Schneider Electric publishes new white paper on liquid cooling for AI data centers

Schneider Electric heeft een nieuwe whitepaper uitgebracht: “Navigating Liquid Cooling Architectures for Data Centers with AI Workloads”. De whitepaper bevat een diepgaand onderzoek naar technologie├źn voor vloeistofkoeling en hun toepassingen in moderne datacenters, vooral in datacenters met high density AI-work loads.

The demand for AI is increasing. As a result, data centers generate significant heat, especially data centers with AI servers and accelerators used for training large language models (LLMs) and inference workloads. This heat generation increases the need for liquid cooling to maintain optimal performance, durability and reliability.

Schneider Electric’s latest white paper guides data center administrators and IT managers through the complexities of liquid cooling and provides clear answers to critical questions about system design, implementation and use.


Understanding architectures for liquid cooling

In the white paper, authors Paul Lin, Robert Bunger and Victor Avelar identify two main categories of liquid cooling for AI servers: direct-to-chip and immersion cooling. They describe the components and functions of a coolant distribution unit (CDU), which are essential for managing temperature, flow, pressure and heat exchange within a cooling system.

“AI workloads present unique cooling challenges that air cooling alone cannot solve,” said Robert Bunger, Innovation Product Owner, CTO Office, Data Center Segment at Schneider Electric. “Our white paper aims to make liquid cooling architectures understandable and provide data center managers with the knowledge to make informed decisions when deploying liquid cooling. Our goal is to provide data center professionals with practical insights to optimize their cooling systems. By understanding the limitations and benefits of each architecture, operators can improve the performance and efficiency of their data centers.”

The white paper outlines three key elements of liquid cooling architectures:

1. Capturing heat within the server: using a liquid medium (e.g., dielectric oil, water) to absorb heat from IT components.
2. Type of CDU: selecting the appropriate CDU based on heat exchange methods (liquid-to-liquid, liquid-to-air) and placement (rack-mounted or, floor-mounted).
3. Heat dissipation method: determine how heat can be effectively dissipated to the outside, through existing systems or special setups.

Choosing the right architecture
The whitepaper describes six common architectures for liquid cooling, combining different CDU types and heat dissipation methods. In addition, the whitepaper provides guidelines for choosing the best option, considering factors such as existing infrastructure, size of implementation, speed and energy efficiency.

With increasing demand for AI processing capacity and rising thermal loads, liquid cooling is becoming a critical component in data center design. The white paper also addresses industry trends such as the need for increased energy efficiency, environmental regulatory compliance and the shift to sustainable business processes.

“As the need for advanced cooling solutions continues to grow due to AI, our white paper provides a valuable tool in dealing with these changes,” Bunger said. “We are committed to helping customers achieve their high-performance goals. At the same time, we help them improve sustainability and reliability.”

Reference designs for AI data centers

This white paper is particularly timely and relevant in light of Schneider Electric’s recent partnership with NVIDIA to optimize data center infrastructure for AI applications.

This collaboration has resulted in the first publicly available AI data center reference design, which leverages NVIDIA’s advanced AI technologies and Schneider Electric’s data center infrastructure expertise.

These reference designs set new standards for AI deployment and operation, providing data center managers with innovative solutions to efficiently manage high density AI workloads.

Read the white paper on their website

Lees meer

Gerelateerd berichten



Stay informed about news, events, updates through our newsletter