Arm Takes On Intel With Neoverse Platforms For Edge, Cloud And 5G

Arm is upping its competition against Intel in data centers, edge servers and 5G networks with the newly revealed Neoverse N1 and E1 CPU platforms.

While Arm doesn't directly compete with Intel, the move is meant to enable the SoftBank-owned company's silicon partners to design new kinds of processors that power the increasingly growing number of cloud and edge servers in the world where Intel has staked its future growth. Since unveiling its Neoverse architecture last fall, Arm has landed multiple silicon partners making Neoverse-based processors, including Amazon Web Services, Huawei and Ampere, which pay Arm for IP licensing costs.

[Related: Watch Out, Intel: AMD To Release 7nm Ryzen, EPYC CPUs In Mid-2019]

Drew Henry, the head of Arm's infrastructure business, said Arm's Neoverse N1 platform, which is based on its Ares microarchitecture, will provide scalable and "world-class server compute performance levels" for everything from edge servers to hyperscale data centers. He added that N1 is optimized for next-generation 7-nanometer process technology, a significant step from the company's previous chips that are based on 16nm and 14nm.

"With our platform, we give our customers the flexibility to design a very large CPU core complex that is literally up to 128 cores" down to four cores for edge workloads, he said. "We're able to do that because we redesigned completely the interconnect, the way the cores communicate."

The Neoverse N1 platform will provide performance and a total cost of ownership that are competitive with existing processors available on the market today, according to Henry. He said the platform can enable silicon partners "to deliver market-leading socket performance at much lower power over production and sampling x86 servers" based on a comparison with existing systems.

"We believe this will drive significant improvements in TCO never seen before from cloud data centers out to edge compute platforms," Henry said.

Henry said Neoverse N1 will provide a roughly 60 percent performance boost over the company's current-generation Cosmos platform, which is double the performance gain the company first promised when it revealed its multi-year Neoverse product roadmap last fall.

Among the figures provided, Henry said Neoverse N1 will provide a 2.5X performance gain over Arm's Cosmos-based Cortex A72 for key cloud workloads, such as the popular web server Nginx. The new platform is also 1.7X faster than its predecessor for Java applications.

"We believe that this is one of the key products for us to accelerate this whole transformation to this new internet infrastructure," Henry said.

Arm's Neoverse E1 platform, on the other hand, is focused on providing high throughput for network infrastructure, from multi-100 gigabit routers to 5G base stations. Henry said the platform will provide 2.7X more throughput performance, 2.4X more throughput efficiency and over 2X more compute performance when compared to the Arm Cortex-A53.

Henry said Arm's silicon partners are expected to begin releasing processors based on the Neoverse N1 and E1 platforms by the end of the year with a ramp-up of production next year, though he was unable to disclose any of the partners’ names.

Beyond the British company's announced silicon design wins with partners like AWS and Huawei, Henry boasted a growing ecosystem of software partners for Neoverse that includes VMware, Docker, Kubernetes, MySQL, Apache, Ubuntu, Oracle and Red Hat.

"We're working with more and more people in the ecosystem and getting a substantial amount of support," Henry said.