OpenAI Researcher Praises AMD and Cerebras Joint AI Inference Solution, Estimates 5x Performance Per Watt Improvement

TechFlow reports, on July 29, OpenAI researcher Jeffrey Wang commented that the performance of the AI inference solution jointly developed by AMD and Cerebras was "incredible". The solution integrates AMD Helios racks and the Cerebras Wafer-Scale Engine, processing prompts and token generation stages respectively, and is estimated to increase tokens per second per watt by 5 times. This technical approach optimizes inference performance through heterogeneous computing, reflecting the continued focus of leading AI labs on reducing inference costs.