STMicroelectronics boosts AI at the edge with NPU-accelerated microcontroller

The latest generation of microcontroller units, or MCUs, promises to enable developers to push more artificial intelligence to edge devices such as robots. STMicroelectronics International NV this month announced a new MCU series that it claimed is its most powerful to date and integrates accelerated machine learning capabilities.

“We are on the verge of a significant transformation at the tiny edge. This transformation involves the increasing augmentation or replacement of our customers’ workloads by AI models,” stated Remi El-Ouazzane, president of the Microcontrollers, Digital ICs and RF Products Group (MDRF) at STMicroelectronics. “Currently, these models are used for tasks such as segmentation, classification, and recognition. In the future, they will be applied to new applications yet to be developed.”

Geneva, Switzerland-based STMicroelectronics said it has more then 50,000 staffers making semiconductor technologies. The integrated device manufacturer works with over 200,000 customers and thousands of partners to design and build products, systems, and ecosystems that address their challenges.

Register today to save 40% on conference passes!

STM32N6 includes new features

STMicroelectronics said its STM32N6 series is the first MCU to embed its proprietary neural processing unit (NPU), the Neural-ART accelerator, which it designed for power-efficient edge AI applications. The company asserted that the MCU can deliver 600 times more machine-learning performance than a high-end STM32 MCU today.

The STM32N6 features an 800 MHz Arm Cortex-M55 core and boasts 4.2 MB of RAM for data-intensive workloads. The Neural-ART Accelerator runs at 1 GHz and provides up to 600 GOPS (Giga operations per second), enabling real-time neural network inference for computer vision and audio applications, said STMicroelectronics.

For the first time on an MCU, the STM32N6 incorporates an image signal processor (ISP) that provides direct signal processing, enabling the use of simple and affordable image sensors, the company said. This ISP can be configured with the free STM32-ISP-IQTune software for customizable parameters such as exposure time, contrast, or color balance.

“[The STM32N6] will utilize a new release of our unique AI software ecosystem package,” said El-Ouazzane. “This marks the beginning of a long journey of AI hardware-accelerated STM32, which will enable innovations in applications and products in ways not possible with any other embedded processing solution.”

STMicroelectronics added that a dedicated computer vision pipeline with a MIPI CSI-2 interface ensures compatibility with a wide range of cameras. The STM32N6 also features an H264 hardware encoder and the NeoChrom Accelerator for graphics, making it suitable for feature-rich products.

Available in six different packages ranging from 169 to 264 pins and from 0.4 to 0.8 mm pitch, the MCU supports up to 125°C ambient temperature, making it suitable for harsh environments.

“It is a common misconception that AI is purely a big data center, power-hungry application,” observed Tom Hackenberg, principal analyst in the Memory and Computing Group at Yole Group.

“This is no longer true,” he said. “Today’s IoT [Internet of Things] edge applications are hungry for the kind of analytics that AI can provide. The STM32N6 is a great example of the new trend melding energy-efficient microcontroller workloads with the power of AI analytics to provide computer vision and mass sensor-driven performance capable of great savings in the total cost of ownership in modern equipment.”

STMicroelectronics MCU supports AI, app dev

STMicroelectronics said its Edge AI Suite includes software tools for the development of edge machine-learning applications, including the possibility to “bring your own model” in formats such as TensorFlow Lite, Keras and ONNX.

In addition, the new microcontroller supports a growing AI “model zoo,” offering enhanced performance and versatility for various applications. Users can explore the expanding collection of AI models to achieve better results and shorten their time to market.

The STM32N6 offers advanced security features, meeting the latest standards with target SESIP Level 3 and PSA Level 3 certifications.

The STM32N6 has been available to select customers since October 2023 and is now ready to be offered in high volumes. Companies such as LG Electronics are integrating the STM32N6 into their products. 

“The STM32N6 delivers remarkable AI performance and provides excellent flexibility in a small silicon package ideal for embedded systems and wearable devices,” said Yehan Ahn, research and development task leader at LG’s CTO Division. “The inference speed, powered by the groundbreaking Neural-ART Accelerator, has exceeded our expectations and ST’s developer-friendly software tools let us seamlessly integrate our AI models into the MCU.”

“We have rigorously evaluated ST’s latest microcontroller, STM32N6, in our labs, noting its outstanding neural processing performance and excellent power efficiency,” said Seiichi Kawano, principal researcher at Lenovo Research. “This microcontroller represents the beginning of a new era, accelerating our research of ‘AI for all’ technologies at the edge.”