Arm has recently launched its latest neural processing unit (NPU), Ethos-U85, which aims to significantly enhance the AI performance of edge devices. At the same time, Arm has also introduced a new reference design platform, Corstone-320, to further simplify the development process of IoT devices in terms of AI capabilities.
Compared to its predecessor, Ethos-U65, Ethos-U85 achieves a performance leap while maintaining high efficiency. The MAC unit of this NPU (neural processing unit) can be flexibly expanded from 128 to 2048, and its performance can reach an astonishing 4 TOPS at a frequency of 1GHz. This makes Ethos-U85 particularly suitable for compute-intensive edge AI applications, such as factory automation, commercial security cameras, and smart home devices.
One of the highlights of Ethos-U85 is that it not only supports traditional convolutional neural networks (CNN), but also supports the currently popular Transformer networks. In recent years, the Transformer architecture has made breakthrough progress in the field of AI, injecting new momentum into the development of natural language processing, computer vision, and generative AI. By accelerating these workloads, Ethos-U85 provides new possibilities for edge devices to handle advanced AI tasks such as video understanding and image generation.
To simplify the development process of IoT devices in terms of AI capabilities, Arm has also launched the Corstone-320 reference design platform. This platform perfectly combines the Ethos-U85 NPU with Arm's high-performance Cortex-M85 CPU and Mali-C55 image signal processor.
Corstone-320 platform has received extensive software, tool, and virtual hardware support, allowing developers to start software development before the silicon chip is launched. This initiative will greatly shorten the time to market for complex edge AI products such as smart speakers with voice assistants and vision-based IoT devices.
Arm Machine Learning Group's Vice President of Business and Marketing, Dennis Laudick, said, "With the continuous development of edge AI, chip innovators need to face the growing complexity of systems and software, the increasing demand for AI performance, and the pressure to accelerate product launches. Ethos-U85 and Corstone-320 platform are designed to address these challenges. They provide high-performance, energy-efficient solutions and bring a more convenient development experience."
Early adopters of Ethos-U85 include semiconductor companies Alif and Infineon. Reza Kazerounian, Co-founder and President of Alif Semiconductor, said, "Alif is the first company to bring edge AI solutions based on Arm Cortex-M55 and Ethos-U55 to the market. We are very excited about Arm's latest AI technology, Ethos-U85, which will provide the required computing performance for our next-generation Ensemble series microcontrollers and fusion processors to meet the needs of future edge AI and vision applications."
Arm has a unique advantage in driving edge AI deployment with its ecosystem of over 15 million developers worldwide. The company predicts that devices based on Ethos-U85 and Corstone-320 platform will start running generative AI models as early as next year. With the rapid development of AI technology, Arm's latest innovations aim to unleash new AI capabilities for edge devices while providing a more convenient development experience for its vast partner network.
