC, memory allocation for the tensor arena within the 5.5MB NPU-accessible SRAM, and vision application support using the OV7670 camera sensor.

Use Cases

  • Edge machine learning inference
  • Keyword spotting (KWS)
  • Visual Wake Words (VWW)
  • Anomaly detection
  • Image classification
  • Low-power AI applications

Topics

eIQ Neutron
N3-64
i.MX RT700
Neural Processing Unit
NPU
TensorFlow Lite
LiteRT
Machine Learning
Edge AI
PCA9422UK
OV7670
MLPerf Tiny

Referenced Parts

PCA9422UK

NXP Semiconductors

This voltage must be supplied by an external PMIC, in this case the PCA9422UK found on the MIMXRT700 EVK

i.MX RT700

NXP Semiconductors

The eIQ Neutron N3-64 NPU is integrated into the i.MX RT700 microcontroller family.

OV7670

OmniVision Technologies

support vision models, there will be examples using both a parallel camera with the OV7670 sensor