Entering the IoT AI market, Arm launches a new Cortex-M52 processor design

In order to adapt to the AIization of the Internet of Things, Arm announced the launch of Cortex-M52, which handles higher digital signal processing (DSP) and machine learning performance requirements. Among them, the Cortex-M52 processor can provide higher AI inference performance on the smallest device, making it the smallest, most cost-effective processor among all supported Arm Helium technologies, providing stronger AI functionality for lower cost IoT devices.

Arm stated that with the evolution and development of AI IoT technology, terminal intelligence is being deployed at the lowest price point on smaller, more cost conscious, and typically battery powered devices, and with higher privacy and reliability due to reduced reliance on the cloud. As the shipment volume of IoT products supporting AI continues to increase, partners need to acquire more machine learning features and simplified development processes in order to have the flexibility to innovate and expand quickly.

Arm pointed out that the addition of Arm Helium technology to Cortex-M52 improves the performance of digital signal processing and machine learning applications for small low-power embedded devices. It can provide digital signal processing functions without the need for dedicated NPUs, which helps save chip area and costs, and deploy more machine learning inference algorithms on terminal devices; Helium technology has been successfully deployed to various products at the edge of network terminals, and Cortex-M52 can assist Arm partners in importing this technology into lower cost and more power limited devices.

Cortex-M52 simplifies the transfer path from Cortex-M33 and Cortex-M4, and can meet the needs of various AI IoT applications, assisting in achieving richer user interfaces, voice and visual experiences, such as automotive and industrial control, predictive maintenance, wearable sensor fusion, and providing expansion flexibility to cover various performance points and configurations.

This processor features the smallest area and lowest power consumption among all Cortex-M processors that support Helium, providing another option for partners to choose from in terms of performance and cost. Compared to previous generations of Cortex-M processors, the machine learning performance of Cortex-M52 has increased by 5.6 times, and the digital signal processing performance has improved by 2.7 times. In addition, Cortex-M52, Cortex-M55, and Cortex-M85 are fully compatible in software, allowing developers to benefit from the continuously growing Helium software and tool ecosystem.