Breaking the 1000-core barrier: Ampere unveils roadmap for 512-core CPU as it battles for cloud-native market – but is it too late for the sole Arm chip vendor?
Ampere Computing introduced the AmpereOne processor family in 2023, featuring a whopping 192 single-threaded cores, the highest number in the industry at the time.
Founded in 2017 by Renée James, former president of Intel, Ampere is pushing the boundaries even further with the announcement of the AmpereOne Aurora, a 512-core CPU designed to meet the evolving needs of cloud-native workloads.
Ampere has focused on developing innovative technologies for cloud infrastructures in recent years. As general and AI workloads converge in the cloud, the need for a platform that is efficient, air-cooled, and integrated with AI acceleration has become clear.
Tackling the Global Challenge of AI Power
AmpereOne Aurora integrates several Ampere innovations, including custom cores, a patented mesh, and die-to-die interconnections across chiplets.
By adding Ampere’s proprietary AI acceleration directly into the silicon hardware, the new processor aims to significantly improve AI Compute capabilities. It features up to 512 Ampere cores, delivering more than three times the performance of its predecessors.
The scalable AmpereOne mesh ensures seamless connectivity between different types of computers, while integrated Ampere AI IP and high-bandwidth memory further improve performance.
The new processor can scale across a range of AI inference and training use cases, and provides robust AI Compute capabilities for workloads such as RAG and vector databases. Ampere says it will deliver industry-leading per-rack performance for AI Compute, and its air-cooling capability enables deployment in any existing data center, including public clouds, enterprise settings, hyperscale data centers, and edge locations. This feature addresses the global AI power challenge by enabling efficient and versatile deployment.
Ampere’s collaborations with various partners and the AI Platform Alliance are focused on creating hardware solutions that deliver a full spectrum of AI Compute in an open and interoperable environment. Users can rely on accessible and affordable technologies to develop their own innovative AI products and services.
Despite these advances, the market for cloud-native solutions is highly competitive. With established players continually pushing the boundaries, some industry analysts may wonder whether Ampere’s efforts are too late to gain significant traction.