The British chip designer Arm Holdings Ltd. introduced a roadmap for its next-generation central processing unit architecture today.

Its new breed of chips will be optimized for specialized workloads such as artificial intelligence and digital signal processing and come with greater security built right into the silicon. Arm, which is in the process of being acquired by Nvidia Corp., creates and licenses the basic architecture that’s used in most mobile processors today, including the chips that power smartphones sold by companies such as Apple Inc. and Samsung Electronics Ltd.

Armv9 is well overdue, the first major update to its CPU architecture in a decade, and its primary focus is on two things: better performance and better security.

Arm Chief Executive Simon Segars said in a video presentation that greater performance is necessary because the chip industry is rapidly moving away from general-purpose compute to more specialized applications as AI, the “internet of things” and 5G gain momentum. He noted that more than 100 billion devices powered by Arm-based chips were shipped in the last five years.

“At the current rate, 100% of the world’s shared data will soon be processed on Arm, either at the endpoint, in the data networks or in the cloud,” Segars said. “Such pervasiveness conveys a responsibility on Arm to deliver more security and performance.”

Arm said it will drive that greater performance by integrating the Scalable Vector Expansion technology at the heart of Fugaku, the world’s fastest supercomputer built by Fujitsu Ltd., with its Armv9 chips. The SEV2 technology is designed to speed up specialized machine learning and digital signal processing workloads across a range of applications. It does this by enhancing the processing ability of 5G systems, virtual and augmented reality and ML workloads such as image processing that run locally on CPUs.

“Over the next few years, Arm will further extend the AI capabilities of its technology with substantial enhancements in matrix multiplication within the CPU,” the company said.

Segars said Arm has been pushing hard on the AI capabilities of its processors for a number of years and that those efforts have paved the way for a vast range of specialized applications already. “But AI remains hungry for ever more efficient compute power,” Segars said. “Our partners want to realize their own individual AI futures and it’s up to us to take every opportunity to help them.”

Arm said it will be able to deliver CPU performance gains of more than 30% over the next two generations of its mobile and server chips. In addition, the Armv9 architecture incorporates what the company calls a “Total Compute design methodology” that will further increase performance via a series of system-level hardware and software optimizations.

“Arm is making it easier to integrate machine learning capabilities into the end product,” Patrick Moorhead, president and principal analyst at Moor Insights & Strategy, told SiliconANGLE. “It’s important to recognize that for most performance cases, especially CPU, it’s more about the architecture of the design versus the instruction set. So in other words, chip designers still need to architect something performant.”

Arm also said it plans to apply the Total Compute design principles across its entire portfolio of automotive, client, infrastructure and IoT chips. And it’s promising even more enhancements in future too, with work underway to increase the frequency, bandwidth and cache size, and reduce the memory latency of Arm9 chips.

“Chip architectures power infinite computing, the pervasive and ubiquitous cheap availability of compute resources, so when a major player like Arm releases a new reference architecture, it is time to pay attention,” said Constellation Research Inc. analyst Holger Mueller. “Not surprisingly, the key architecture focus areas are centered on AI, with Arm partnering with Fujitsu to deliver SVE2. We are probably going to hear a lot more about that soon.”

On security, Arm said it’s embracing a new Confidential Compute Architecture in the Armv9 roadmap. Confidential computing is about encrypting data while it’s being processed in-memory, so it cannot be exposed to other parts of the computer system. It’s often referred to as the Holy Grail of data security, because though methods for encrypting data at rest and data in transit have already been widely adopted, there’s no reliable way to secure data as it’s actually being used.

Arm says its Confidential Compute Architecture will change that by shielding code and data from being accessed or modified while it’s being processed, by doing the computations in a hardware-based secure environment. More specifically, Arm said its CCA introduces the concept of “Realms” that can be used by all applications.

Realms are a region of the chip that is kept separate from both the secure and non-secure worlds, Arm said. With business applications for example, Realms will be able to protect sensitive data and code from the rest of the app while it’s being used, and also while it’s in transit or at rest. That, Arm says, is a big deal, citing a recent Pulse survey that found 90% of enterprise executives believe confidential computing will substantially reduce the cost of security.

In his presentation, Segars spoke of some of the immediate benefits of CCA. He talked about the post-COVID-19 world and the need for people to store medical information on their devices so that critical information might be accessed quickly.

“But for me to get comfortable with something like that, I’d want advanced encryption running on my device, beyond what’s possible today,” Segars said. “That’s where our Armv9 architecture comes in with features such as memory tagging to eliminate memory safety issues and Realm management that provides a vault for secure data manipulation.”

David Floyer, an analyst with SiliconANGLE sister market research firm Wikibon, said it was clear that Arm’s focus on end-to-end security fits with its vision that every bit of data will go through Arm processors at some stage of its life, either at the edge, in the network or the cloud.

“It’s placing an emphasis on hardware, firmware and software to deliver a platform that will guarantee that containers, virtualization, Linux software platforms and the software that runs on them, will run without change on Arm processors,” Floyer said. “Arm has the best end-to-end vision for security that I have seen.”

Image: Arm