First ARMv9 processor for edge AI is fundamental shift

ARM has made a significant advancement in the field of edge AI and embedded applications by introducing its first processor core based on the ARMv9 instruction set. This development marks a departure from the traditional use of high-end microcontrollers in such applications, according to Paul Williamson, the general manager of the project.

The transition to a microprocessor from a microcontroller is primarily driven by the need for efficient memory management in AI frameworks. The new ARM Cortex-A320 core can be clustered with up to four cores and a U85 neural processing unit (NPU) to create a powerful embedded processor tailored for edge AI tasks.

Williamson emphasized the groundbreaking nature of this development, stating, “This isn’t just an incremental step forward, it’s a fundamental shift and we believe it will drive the edge AI forward for many years to come. We developed this from the ground up specifically for edge AI.”

• Paul Williamson interview: ARM moves to support transformer AI models in embedded applications
• Low power M55 microcontrollers with U85 for edge AI
• Renesas ships M85 microcontroller

The ARM Cortex-A320 is designed to be versatile for embedded developers, capable of running real-time operating systems like FreeRTOS and Zephyr, as well as high-level Linux and Android operating systems. It can also handle AI models with over 1 billion parameters, such as large language models (LLMs), enabling a wide range of AI applications.

According to Williamson, the increasing complexity of AI workloads necessitates higher performance at the edge. He highlighted the importance of running real-time operating systems on microcontrollers to enhance memory flexibility, with the ARM Cortex-A320 offering maximum flexibility for developers.

While the introduction of the ARM Cortex-A320 represents a significant step forward in edge AI capabilities, Williamson clarified that it does not signal a shift away from microcontrollers. He acknowledged the strengths of microcontrollers in power optimization and highlighted the complementary nature of the ARMv9 architecture enhancements for machine learning performance.

Furthermore, the ARMv9 architecture brings additional security features, such as memory tagging extensions and pointer authentication, to enhance the overall security of AI frameworks. The ARM Cortex-A320 is positioned as the entry-level core in the ARMv9 family, with higher-tier cores like the A520 and A725 offering increased performance for diverse applications.

• ARMv9 architecture looks to a decade of chips
• ARM details updated v9 architecture

Looking ahead, Williamson envisions the ARM Cortex-A320 being utilized in a wide range of devices, from consumer wearables to autonomous vehicles and infrastructure applications. The core's scalability and performance make it suitable for various use cases, with partners already engaged in silicon development for future implementations.

In conclusion, the introduction of the ARM Cortex-A320 core represents a significant milestone in advancing edge AI capabilities and embedded applications. With a focus on memory management, performance optimization, and security enhancements, ARM is poised to drive innovation in the field of AI processing for years to come.