NVIDIA Introduces Vera Rubin Architecture at OCP Summit

During the Open Compute Project (OCP) Global Summit, NVIDIA presented new collaborations and contributions to create a common foundation for the next generation of AI infrastructure. The adoption of open standards accelerates the scalability and efficiency of AI factories, which are becoming larger and require more computing power. NVIDIA has contributed key designs for racks and computing systems to OCP via the open NVIDIA MGX architecture, providing system builders with a solid foundation for the development of AI factories.

Discover various announcements from NVIDIA and partners at the OCP event. This includes, among other things, open standards for AI factories, a new version of the open-source software component for the NVIDIA DSX AI factory, and an innovative 800V DC design from Visionbay for the Vera Rubin era. Additionally, NVIDIA and Cisco jointly presented the Ultra-Dense Small Form Factor Pluggable (USFP) standard, marking an important milestone for network infrastructure with 400Gb/s capacity. These developments are crucial for improving the scalability and power of AI infrastructures worldwide.

NVIDIA emphasized the importance of open infrastructures during a session led by Ian Buck, vice president of high-performance computing. The new MGX specifications enable a larger modular architecture that supports community standards and promotes the interoperability of different components. This facilitates not only assembly and testing but also service provision when scaling AI factories.

Furthermore, collaborations between NVIDIA, Meta, Google, and other players were discussed. This includes the publication of the Open Data Center Specification, which provides a blueprint for the design standards of AI factories. The specifications help manufacturers develop a standardized infrastructure that is essential for the growth of AI solutions. Through open standards and joint innovations, a sustainable foundation is laid for the accelerated development of AI factories capable of processing millions of decisions per second.

Read the full article from Nvidia.