The AmpereOnes promise two to three times the power savings over x86 with two to three times the cores. Ampere has announced it has begun shipping its next-generation AmpereOne processor, a server chip with up to 192 cores and special instructions aimed at AI processing. It is also the first generation of chips from the company using homegrown cores rather than cores licensed from Arm. Among the features of these new cores is support for bfloat16, the popular instruction set used in AI training and inferencing. “AI is a big piece [of the processor] because you need more compute power,” said Jeff Wittich, chief products officer for Ampere. ”AI inferencing is one of the big workloads that is driving the need for more and more compute, whether it’s in your big hyperscale data centers or the need for more compute performance out at the edge.” The company claims AmpereOne is two to three times more power efficient than Intel and AMD server processors. Wittich said customers buying products from Ampere are struggling to build out more capacity because they lack access to additional power or the ability to add new data centers. “So at the end of the day, all compute capacity in the cloud is constrained by the amount of available power. And so we need to deliver more compute in the same footprint,” he said. AmpereOne uses a chiplet design to break up the CPU into multiple chips rather than one monolithic piece of silicon. They are linked together by a mesh connecting the cores that acts as a ”sophisticated traffic cop,” as Wittich put it. It foresees where bottlenecks might occur and routes traffic to prevent them. All those cores might lead to contention for shared resources like memory or the system-level cache. So AmpereOne has a quality-of-service enforcement feature that restricts the amount of memory bandwidth that a specific processor user is taking up. Or it can prioritize a specific process to ensure it gets ample memory bandwidth. AmpereOne also supports nested virtualization, which runs VMs inside VMs—something customers asked for, Wittich said. While the Ampere line of processors is primarily aimed at cloud service providers, Wittich said enterprise customers use it for cloud infrastructure within their data centers running platform as a service and containerized apps. AmpereOne CPUs are shipping today, manufactured by TSMC and built on a 5nm process. Wittich said cloud providers and OEM licensees using the processors will make announcements about it when they are ready to launch their services or platforms. Related content news High-bandwidth memory nearly sold out until 2026 While it might be tempting to blame Nvidia for the shortage of HBM, it’s not alone in driving high-performance computing and demand for the memory HPC requires. By Andy Patrizio May 13, 2024 3 mins CPUs and Processors High-Performance Computing Data Center news CHIPS Act to fund $285 million for semiconductor digital twins Plans call for building an institute to develop digital twins for semiconductor manufacturing and share resources among chip developers. By Andy Patrizio May 10, 2024 3 mins CPUs and Processors Data Center news HPE launches storage system for HPC and AI clusters The HPE Cray Storage Systems C500 is tuned to avoid I/O bottlenecks and offers a lower entry price than Cray systems designed for top supercomputers. By Andy Patrizio May 07, 2024 3 mins Supercomputers Enterprise Storage Data Center news Lenovo ships all-AMD AI systems New systems are designed to support generative AI and on-prem Azure. By Andy Patrizio Apr 30, 2024 3 mins CPUs and Processors Data Center PODCASTS VIDEOS RESOURCES EVENTS NEWSLETTERS Newsletter Promo Module Test Description for newsletter promo module. Please enter a valid email address Subscribe