[HARDWARE]■ STORY TIMELINE

VERA RUBIN NVL72 CRUSHES BLACKWELL ON AI INFERENCE

AMD's Vera Rubin NVL72 accelerator delivers up to 7x better token throughput per megawatt than Nvidia's Blackwell on large language model inference, according to SemiAnalysis testing. The result exceeds Nvidia CEO Jensen Huang's stated 3x efficiency gains for models in the 1T-3T parameter range.

1 SOURCEFIRST SEEN SEP 15, 12:40 AM► READ THE ARTICLE

Techmeme+0m

Vera Rubin NVL72 inference tests show up to 7x better token throughput per MW vs. Blackwell on a 1.6T DeepSeek model, above Huang's 3x claim for 1T-3T LLMs (Bryan Shan/SemiAnalysis)

Bryan Shan / SemiAnalysis: Vera Rubin NVL72 inference tests show up to 7x better token throughput per MW vs. Blackwell o…

◄ BACK TO ARTICLE