Arm Releases Ethos-U85 Neural Processing Unit and IoT Reference Design Platform

Author photo: Patrick Arnold
ByPatrick Arnold
Company and Product News

Semiconductor and software designer Arm has announced the release of the Ethos-U85, their most efficient Neural Processing Unit (NPU) to date. The Ethos-U85 delivers a 4 times performance increase and higher power efficiency compared to its predecessor while scaling from 128 to 2048 Multiply-Accumulate (MAC) units, hardware components that are fundamental to some machine learning processes. Arm’s new NPU is designed to address applications with growing performance demands such as factory automation and smart cameras. Ethos-U85 offers a consistent toolchain so partners can leverage existing investments for a seamless developer experience, and it provides support for established AI frameworks such as TensorFlow Lite and PyTorch. 

Neural Processing Unit

As AI deployed on edge systems becomes more sophisticated, silicon designers face growing challenges in system and software complexity and higher expectations for AI performance. At the same time, software developers require streamlined, uniform experiences and seamless integration with new AI frameworks and libraries. 

Ethos-U85 supports Transformer Networks as well as Convolutional Neural Networks (CNNs) for AI inference on the edge. Applying these machine learning techniques will drive new applications in vision and generative AI use cases for tasks like enhancing video understanding, correcting and filling in images, and analyzing camera data for image classification and object detection. As enterprises increasingly deploy microprocessors in high-performance IoT systems for emerging use cases such as industrial machine vision, wearables and robotics, the Ethos-U85 synergizes with the Armv9 Cortex-A CPUs to accelerate ML tasks and spread power-efficient edge inference to a broader range of hardware platforms.

New IoT Reference Design Platform, Corstone-320, Accelerates Voice, Audio, and Vision Systems

Arm has additionally announced the Corstone-320 IoT Reference Design Platform, combining the Cortex-M85 CPU, the Mali-C55 Image Signal Processor, and the new Ethos-U85 NPU to deliver the necessary performance required for edge AI applications such as real time image classification, object recognition, and enabling voice assistants. The platform includes software, tools, and support to accelerate software development ahead of hardware availability, rapidly improving time to market for complex edge AI devices.

For more ARC coverage on industrial AI, please visit our website.

Engage with ARC Advisory Group

Representative End User Clients
Representative Automation Clients
Representative Software Clients