British semiconductor firm Arm Ltd. today unveiled a revamped product roadmap with new processor designs that it says will provide a foundation for multiple kinds of workloads over the next few years.

Arm’s roadmap sees the addition of Arm Neoverse V2, code-named Demeter, a new processor that’s expected to play a key role in data centers years into the future. The company said its latest core is built to provide the highest per-thread performance for cloud, hyperscale and high-performance computing workloads.

In an online presentation, Arm Chief Executive Rene Haas said industry leaders across cloud, HPC, 5G and edge computing have selected Arm Neoverse as the compute foundation for multiple next-generation infrastructure platforms. Arm Neoverse was introduced in 2018 as the company’s answer to Intel Corp.’s Xeon processors and Advanced Micro Devices Inc.’s Epyc chips, and has since evolved to three distinct platforms — the Arm V-series, N-series and E-series — each of which is aimed at different compute tasks.

Arm Vice President of Product Solutions Dermot O’Driscoll said during the presentation that the V-series chip designs have been adopted by more than 20 chipmakers. The blueprints enable chipmakers to create new solutions at an extremely rapid pace and quickly meet the ever-changing needs of both customers and the market, he said.

O’Driscoll explained that the cloud infrastructure of tomorrow will need to be able to handle an explosion of new data while effectively processing increasingly more complex workloads. At the same time, customers want to see increased power efficiency too, as they look to minimize their carbon footprints.

In particular, he said customers are looking for solutions that push cloud workload performance without additional power and area requirements. Arm’s response to these demands is the Neoverse V2 platform, he said, which will deliver market leading integer performance for cloud and HPC workloads.

What’s more, it will be coming soon. Arm said multiple partners are already working to get their chip designs based on Neoverse V2 to market. One of these is Nvidia Corp., which is using the V2 as the foundation for its new Grace data center central processing unit. Grace is said to combine the power efficiency of V2 with LPDDR5X memory to deliver twice the performance per watt compared with servers powered by traditional architectures.

Arm said Nvidia’s Grace CPUs are already in manufacturing and that several other customers also have chips based on the V2 blueprint in late-stage development. Those customers include the likes of Ampere Computing LLC, Marvell Technologies Inc. and Amazon Web Services Inc.

“The new datacenter roadmap is the most aggressive I have ever seen from Arm,” said analyst Patrick Moorhead of Moor Insights & Strategy. “Arm is projecting it will have the highest-performance CPU versus AMD’s or Intel’s equivalent CPUs. As I have said before, industry standard-architecture does not connote performance, but rather design. Arm reinforced that today.”

Less information was provided about the upcoming N-series and E-series chip designs, which are set for a 2023 launch. While the V-series cores are designed to push the limits of performance, the N-series chips are tailored for applications where thread count is prioritized over single-threaded performance. Meanwhile, the Arm E-series cores are meant for data plane processing applications such as edge networking, 5G RAN and other kinds of acceleration.

Arm said the third iteration of its N-series core designs is in development and will be available to partners later next year. The E2’s successor is currently under development and should also be available next year.

Photos: Arm