STMicroelectronics has rolled out its latest series of microcontrollers designed to bring advanced artificial intelligence processing directly to edge devices like robots. This new series, its most powerful to date, integrates a custom neural processing unit (NPU) that dramatically increases machine-learning capabilities at the edge.
According to Remi El-Ouazzane, president of the Microcontrollers, Digital ICs and RF Products Group at STMicroelectronics, the era of edge AI is dawning. He explained that while current AI models perform tasks such as segmentation, classification, and recognition, they will soon be harnessed for a wider range of future applications.
Headquartered in Switzerland and employing over 50,000 people worldwide, STMicroelectronics serves more than 200,000 customers and collaborates with thousands of partners to develop innovative semiconductor solutions. The new microcontroller, part of the STM32N6 series, is the first to feature the company’s proprietary Neural-ART accelerator. This dedicated accelerator is engineered for power-efficient AI tasks, claiming to deliver 600 times more machine-learning performance than previous high-end STM32 microcontrollers.
Key highlights of the STM32N6 include:
- Advanced Processing Core: It utilizes an 800 MHz Arm Cortex-M55 processor paired with 4.2 MB of RAM to handle data-intensive applications.
- Neural-ART Accelerator: Operating at 1 GHz and capable of up to 600 GOPS (giga operations per second), it enables real-time neural network inference for computer vision and audio processing.
- Integrated Image Signal Processor (ISP): For the first time in an MCU, an ISP is built-in to support direct image processing. This feature allows the use of affordable image sensors and is customizable via free software to adjust parameters like exposure, contrast, and color balance.
- Enhanced Connectivity and Graphics: A dedicated computer vision pipeline with a MIPI CSI-2 interface ensures broad camera compatibility, while an H264 hardware encoder and a graphics accelerator (NeoChrom) support feature-rich applications.
- Robust Packaging: Available in six package options, the MCU is built to operate in harsh environments with ambient temperatures up to 125°C.
El-Ouazzane emphasized that the introduction of the STM32N6 marks just the beginning of a broader initiative to integrate hardware-accelerated AI into future STM32 products, promising innovations that were previously out of reach with standard embedded processing.
The MCU is further supported by STMicroelectronics’ Edge AI Suite, a collection of software tools that facilitate the development of edge machine-learning applications. Developers can import models in formats like TensorFlow Lite, Keras, and ONNX, and take advantage of an expanding repository of pre-trained AI models. Additionally, the STM32N6 is designed with advanced security in mind, meeting rigorous standards with SESIP Level 3 and PSA Level 3 certifications.
Already available to select customers since October 2023 and now offered in high volumes, the STM32N6 is finding its way into products from major companies such as LG Electronics and research initiatives at Lenovo. Early feedback highlights its exceptional neural processing performance, power efficiency, and ease of integration—paving the way for a new generation of smart, cost-effective edge devices.