Data centers are facing a growing problem: the rapid advance of AI and its increasing energy requirements. Ampere Computing, a server CPU designer, aims to address this issue with its AmpereOne chip family and its partnership with Qualcomm Technologies. The new Ampere CPU is expected to provide 40% more performance than any CPU currently on the market. Ampere CEO Renee James emphasized that the power requirements and energy challenge of AI have brought the company’s focus on performance and efficiency into sharper focus.
Ampere is collaborating with Qualcomm Technologies to develop a joint solution for AI inferencing using Qualcomm’s Cloud AI 100 inference solutions and Ampere CPUs. This collaboration aims to create more efficient solutions for data centers, as Jeff Wittich, Ampere’s chief product officer, explained. He highlighted the need to build broad horizontal solutions involving ecosystem partners to address power consumption issues in data centers.
Ampere’s vision for “AI Compute” encompasses a range of workloads, from traditional cloud native applications to AI. The company’s CPUs are designed to run various workloads, including data processing, web serving, media delivery, and AI integrated with cloud native applications. Ampere has an ambitious roadmap for CPUs for the data center, including the upcoming AmpereOne platform with a 256 core CPU.
Ampere’s partnership with Qualcomm also includes the development of a joint solution for efficient CPUs and high-performance AI accelerators. This joint solution, known as a super micro server, aims to optimize performance for AI inferencing tasks, especially on large models. Ampere is working closely with Qualcomm to create an innovative solution that will be easy for customers to adopt.
In terms of data center energy efficiency, Ampere’s CPUs offer significant advantages. The company’s 192-core Ampere Altra CPU running Llama 3 on Oracle Cloud delivers the same performance as an Nvidia A10 GPU paired with an x86 CPU, while using only a third of the power. Ampere’s CPUs also outperform AMD’s Genoa CPUs, providing up to 50% more performance per watt.
Ampere’s commitment to flexibility and incorporating customer IP into future CPUs is demonstrated by its participation in the UCIe working group as part of the AI Platform Alliance. The company aims to leverage open interface technology to enable the incorporation of other customer IP into its CPUs.
Overall, Ampere Computing is positioning itself as the go-to technology for the AI era. Its focus on performance, efficiency, and collaboration with industry partners like Qualcomm is driving innovation in data center CPUs and addressing the energy challenges posed by AI. With its upcoming AmpereOne platform and the promise of more efficient solutions, Ampere is poised to make a significant impact in the data center industry.