ARM Announces Next-Gen Processor Architecture

Arm Unveils the Armv9 Architecture

Arm has officially announced Armv9, representing the next evolution in its chip architecture. The previous generation, Armv8, debuted a decade ago and, despite numerous refinements, the advancements within Armv9 necessitate a version number update.

Armv9 builds upon the foundation of v8, maintaining backward compatibility. However, it introduces significant enhancements focused on security, artificial intelligence, signal processing, and overall performance.

A Decade of Growth and Future Projections

Over the past five years, more than 100 billion chips based on Arm technology have been shipped globally. Arm anticipates that its partners will ship over 300 billion such chips in the coming decade.

The first devices powered by Armv9-based chips are expected to become available later this year.

The Vision Behind Armv9

Ian Smythe, Arm’s VP of Marketing for its client business, expressed his belief that this new architecture will fundamentally reshape the landscape of computing over the next ten years.

He stated that the goal is to deliver increased performance, bolster security capabilities, and improve workload handling in response to the evolving demands of modern computing.

This shift is driven by the exponential growth of data, the necessity for efficient processing, the need for rapid data transfer, and the critical importance of data protection.

Enhanced Security with Confidential Computing and Realms

A core element of the Armv9 updates is the introduction of Arm’s confidential compute architecture and the innovative concept of Realms.

Realms empower developers to create applications where data is shielded from the operating system and other applications running on the device.

For instance, a business application could utilize Realms to safeguard sensitive data and code from unauthorized access.

The Architecture of Trust

Richard Grisenthwaite, Arm’s chief architect, explained that Arm Confidential Compute Architecture addresses concerns about the inherent vulnerabilities within operating systems and hypervisors.

These systems, while complex, can be susceptible to breaches. The architecture aims to move certain workloads to a significantly smaller and more secure code base.

Only the Realm manager would have access to the data while it’s in use, representing a fraction of the size of a typical hypervisor or operating system.

Addressing Past Challenges

Grisenthwaite acknowledged that developing this security architecture required several years of meticulous work and refinement.

The emergence of vulnerabilities like Spectre and Meltdown prompted a reassessment of initial solutions to ensure robustness against similar attacks.

Boosting AI Capabilities with SVE2

Another key focus area was enhancing the CPU’s ability to handle artificial intelligence workloads, which are becoming increasingly prevalent.

Arm previously introduced its Scalable Vector Extension (SVE) for high-performance computing applications, such as the Arm-powered Fugaku supercomputer.

Now, Arm is introducing SVE2 to expand AI and digital signal processing (DSP) capabilities.

These enhancements can be applied to image processing, IoT devices, and smart home solutions.

While dedicated AI chips exist, Arm believes that optimizing the entire computing stack for AI workloads is crucial, and that CPUs remain a viable choice for many applications, particularly smaller ones.

A Holistic Approach to Machine Learning

Grisenthwaite emphasized that machine learning will be integrated across various computing platforms, including GPUs, dedicated processors, neural processors, and CPUs.

Improving the performance of all these components in machine learning tasks is paramount.

Performance Gains and Industry Collaboration

Arm projects that its new architecture will enable chip manufacturers to achieve over 30% improvements in compute power over the next two chip generations.

These gains will benefit both mobile CPUs and infrastructure CPUs used by cloud providers like AWS.

Min Goo Kim, the executive vice president of SoC development at Samsung Electronics, stated that Armv9 offers a substantial leap forward in security and machine learning.

He anticipates that this architecture will drive innovation in Samsung’s next-generation Exynos mobile processors through continued collaboration with Arm.