Microcontroller With Built-In AI Processing Unit

The microcontroller series adds ML to small, power-efficient devices, allowing AI features like computer vision, audio processing, & real-time edge AI applications.

STMicroelectronics has launched a new microcontroller (MCU) series that adds embedded artificial intelligence (AI) capabilities to cost-sensitive and power-efficient devices. The STM32N6 series integrates machine learning (ML) features, enabling applications like computer vision, audio processing, and sound analysis in embedded systems.

The series features the Neural-ART Accelerator, a neural processing unit (NPU) that provides up to 600 times the ML performance of existing STM32 MCUs. This allows consumer and industrial devices to use previously impossible AI features in small embedded systems.

– Advertisement –

The microcontroller series is designed for a wide range of users, including consumer electronics developers, industrial IoT engineers, and embedded systems designers who need compact, power-efficient AI solutions for applications like computer vision, audio processing, and sound analysis. AI and machine learning developers can leverage its compatibility with frameworks, while manufacturers of vision-related products can use its integrated Image Signal Processor (ISP) for image sensor processing. Prototypers and researchers developing edge AI applications will benefit from its support for real-time, local model execution.

The company claims the series provides up to 600x machine-learning performance compared to high-end STM32 MCUs. This is achieved through ST’s Neural-ART Accelerator, which includes nearly 300 configurable multiply-accumulate units, delivering up to 600GOPS.

The MCU features an 800 MHz Arm Cortex-M55 core, achieving a CoreMark score of 3,360. For applications that do not need AI algorithms, the series is also available in a version without the Neural-ART Accelerator while still offering the performance, interfaces, and functionality of the MCU.

With 4.2 MB of RAM, the MCU supports data-intensive AI and multimedia workloads. Two 64-bit AXI interfaces provide bandwidth to sustain high computation rates and fully leverage the power of the Neural-ART Accelerator. The MCU is supported by ST’s Edge AI Suite, offering tools for developing edge machine-learning applications. The suite lets users bring their models in formats like TensorFlow Lite, Keras, and ONNX.