Ampere Computing, the emerging ARM architecture-based chip company, has unveiled its next generation of processors, which will reach 512 cores in the coming years and include an AI processing unit.
A glimpse into the future of computing
The new processor, named Aurora, won’t be available until 2026. Currently, the company sells the AmpereOne with 192 cores, and the AmpereOne MX with 256 cores is expected for next year. Aurora will represent a significant leap in this product line, also incorporating Ampere’s AI processor.
Jeff Wittich, Ampere Computing’s product director, highlighted: “For the first time, using our own integrated Ampere AI IP directly in the SOC via our interconnect and also high-bandwidth memory connected to this platform, we are really addressing those critical AI use cases, starting with inference, but also scaling towards training.”
Key features of the AmpereOne Aurora
– Up to 512 Ampere cores
– Scalable AmpereOne Mesh
– Integrated Ampere AI IP directly in the silicon hardware
– High-bandwidth memory
Aurora promises to offer three times the performance per rack compared to current AmpereOne processors. Additionally, Ampere claims the chip will be compatible with all types of enterprise applications, not just the cloud.
Focus on efficiency and adaptability
A highlight of Aurora is its ability to be air-cooled, allowing for implementation in any existing data center without the need for adaptations to support liquid cooling. Wittich emphasized the energy efficiency of the AmpereOne product line, noting that it is more energy-efficient for current data centers.
Impact on the AI market
Aurora is positioned as a powerful solution for AI workloads like Retrieval-Augmented Generation (RAG) and vector databases. Integrating AI capabilities directly into the chip could be a turning point in the development of computing systems for artificial intelligence.
Future outlook
Although Aurora won’t be commercially available until 2026, its announcement sends a clear signal about the direction Ampere Computing is taking in the competitive market of processors for data centers and AI applications. The company seems to be betting on a strategy that combines high performance, energy efficiency, and versatility in implementation.
Source: Ampere (https://amperecomputing.com/blogs/introducing-ampereone-aurora)