Arm has announced its latest neural processing unit, the Ethos-U85, aimed at boosting AI performance at the edge. The company also introduced a new Reference Design Platform, Corstone-320, to streamline development of AI-enabled IoT devices.
The Ethos-U85 delivers a significant performance uplift compared to its predecessor, the Ethos-U65, while maintaining high energy efficiency. Capable of scaling from 128 to 2048 MAC units, the NPU can achieve up to 4 TOPS of performance at 1GHz. This makes it well-suited for compute-intensive edge AI applications such as factory automation, commercial security cameras, and smart home devices.
One of the key advancements in the Ethos-U85 is native support for transformer networks, in addition to convolutional neural networks (CNNs). Transformer architectures have revolutionized AI in recent years, enabling breakthroughs in natural language processing, computer vision, and generative AI. By accelerating these workloads, the Ethos-U85 opens up new possibilities for edge devices to handle advanced AI tasks like video understanding and image generation.
To simplify the development process for AI-enabled IoT devices, Arm also introduced the Corstone-320 Reference Design Platform. This platform brings together the Ethos-U85 NPU with Arm's highest-performance Cortex-M85 CPU and the Mali-C55 image signal processor.
The Corstone-320 platform is supported by an extensive suite of software, tools, and virtual hardware, allowing developers to begin software development even before silicon is available. This can significantly accelerate time-to-market for complex edge AI products like smart speakers with voice assistants and vision-based IoT devices.
"As edge AI scales, silicon innovators must navigate growing system and software complexity, an ever-increasing demand for AI performance and pressure to accelerate their time-to-market," said Dennis Laudick, VP Commercial and Marketing, Machine Learning Group, Arm. "The Ethos-U85 and Corstone-320 platform address these challenges, providing a high-performance, energy-efficient solution with a streamlined development experience."
Early adopters of the Ethos-U85 include semiconductor companies Alif and Infineon. "Alif was the first to market with an edge AI solution based on Arm Cortex-M55 and Ethos-U55, and we welcome Arm's latest AI technology, Ethos-U85, which will deliver the compute performance required for our next generation Ensemble family of microcontrollers and fusion processors to address future edge AI and vision use cases," said Reza Kazerounian, co-founder and president, Alif Semiconductor.
With a global ecosystem of over 15 million developers, Arm is well-positioned to scale edge AI deployment. The company expects to see devices based on the Ethos-U85 and Corstone-320 platform running generative AI models as early as next year. As AI continues its rapid evolution, Arm's latest innovations aim to unleash new AI capabilities at the edge while simplifying development for its vast partner network.