Most enterprise data centers have gotten by just fine by air cooling their CPUs and servers, but AI is forcing IT to consider new cooling types, including liquid cooling. With power draws for CPUs hitting 400 watts and GPUs hitting 700 watts, air cooling is simply not sufficient for the extremely hot-running, power-hungry chips used in AI.
Today, 22% of data centers are using liquid cooling, according to IDC numbers. A decade of growth is anticipated, according to Global Market Insights, which projects the global data center liquid cooling market will grow at a compound annual growth rate of 15% between 2023 and 2032. (See also: Pros and cons of air, liquid and geothermal cooling systems)
A new resource from Schneider Electric is aimed at helping enterprises evaluate liquid cooling technologies and their applications in modern data centers, particularly those handling high-density AI workloads.
Network World
While Schneider Electric has a vested interest in liquid cooling, since it sells the equipment, it strove to be vendor- and product-neutral in its advice, said Robert Bunger, innovation product owner for the data center segment at Schneider Electric and a co-author of the publication, “Navigating Liquid Cooling Architectures for Data Centers with AI Workloads.”
Schneider Electric’s paper covers two types of liquid cooling: direct-to-chip, the overwhelming choice for those who have adopted liquid cooling; and immersion, which is still largely a fringe solution. Immersion cooling involves dunking very expensive electronics in a liquid bath, but it is a non-conductive liquid, so it’s safe to use with electronics.
The Schneider Electric paper addresses three main topics:
- 1. Heat capture within the server: How to use a liquid medium (such as dielectric oil, or water) to absorb heat from IT components.
- 2. Types of coolant distribution units (CDU): There are multiple ways to move liquid around server racks. Schneider’s paper covers how to select the appropriate CDU based on heat exchange methods (liquid-to-air, liquid-to-liquid) and form factors (rack-mounted, floor-mounted).
- 3. Heat rejection methods: Determining how to effectively transfer heat to the outdoors, either through existing facility systems or dedicated setups.
In addition, Schneider Electric details six common liquid cooling architectures, combining different CDU types and heat rejection methods, and it provides guidance on selecting the best option based on factors such as existing infrastructure, deployment size, speed, and energy efficiency.
Bunger said there is still a fear of leaks when companies consider liquid cooling, and it’s okay to have reasonable anxiety about that. “But I think people still kind of forget that there is water everywhere in the data center, with chilled water systems and everything like this. And this is just a little bit of an extension of that. So, you know, it’s not as scary as people think,” he said.
Migrating to liquid cooling is not something that is casually done and typically requires an experienced IT integrator, Bunger noted. “If it’s a first-time deployment for somebody doing liquid cooling, for sure, [I would recommend] to lean on the IT integrator,” he said.
Read more about liquid cooling:
- Data centers warm up to liquid cooling
- Is immersion cooling ready for mainstream?
- Data center cooling: Pros and cons of air, liquid and geothermal systems
- Intel and Vertiv partner to liquid cool AI processors
- Accelsius offers liquid cooling without a data center retrofit
- ZutaCore launches liquid cooling for advanced Nvidia chips
- Supermicro has a new liquid-cooled server for AI
- LiquidStack expands into single-phase liquid cooling