Trane Technologies and Eaton have developed the first-of-its-kind design solution that integrates thermal and electrical systems for AI data centers. Using NVIDIA DSX AI Factory Reference Design, the solution is expected to enhance energy efficiency by up to 15%, reduce installation costs by up to 30%, and save up to 80% of copper used.
According to the companies, the solution allows integrating power and cooling systems instead of using separate and manual solutions. The solution is designed to help data center owners speed up the deployment of high-powered AI infrastructure and make it less complex and resource-intensive.
Integration of power and cooling system:
The reference design enhances medium voltage architecture for high-powered density AI factories. It is incorporated into Trane® Continuum Rubin DSX and Eaton Beam Rubin DSX systems and is compatible with NVIDIA DSX systems.
Combining thermal and electrical systems in terms of the grid-to-chip approach allows the systems to exchange information and respond to changes dynamically.
“AI and high-performance computing are changing the requirements for data center technology, and customers are seeking solutions that can support them in meeting those requirements,” commented Mauro J. Atalla, Senior Vice President, Chief Technology and Sustainability Officer, Trane Technologies. “Through the combination of our advanced thermal management solutions with Eaton’s innovative power management solutions, we are providing a system design to help customers deploy faster and scale confidently.”
Designed for the next generation of AI infrastructure:
“We are improving the industry benchmark with respect to speed of deployment by maturing reference architectures into systems that can be deployed repeatedly,” noted Michael Regelski, Senior Vice President and Chief Technology Officer, Electrical Sector, Eaton. “With NVIDIA DSX, we are combining our medium voltage power systems and white space thermal management solutions with Trane’s advanced thermal management solution architecture to help customers deploy AI factory at scale,”
The design supports NVIDIA® Omniverse™ DSX Blueprint and is designed to give a more systematic way of planning electrical, thermal and digital control infrastructure. The design is also scalable in case of emerging liquid cooling and direct current technologies.
AI factories require a well-coordinated power, cooling, and computing infrastructure to function effectively. The companies are ensuring that their collaboration with the NVIDIA Omniverse DSX Blueprint is able to enable customers to simplify and fast-track the deployment of the next-gen AI data centers,” said Vladimir Troy, Vice President of AI Infrastructure at NVIDIA. “The new approach will facilitate the solid and scalable frameworks needed for enterprises to take full advantage of generative and reasoning AI.”
