Arm pushes the case for more robust AI at the IoT edge

Arm has been pushing the case for more AI and machine learning (ML) capabilities at the edge for a while now, with past product launches like its Cortex-M52 portfolio designed to extend these capabilities even into the smallest edge IoT form factors. This week, the company came to Embedded World with two more key announcements aimed at furthering edge AI and ML in IoT scenarios.

“Edge AI presents opportunities and challenges for our partners, who are navigating the growing performance demands of AI, balanced with the need for low-power designs, pressure to speed up innovation, and the need for machine learning integration in an increasingly complex scape,” said Paul Williamson, senior vice president and general manager of the IoT Line of Business at Arm, during a press preview.

Arm’s latest device aimed at fulfilling some these needs is Arm Ethos-U85 neural processing unit, which compared to its earlier Arm devices delivers a. a 4x performance uplift, but also boosts power efficiency by 20%, better positioning it for a broader variety of edge IoT applications where power efficiency is key.

The Ethos-U85 NPU, which already has been licensed by more than 20 Arm partners, including Alif Semiconductor and Infineon, scales from 128 to 2048 MAC units (4 TOPs @1GHz), and Williamson said it is targeted at rapidly emerging edge applications that are most in need of AI processing, such as surveillance cameras and  the automated machines in smart factory settings. The Ethos-U85 also provides support for AI frameworks such as TensorFlow Lite and PyTorch, and supports transformer networks enabling smart computer vision and generative AI use cases.

“With the deployment of microprocessors into more high-performance IoT systems for use cases such as industrial machine vision, wearables and consumer robotics, we’ve designed the Ethos-U85 to work with our leading Armv9 Cortex-A CPUs, to accelerate ML tasks and bring power-efficient edge inference into a broader range of higher-performing devices,” Williamson stated in a blog post.

“Edge AI use cases are becoming increasingly sophisticated and require secure, high performance compute systems to deliver on the opportunities of the AI era,” said Steve Tateosian, SVP of Industrial MCUs, IoT, Wireless and Compute Business, Infineon. “We look forward to building on Infineon’s long-standing partnership with Arm and enabling these sophisticated systems with Arm Ethos-U85 and the transformer network support it provides for edge AI deployments.” 

Arm also this week announced a new IoT reference design platform called Corstone-320, which Williamson said packages the high-performance Arm Cortex-M85 CPU, the Arm Mali-C55 Image Signal Processor and the new Ethos-U85 NPU improve performance of edge AI applications for voice, audio, and vision, such as real time image classification and object recognition, or enabling voice assistants with natural language translation on smart speakers. 

The platform includes software, tools, and support including Arm Virtual Hardware. This combination of hardware and software will accelerate product timelines by enabling software development to start ahead of silicon being available, rapidly improving time to market for increasingly complex edge AI devices, the company stated.