Vera Rubin Accelerator Smashes Performance Expectations
Nvidia's Vera Rubin AI accelerator has delivered impressive results in recent tests, according to SemiAnalysis. The platform generated 39% more annual revenue and 42% more profit per gigawatt than the strongest GB300 setup.
SemiAnalysis credits Nvidia's integrated compute, networking, and interconnect design for Rubin's performance. This 'extreme co-design' of the Rubin GPU, Vera CPU, NVLink 6, ConnectX-9, BlueField-4, and Spectrum-6 networking stack is designed around agentic workloads involving long context and repeated turns.
The analysis highlights Rubin's performance-per-dollar advantage, with the platform delivering up to 67x the total throughput per Total Cost of Ownership (TCO) of GB300 in a 170 tokens-per-second (TPS) configuration. This is particularly important as AI data centers are increasingly constrained by available electricity rather than simply the number of accelerators that can be purchased.