As the data center boom continues, cooling solutions have a significant role to play. Data centers require cooling systems to keep chips running smoothly and prevent equipment failure. There are many different types of data center cooling systems that address different needs for data centers focused on running different kinds of chips.
Data centers managers continue to chase performance for their customers. To take advantage of the better performance, new business cases and new types of workloads that next-generation technology enables, data center managers need to maximize cooling performance. In many situations, including deploying artificial intelligence (AI) and machine learning (ML) applications, data centers have crossed a line where cooling high-powered equipment with air cooling is not sufficient.
When air cooling systems need to work overtime to maintain necessary operating temperatures in the face of growing demand, facilities can experience equipment failures, unplanned downtime and high energy costs. For many data centers, liquid cooling can offer better performance while saving energy and helping data centers operate more sustainably.
Liquid cooling is rapidly growing in popularity among data center managers and is required for many AI and ML applications. Liquid cooling uses precisely chilled liquid to capture and transport heat away from IT equipment. Liquid cooling is more efficient because liquid has a much greater heat transfer capacity than air and can be brought closer to the source of heat. Liquid cooling can help data center managers increase capacity, improve efficiency and reduce energy consumption of cooling systems. In 2023, the data center liquid cooling market was valued at about 3.2 billion dollars and is estimated to reach 16.4 billion dollars in 2032.
There are four main types of liquid cooling:
Liquid-to-Air
Liquid-to-air cooling systems bring liquid cooling to chips even within air-cooled facility infrastructure by circulating coolant within rack from a unit that sits beside a data center rack. This type of cooling does not require facility-level liquid cooling infrastructure which makes it ideal for older data centers trying to augment their current systems to accommodate next-generation chips. Liquid-to-air cooling is an important sustainability technology because it can be quickly deployed to cool the hottest parts of the data center with energy efficient liquid cooling.
Air-to-Liquid
Air-to-liquid cooling systems cool IT equipment within a rack using air and then cools the hot air with liquid lines before releasing it into the room. This makes the whole system more efficient because the entire room does not need to be cooled to the temperature required for chips. Air-to-liquid cooling requires facility liquid cooling lines but brings efficient liquid cooling to data centers that are running chips meant to be air cooled.
Direct-to-Chip
In direct-to-chip cooling, engineers place a heat sink directly on server boards enabling liquid lines to cool the chip directly. These systems must be carefully engineered due to their proximity to IT systems. They require an external coolant distribution unit (CDU) to precisely regulate the flow rate and temperature of liquid. CDUs can be known as the “heart and brain” of liquid cooling systems.
Immersion
Immersion cooling is the most efficient form of cooling, with servers immersed in cooled dielectric liquid. This system both cools and shields chips from dust and humidity. It provides the highest heat removal capacity across wider operating temperatures and offers quieter operation. The modularity of this kind of cooling system allows flexibility in IT advancement adoption.
nVent offers solutions to support all different types of data center cooling systems—from helping air cooled systems become more efficient to using immersion to cool the cutting edge of AI chips. Our experts can help determine the best solutions for your data center.