Arm Neoverse CSS N4: Fast, Scalable, and Flexible (www.arm.com)

Arm has introduced the Neoverse CSS N4, the latest addition to their compute subsystem lineup, designed to accelerate the development of AI data centers. This innovative offering allows silicon designers to configure CPU cores, cache sizes, memory, and I/O options, providing the flexibility needed to build both single and multi-chiplet systems. With significant performance improvements—up to 2x performance per socket and 1.75x memory bandwidth compared to its predecessor—the Neoverse CSS N4 is optimized for next-generation agentic AI workloads, supporting faster integration and reduced time-to-market through a pre-integrated architecture. The Neoverse CSS N4 is particularly relevant for the AI/ML community as it lays the groundwork for enhanced AI infrastructure. It supports critical advancements such as LPDDR6 memory and PCIe Gen 7, facilitating high-bandwidth networking and efficient control-plane processing. This subsystem is engineered to drive accelerated AI capabilities, offering essential building blocks for cloud servers and data processing units (DPUs) that can offload tasks and optimize overall performance. By leveraging nearly a decade of Neoverse deployment experience, Arm aims to provide a streamlined path to custom silicon tailored for the demands of modern AI data centers.