NVIDIA GB200 NVL72 design contributions and NVIDIA Spectrum-X help accelerate the next industrial revolution
OCP Global Summit — To advance the development of open, efficient, and scalable data center technologies, NVIDIA today announced the foundational elements of the NVIDIA Blackwell Accelerated Computing Platform design to the Open Computing Project (OCP) and the NVIDIA We are pleased to announce the expansion of Spectrum-X™. Support for OCP standard.
At this year’s OCP Global Summit, NVIDIA shares key parts of the electromechanical design of the NVIDIA GB200 NVL72 system with the OCP community, including rack architecture, compute and switch tray mechanics, liquid cooling and thermal environment specifications, and NVIDIA NVLink I will. ™ Cable Cartridge Volumetric—Supports higher computing density and network bandwidth.
NVIDIA has already made several official contributions to OCP across multiple hardware generations, including the NVIDIA HGX™ H100 baseboard design specification, which brings the widest selection of products from the world’s computer manufacturers to the ecosystem. and help expand the adoption of AI.
Additionally, by extending the NVIDIA Spectrum-X Ethernet networking platform to specifications developed by the OCP community, enterprises can preserve investment and maintain software consistency while improving the performance of AI factories deploying OCP-certified equipment. You can maximize your potential.
NVIDIA Founder and CEO Jensen Huang said: “Building on a decade of collaboration with OCP, NVIDIA is collaborating with industry leaders to shape specifications and designs that can be widely adopted across data centers.” We are helping organizations harness the full potential of accelerated computing and build the AI factories of the future.”
High-speed computing platform for the next industrial revolution
NVIDIA’s accelerated computing platform is designed to power a new era of AI.
The GB200 NVL72 is based on the NVIDIA MGX™ modular architecture, allowing computer manufacturers to quickly and cost-effectively build massive data center infrastructure designs.
The liquid cooling system connects 36 NVIDIA Grace™ CPUs and 72 NVIDIA Blackwell GPUs in a rack-scale design. Featuring a 72-GPU NVIDIA NVLink domain that acts as a single large-scale GPU for real-time, trillion-parameter, large-scale language model inference that is 30x faster than NVIDIA H100 Tensor Core GPUs.
The NVIDIA Spectrum-X Ethernet networking platform, including the next-generation NVIDIA ConnectX-8 SuperNIC™, supports OCP’s Switch Abstraction Interface (SAI) and Software for Open Networking in the Cloud (SONiC) standards. This enables customers to use Spectrum-X’s adaptive routing and telemetry-based congestion control to accelerate Ethernet performance for scale-out AI infrastructure.
ConnectX-8 SuperNIC features accelerated networking with speeds up to 800Gb/s and a programmable packet processing engine optimized for large-scale AI workloads. ConnectX-8 SuperNIC for OCP 3.0 will be available next year, allowing organizations to build flexible networks.
Data center critical infrastructure
As the world moves from general-purpose computing to high-speed AI computing, data center infrastructures are becoming increasingly complex. To simplify the development process, NVIDIA works closely with more than 40 global electronics manufacturers that provide key components for building AI factories.
Additionally, a wide range of partners are innovating and building on the Blackwell platform, including Meta, which will provide OCP with its Catalina AI rack architecture based on the GB200 NVL72. This gives computer manufacturers flexible options to build systems with high computing density and meet growing data center performance and energy efficiency needs.
“NVIDIA has made significant contributions to open computing standards over the years, including with our high-performance computing platform that has been the foundation of our Grand Teton server for the past two years,” said Yee Jiun Song, vice president of engineering at Meta. says. “NVIDIA’s latest contributions in rack design and modular architecture will help speed the development and implementation of AI infrastructure across the industry as we move to meet the growing computational demands of large-scale artificial intelligence.”
Learn more about NVIDIA’s contributions to open computing projects at the 2024 OCP Global Summit, October 15-17 at the San Jose Convention Center.
