ARM updates Neoverse enterprise processor roadmap

Nvidia’s Grace processor will be one of the first chips to use Arm’s upcoming Neoverse V2 CPU cores.

arm cortex chip image
Arm

Arm has introduced the next generation of its Neoverse high performance core technology, and Nvidia will be one of the first licensees to offer it out of the gate.

Arm introduced the Neoverse N-Series processors for data center use in 2019, along with the Neoverse E-Series for edge computing and the Neoverse V-Series for high performance computing (HPC).

For Neoverse V2, Arm is claiming higher per-thread performance at half the power consumption of its x86 competitors. Dermot O’Driscoll, vice president of product solutions at Arm, said on a conference call with journalists that the main aim of V2 is improved performance for cloud and single-thread workloads while balancing power consumption.

“We engage in close collaboration with the key cloud HPC and wireless infrastructure players, so we really understand their workloads and challenges,” O’Driscoll said. “Almost everything we do is to enable a customization specific to market needs, driven by our close collaboration with customers on their requirements for the future designs.”

The updated Neoverse V2 processor cores, code-named Demeter, are due in 2023 and will outperform the first generation in per-thread performance for cloud, hyperscale and HPC workloads. The new V-series core will feature a high bandwidth, low-latency mesh interconnect system with up to 4TB/sec of bandwidth.

Nvidia taps latest Neoverse version for Grace chip

With the introduction of the V2 cores came the announcement of a licensee: Nvidia, which will use the V2 core in the Grace processor, announced last spring. Nvidia is planning a combined Grace chip with its Hopper GPU connected by high-speed link, as well as a SuperChip that consists of two Grace dies also connected by high-speed interconnect.

“We're seeing a new kind of data center that's designed to take massive amounts of data and create artificial intelligence for business decisions. These data centers are like AI factories at Nvidia,” said Ian Buck, vice president of accelerated computing at Nvidia, during the media briefing.

“We were very impressed with V2 SPECfp and SPECint performance in order to keep up with the data parallel operations that are happening on top of the GPU itself. Also, its power performance is excellent. Its performance-per-watt of a core itself, combined with Grace's LPDDR memory, is showing upwards of two times the perf-per-watt performance compared to alternatives,” he added.

Nvidia's GTC event will be held next week, during which the company will disclose more information on Grace and other technologies.

Additional Neoverse cores to be updated

In addition to the V2 cores, Arm says that it will update the N-series and E-series next year, but it did not go into detail. The new core designs are in development and will be available to partners later next year.

The N-series is designed for applications where thread count is important, while the E-series is designed for data plane processing applications, such as 5G RAN, edge networking, and other forms of network acceleration.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:

Copyright © 2022 IDG Communications, Inc.