“
Vlad Cherevko I have been interested in all kinds of electronics and technologies since 2004. I like to play computer games, and I understand the work of different gadgets. I regularly monitor the news of the technology in the world and write materials about it.
The new platform includes R100 graphic processors, central processors, NVLink switches, silicon-photon processors and other components, which indicates a complete restructuring of the technical stack.
All Rubin chips are created for the 3nm TSMC N3P technical process and use CowOS-L housing. For the first time in NVIDIA history, the design of chiplets, as well as a new 4X targeting grid, exceeding 3.3x in the previous Blackwell architecture. Rubin R100 graphic processors will receive HBM4 memory support, which has a higher bandwidth compared to HBM3E.
The Rubin platform is focused on highly productive calculations, in particular for artificial intelligence systems. NVIDIA conducts a TSMC factories checking the thermal characteristics of architecture, energy consumption and efficiency of inter -protach joints. Rubin chips are expected to begin in 2026, and the extended version of Rubin Ultra in 2027.
Jensen Juang noted that the main purpose of Taiwan’s trip was to assess the progress of Rubin architecture at TSMC capacity. According to him, the new platform is “very advanced” and will become the basis for the next generation of NVIDIA server decisions.
”, – WRITE: mezha.media
Vlad Cherevko I have been interested in all kinds of electronics and technologies since 2004. I like to play computer games, and I understand the work of different gadgets. I regularly monitor the news of the technology in the world and write materials about it.
The new platform includes R100 graphic processors, central processors, NVLink switches, silicon-photon processors and other components, which indicates a complete restructuring of the technical stack.
All Rubin chips are created for the 3nm TSMC N3P technical process and use CowOS-L housing. For the first time in NVIDIA history, the design of chiplets, as well as a new 4X targeting grid, exceeding 3.3x in the previous Blackwell architecture. Rubin R100 graphic processors will receive HBM4 memory support, which has a higher bandwidth compared to HBM3E.
The Rubin platform is focused on highly productive calculations, in particular for artificial intelligence systems. NVIDIA conducts a TSMC factories checking the thermal characteristics of architecture, energy consumption and efficiency of inter -protach joints. Rubin chips are expected to begin in 2026, and the extended version of Rubin Ultra in 2027.
Jensen Juang noted that the main purpose of Taiwan’s trip was to assess the progress of Rubin architecture at TSMC capacity. According to him, the new platform is “very advanced” and will become the basis for the next generation of NVIDIA server decisions.