512-core AmpereOne Aurora processor with custom AI engine and HBM memory support announced
Ampere launched the 512-core AmpereOne Aurora with AI engine and HBM memory for cloud AI computing.
Ampere Computing has introduced the 512-core AmpereOne Aurora processor, aimed at addressing the demands of cloud-native AI computing. The processor features an integrated AI accelerator and on-chip High Bandwidth Memory (HBM), providing notable efficiency and power.
The AmpereOne Aurora includes a scalable AmpereOne Mesh for seamless compute connections and custom AI IP cores, marking a first for Ampere products. The chip is designed to enhance workloads like RAG and vector databases, offering three times the performance per rack compared to prior AmpereOne CPUs.
Targeted at a variety of applications from public cloud to hyperscale data centers, the Aurora can be air-cooled, adding flexibility for deployment. Although it remains to be seen if Ampere can effectively compete with x86-64 chips from Intel and AMD, it aims to stand out with its focus on AI and cloud-native features.