Seven new chips are now fully operational as part of NVIDIA's newly revealed AI platform, the Vera Rubin. This initiative is designed to support the largest AI factories globally, offering a flexible infrastructure that enhances various stages of artificial intelligence development, including training and inference.
The platform features several advanced components, such as the NVIDIA Vera CPU, NVIDIA Rubin GPU, and the NVIDIA NVLink™ 6 Switch. According to Jensen Huang, NVIDIA's CEO, this launch represents a significant technological advancement, as the system combines seven innovative chips and five coordinated racks into a powerful AI supercomputer.
Industry leaders like Dario Amodei from Anthropic and Sam Altman from OpenAI have highlighted the platform’s potential to enhance AI capabilities significantly. With the Vera Rubin, organizations can expect improved model performance and efficiency while democratizing access to AI technology across various sectors.
The standout feature, the NVIDIA Vera Rubin NVL72 Rack, incorporates 72 Rubin GPUs and 36 Vera CPUs, aimed at delivering remarkable efficiency for complex model training, achieving performance increases of up to ten times.