Alif ships first Ethos U85 AI microcontroller

Alif Semiconductor has made a significant stride in the field of microcontrollers by shipping the first device equipped with the ARM Ethos U85 neural processing unit, capable of handling transformer AI models at the edge. This achievement marks a milestone in the industry, showcasing the cutting-edge technology integrated into the second-generation multicore microcontroller unveiled earlier this year. The Ethos U85 core, in conjunction with two Ethos U55 NPU cores in the high-end E8 chip, enables the utilization of small language models (SML) and transformer-based vision models, setting a new standard for performance and efficiency.

One of the key advantages of the Alif Semiconductor family of microcontrollers is its utilization of a fully depleted silicon on insulator (FD SOI) process, designed to optimize power consumption. With just 1.3 µA consumed in STOP mode and as low as 27 µA/MHz in operation, the microcontrollers offer a compelling solution for power-conscious applications. Additionally, the integration of MRAM non-volatile memory further contributes to minimizing power usage, making these devices ideal for energy-efficient operations.

• ARM supports embedded transformer models with U85 NPU
• How transformer models impact edge AI chip designs

According to Raza Kazerounian, president and co-founder of Alif Semiconductor, the focus of the company lies in delivering a scalable platform that prioritizes power efficiency and security for endpoint applications. Kazerounian emphasizes the significance of incorporating an NPU that supports transformers in the second-generation microcontroller, highlighting the innovative approach taken by Alif Semiconductor in catering to diverse AI requirements.

Despite the challenges posed by the complexity of AI technologies, Kazerounian underscores the importance of utilizing common IP such as the U85 to facilitate broader developer adoption. By leveraging established ARM IP, Alif Semiconductor aims to expand the ecosystem of AI applications, enabling a wider range of functionalities and use cases. The introduction of the U85 core opens up new possibilities for diverse applications, signaling a significant advancement in the field of edge AI.

The latest generation of microcontrollers from Alif Semiconductor not only features enhanced performance capabilities but also incorporates advanced functionalities such as an image pipeline, two MIPI CSI lines, and a GPU onboard for camera integration. With the current U85 core delivering 204GOPS of performance, plans are already in place for higher performance versions in subsequent generations, offering increased flexibility and efficiency for a variety of applications.

Camera integration

Looking ahead, Kazerounian reveals that future iterations of the microcontrollers will further enhance performance, with the potential for implementing cores with varying performance levels. The versatility of the U85 core, ranging from 0.5TOPS to 3TOPS, underscores its adaptability to different application requirements, ensuring that the microcontrollers remain at the forefront of technological innovation.

With the availability of the Ensemble E4, E6, and E8 series of Ensemble GenAI products, Alif Semiconductor continues to push the boundaries of edge AI applications. By offering AI-ready architecture from the outset, Alif Semiconductor empowers customers to leverage transformer-based models and generative AI in edge and endpoint products powered by small batteries. The E4, E6, and E8 devices represent a significant leap forward in edge computing, providing a platform for implementing advanced AI functionalities in a compact and energy-efficient manner.