Shenzhen Zhilian Weichuang Technology Co., Ltd.

Internal turmoil did not disrupt plans; NVIDIA continues to launch new chips, showcasing Rubin HPC

Publication Date: 2026-11-18 Source:AI ji wei Key words: NVIDIA GPU chip View count:

Although NVIDIA has been deeply troubled by management changes, it continues to show ambition in the field of high-performance computing technology. At the Supercomputing 2026 (SC26) conference, it announced a new GPU chip, Rubin HPC, designed to accelerate work in the fields of science, engineering, and analytics, becoming a powerful tool to aid scientific advancement. This demonstrates its ambition to integrate AI with traditional high-performance computing (HPC), with an expected release in 2028.


Major AI companies continue to accelerate scientific progress with chips and language models. Google is advancing computations of protein structures with the TPU and Gemini model, while NVIDIA demonstrates its position in AI chips and also launches new products designed specifically for science and engineering, showcasing AI technology.


The Rubin HPC GPU is also a product of the Vera Rubin platform, equipped with over 50,000 CUDA cores, setting a record for the number of CUDA cores on a single GPU. It has sparse data acceleration capabilities and uses the seventh-generation NVIDIA NVLink interconnect technology to improve communication efficiency between GPUs. Rubin HPC accelerates simulation calculations, significantly shortening the time from exploration to scientific discovery and industrial design.


Ian Buck, Vice President of Hyperscale and High-Performance Computing at NVIDIA, pointed out that the application of AI agents in the scientific field means faster simulations lead to faster discoveries. Rubin HPC accelerates AI agent testing of concepts and guides simulations for subsequent steps, helping scientists and engineers explore more possibilities and break through product development bottlenecks earlier.