Arm introduces Arm v9 architecture

British Arm has announced a new generation of its Arm architecture. The v9 architecture succeeds the v8 generation released in 2011 and should be used in billions of devices over the next decade.

Socs based on the Arm v9 architecture are divided into three profiles: m, for microcontrollers, a, for application processors, and r, for real-time processors. The v9 architecture puts heavy emphasis on artificial intelligence and security. For better handling of AI-related computations, the Arm v9 architecture has received an updated Scalable Vector Extension, SVE2, co-developed with Fujitsu. In addition, AI workloads would be accelerated with improved matrix calculations in the CPU cores and innovations in the Mali GPU and Ethos-npu, the neural processing unit. Nvidia would also contribute to the acceleration of AI calculations.

The security of data should be guaranteed thanks to the Confidential Compute Architecture, or CCA, with which the soc can effectively compartmentalize data and code, so that different processes cannot access each other’s data. Arm calls the compartments Realms and data can be safely stored in them whether in use or idle. This makes the realms somewhat like virtual machines or sandbox environments and would protect data against commonly used memory exploits such as buffer overflows and reuse after free vulnerabilities. In addition, Arm is involved in the Morello project, a security initiative in collaboration with the British government.

In addition to security and AI, effective design is also a spearhead of Arm. Not only does it aim to get thirty percent more performance from CPU cores in the next two generations, but the company also wants to focus more on system-driven soc design. In this Total Compute design philosophy, socs must be designed specifically for specific applications, in order to produce hardware optimally for its intended purpose. Several partners have indicated that they are working on Arm v9 products, including Google, Marvell, MediaTek and Samsung.

According to Arm, some 100 billion Arm processors have been in use in the past twenty-six years. The company expects to pass the next hundred billion ARM-socs milestone over the next five years. That growth would be driven by increasing compute requirements for IoT and cloud services, with more and more data being processed locally rather than in data centers to stem the explosion of data.