Arm Neoverse V1 and N2, Optimized for Infrastructure, Unveiled
Focus on V1 performance and N2 usability... Support for SVE features
CMN-700 Increases V1·N2-Based SoC Design Flexibility


On the 28th, Arm held an online press conference and unveiled the 'Neoverse V1 and N2' platforms and the 'CMN-700' system IP required for SoC configuration based on them.

Unlike Cortex, which is designed for embedded use, Neoverse is designed to meet the performance and power efficiency requirements of all infrastructure computing, from the cloud to the edge.

The V1, the first product in Arm's "performance-first" compute tier, delivers a 50% performance boost over the N1, delivering a 1.8x performance boost across a wide range of vector workloads and a 4x performance boost for machine learning workloads. It provides SoC design flexibility to build compute solutions tailored to applications that rely on CPU performance and bandwidth. SVE (Scalable Vector Extension) also ensures high performance per core and long code life.

The N2 is the first platform based on the Armv9 architecture, enhancing performance, power efficiency, and security. It delivers a 40% improvement in single-threaded performance while maintaining the same power and area efficiency as the N1 platform. Compared to the N1, NGINX performance is 1.3x faster, and DPDK (Data Plane Development Kit) packet processing speed is 1.2x faster.

It supports everything from hyperscale cloud to edge and 5G use cases, and with SVE2, it provides high performance, ease of programming, and SVE portability for use cases such as machine learning, digital signal processing (DSP), multimedia, and 5G.

The CMN-700 system IP, required for high-performance SoC designs based on Neoverse V1 and Neoverse N2, can incrementally increase performance across all vectors, including core count, cache size, memory capacity, and the number and type of IO devices. It also supports next-generation use cases for multi-chip and memory expansion, as well as accelerators.

Additionally, Arm continues to invest in the development of Cache Coherent Interconnect for Accelerators (CCIX) and Compute Express Link (CXL) technologies to provide tailored options for them. Arm plans to provide high-speed fabric and core scalability to partner solutions, enabling flexibility in tightly connected heterogeneous computing.

“Marvel’s ‘OCTEON’ family of network solutions based on the Neoverse N2 platform will begin sampling later this year,” said Jeong Seong-hun, FAE director at Arm Korea. “India’s Ministry of Electronics and Communications Technology (MeitY), France’s SiPearl, and Korea’s Electronics and Telecommunications Research Institute (ETRI) will license Neoverse V1 for exascale HPC projects.”

He added, “AWS is utilizing its Arm architecture-based ‘Graviton2’ processor for its cloud services,” and “Oracle, Alibaba Cloud, Tencent, and others are considering introducing Arm-based processors and Neoverse technology.”