NVIDIA has unveiled a digital blueprint for building next-generation data centers using its Omniverse digital twin platform. The company demonstrated a fully operational data center based on the NVIDIA GB200 NVL72 liquid-cooled system, which includes 18 NVIDIA Grace CPUs and 36 NVIDIA Blackwell GPUs. The digital twin was created using software tools connected by Omniverse and the Cadence Reality digital twin platform. The technology allows engineers to test, optimize, and validate data center designs before building a physical system, improving efficiency and speed. Partners in the project include Ansys, Cadence, PATCH MANAGER, Schneider Electric, and Vertiv.
NVIDIA’s Innovative Approach to Data Center Design
NVIDIA, a company renowned for its advanced AI supercomputers, has recently unveiled a new approach to designing and simulating data centers. This process involves the use of a digital twin, a virtual replica of a physical system, to test and optimize the design before it is physically implemented. This innovative method is expected to significantly enhance the efficiency and accuracy of data center design and construction.
The company’s latest AI supercomputer, a large cluster based on the NVIDIA GB200 NVL72 liquid-cooled system, was showcased at the GTC event. This system comprises two racks, each containing 18 NVIDIA Grace CPUs and 36 NVIDIA Blackwell GPUs, interconnected by fourth-generation NVIDIA NVLink switches. This fully operational data center was demonstrated as a digital twin in NVIDIA Omniverse, a platform designed for building generative AI-enabled 3D pipelines, tools, applications, and services.
The Role of Digital Twins in Data Center Design
To expedite the construction of new data centers, NVIDIA first created a digital twin of the proposed design using software tools connected by Omniverse. This allowed engineers to unify and visualize multiple computer-aided design (CAD) datasets with full physical accuracy and photorealism. The digital twin was built using the Cadence Reality digital twin platform, which is powered by NVIDIA Omniverse APIs.
The digital twin approach allows for comprehensive testing, optimization, and validation of data center designs before the physical system is built. By visualizing the performance of the data center in the digital twin, teams can better optimize their designs and plan for potential scenarios. This method also enables the balancing of disparate sets of boundary conditions, such as cabling lengths, power, cooling, and space, in an integrated manner. This results in faster and more efficient deployment of clusters.
The Process of Creating a Digital Twin
The process of creating a digital twin for a data center involves several steps. Initially, technology company Kinetic Vision scanned the facility using the NavVis VLX wearable lidar scanner to produce highly accurate point cloud data and panorama photos. Following this, Prevu3D software was used to remove the existing clusters and convert the point cloud to a 3D mesh, providing a physically accurate 3D model of the facility.
Engineers then combined and visualized multiple CAD datasets with enhanced precision and realism using the Cadence Reality platform. The platform’s integration with Omniverse provided a powerful computing platform that enabled teams to develop OpenUSD-based 3D tools, workflows, and applications.
The Role of Various Tools and Partners in the Digital Twin Creation
The creation of the digital twin involved the use of various tools and collaborations with several partners. PATCH MANAGER was used to design the physical layout of the cluster and networking infrastructure, ensuring accurate cabling lengths and proper routing configuration. NVIDIA Air was also integrated into the process through Omniverse Cloud APIs.
The team used Cadence’s Reality Digital Twin solvers, accelerated by NVIDIA Modulus APIs and NVIDIA Grace Hopper, to simulate airflows and the performance of the new liquid-cooling systems. Partners such as Vertiv and Schneider Electric provided the integrated cooling systems in the GB200 trays, which were simulated and optimized using solutions from Ansys. This brought simulation data into the digital twin, further enhancing the accuracy and efficiency of the design process.
External Link: Click Here For More
