ON JANUARY 6, REUTERS REPORTED THAT TODAY, AT THE CES CONSUMER ELECTRONIC EXHIBITION IN LAS VEGASNvidia CEO Jen-Hsun HuangIn the keynote address, it was announced that the company's next-generation chip, Rubin, was in the "full production" phase。

Wong In-hoon said that the Rubin chip would be officially launched later this year. When running chat robots and other AI applications, this chip is five times more computing performance than the previous generation。
The current Vera Rubin platform consists of six independent British Wida chips, of which the flagship equipment consists of 72 graphical processing units (GPU) and 36 new central processors (CPUs). These chips can be connected to a cluster of over 1,000 Rubin chips。
ALTHOUGH THE NUMBER OF TRANSISTORS IS ONLY 1.6 TIMES GREATER THAN THE NUMBER OF PREVIOUS GENERATIONS, THROUGH THE ADOPTION OF A PROPRIETARY DATA FORMAT, THE PERFORMANCE OF THE UNITED KINGDOM OF GREAT BRITAIN AND NORTHERN IRELAND HAS LEAPFROGGED. SENIOR YVETTE CONFIRMED TO REUTERS THAT THE CHIPS WERE CURRENTLY BEING TESTED IN THEIR LABORATORIES BY AI CLIENTS。
In terms of technical details, Hoang In-hoon highlighted two key innovations:
Context Memory: The aim is to help chat robots respond more quickly to long conversations initiated by millions of users at the same time, and to upgrade and develop processing capabilities
Co-packaged Optics: A new generation of network switches is key to connecting thousands of machines and will compete with the same products as Broadcom and Cisco。
In the area of auto-driving, Ying Weidar announced that the software called Alpamayo and its training data would be published more widely. The software helps auto-driving vehicles to make path decisions and provides engineers with a basis for decision-making。
Huang In-hoon said: "We are not only open-source models, but also open-source data for training these models, because only then can you truly trust the source of the models. I'm sorry
It was reported that, while Young Weida still dominates the AI Model Training Market, at the reasoning end (delivery of model results to end users) there is increasing competition from traditional competitors such as AMD and customers such as Google。
Last December, Inweida absorbed the talent and chip technology of the original company Groq, which included helping Google design AI ChipsIn response to the threat of companies like Google United Meta trying to undermine the ruling position of the British International。