전세계의 데이터센터와 네트워크 인프라는 데이터 수요 증가와 더욱 어려워지는 전력 제약으로 인해 극심한 압박에 직면해 있다. 10억명의 5G 구독자, 약 150억 개의 IoT 연결 및 120ZB의 데이터 생성으로 인해 커넥티드 디바이스 및 데이터가 가속화됨에 따라, 더 많은 전력과 성능이 요구되고 있다.
Pre-integrated and validated configurations reduce development costs and shorten time to market
Save 80 years of engineering time with a single partner implementation
Data centers and network infrastructure around the world are under extreme pressure due to increasing data demand and increasingly difficult power constraints. With 1 billion 5G subscribers, approximately 15 billion IoT connections, and 120ZB of data being generated, the acceleration of connected devices and data is demanding more power and performance.
Arm announced on the 29th that it is launching the Arm Neoverse Compute Subsystems (CSS) to support design of custom silicon for time-to-market.
Arm is enabling the ecosystem through CSS to deliver custom designs at lower cost and risk than individual IPs, and with CSS N2, Arm is freeing up partner engineering resources to focus on SoC and system-level innovation.
Arm CSS N2, the first generation CSS product, is a Neoverse N2 platform with a pre-integrated and validated PPA-optimized configuration. CSS N2 is optimized for an advanced 5nm process and is available to partners as a custom compute subsystem, leveraging the performance-per-watt efficiency of the Neoverse N2 platform to accelerate time to market. This gives partners the freedom to implement new levels of innovation, either as-is or with further customization, giving them the opportunity to differentiate in areas such as memory, input/output (IO), acceleration, physical topology, and more – all in the pursuit of custom silicon solutions optimized for their workloads.
For one partner, CSS helped them take a project from ideation to 100+ cores running Linux in just 13 months, while another partner saved 80 years of engineering time with CSS. CSS N2 also supports integration of specialized domain-specific accelerators, including AI, and delivers industry-leading performance-per-watt for cloud-native workloads.
As SoC design costs and complexity increase, achieving sustainability is a key goal for many companies. In response, innovative companies in the industry are responding by offering efficient and specialized processing.
The Arm ecosystem says that to fully leverage the benefits of specialized processing, an architecture is needed that provides the flexibility to innovate and implement diverse and differentiated solutions while maintaining access to a robust software ecosystem. Flexibility and specialization have always been core tenets of Arm and the guiding philosophy of Arm Neoverse, and are realized through decades of investment in software and curation of the Arm architecture.
Arm’s partners have leveraged this unique balance between customization and standardization to achieve numerous industry firsts, including the first CPU with over 1 terabyte per second of aggregate memory bandwidth, the first CPU with over 100 cores per die, and the first CPU to bring DDR5 and PCIe Gen5 to market. These innovations are accelerating every major growth and momentum from wireless networking to the cloud.
CSS is another innovation in how Arm delivers the Neoverse platform, enabling more partners to leverage custom silicon by lowering design costs and accelerating time to market. Arm CSS marks the beginning of a new era of custom silicon for infrastructure, and represents an extension of the Neoverse portfolio with a trusted compute subsystem that allows the Arm ecosystem to focus on differentiation and innovation.
Today’s infrastructure is being custom-built, said Mohamed Awad, senior vice president and general manager of the Infrastructure Business Unit at Arm. “To ease the burden on future infrastructure, we need to build specialized processing built on efficient, scalable multicore compute, backed by a robust software ecosystem,” Awad said. “Arm is committed to providing a scalable and efficient compute foundation, enabling our partners to focus their limited resources on their core differentiation.”