Search
Nvidia has revealed that its cutting-edge Vera Rubin microarchitecture is currently in the research and development phase, with an official launch slated for 2026. Notably, the Rubin CPX variant is specifically engineered to tackle AI workloads that demand extensive context windows, making it an ideal choice for scenarios that involve processing over a million tokens. By the end of 2026, it is anticipated that the Vera Rubin NDL144CPX-class GPU will hit the market, boasting impressive AI performance and robust memory configurations. Furthermore, Nvidia is set to introduce a reference design for gigascale data centers, aimed at facilitating the construction of AI factories. Earlier, in Nvidia's MLPerf inference tests, the Blackwell GPU achieved a new benchmark, with its 'decoupled service' technology significantly boosting hardware performance.
-
C114 Communication Network -
Communication Home
