Author: By Gregory Alexander, Senior Engineer II, Thermal at Motivair by Schneider Electric
Summary
- PG25 is a common direct-to-chip coolant, balancing performance, protection, and ease of maintenance.
- Coolant options involve tradeoffs: deionized water offers higher performance, but at the cost of increased maintenance.
- Regular testing and fluid management help prevent corrosion, biofouling, contamination, and other risks.
- As AI drives higher chip power and rack densities, proper coolant selection and management become increasingly critical.
There’s little question that AI data centers require liquid cooling, as AI workloads drive ever-increasing rack densities. So, coolant selection, fluid management, and cooling system reliability all become paramount. It’s natural, then, that questions arise about the most common cooling fluid, PG25, as well as what ASHRAE and the Open Compute Project (OCP) recommend.
In this blog post, I’ll tackle some of the most common coolant questions we hear from hyperscalers, colocation providers, neoclouds, and other data center ecosystem players.
Let’s start with a basic question:
What coolant should I use for data center liquid cooling?
Companies have three main fluid options for liquid data center cooling: dielectrics, deionized water plus additives, and PG25. Dielectrics are used mainly for immersion cooling, which really can’t keep up with the increasing cooling requirements of AI chips.
Deionized (DI) water performs well, but the maintenance required to keep the fluid clean can significantly increase costs. The industry has thus settled on PG25, which performs nearly as well as DI water but with far less fluid maintenance.
What is PG25 coolant for data centers?
PG25 is a mix of 75% water and 25% propylene glycol (PG). At a 25% concentration, PG inhibits bio growth. Interestingly, at 10%, algae and other biomatter tend to feed on PG, but 25% is the sweet spot at which the solution becomes biostatic, meaning it inhibits growth.
PG25 also generally includes other additives to further prevent bacterial growth and corrosion in the copper within coolant distribution units (CDUs) and cold plates that carry the fluid. Each fluid manufacturer has its own secret sauce in terms of the exact formulation, but the main goals are to prevent corrosion and bio-growth. That’s important because the cold plates have tight fin spacing, around 50 microns wide today with future cold plates possibly going even tighter. Any growth on those plates can affect cooling performance, cause thermal throttling, and potentially damage the expensive IT equipment. So, it’s imperative to prevent growth of algae or any other biological fouling. As an added benefit, PG25 offers freeze protection down to about -10°C (14°F).
What does ASHRAE recommend for liquid cooling fluids?
ASHRAE doesn’t generally make recommendations. It does write standards, but they’re not mandates. That said, they do have groups that look at issues like data center cooling, namely Technical Committee 9.9 (TC 9.9), for Mission Critical Facilities, Data Centers, Technology Spaces, and Electronic Equipment. Many of the players working within ASHRAE to guide data center design are involved with TC 9.9, as well as ANSI/ASHRAE Standard 90.4, which is the main standard covering data center efficiency. While that standard doesn’t recommend any specific fluid, it mainly focuses on PG25 and DI water.
What is the OCP coolant standard?
Unlike ASHRAE, OCP does make recommendations, with the goal of providing a framework for the data center industry to deploy liquid cooling at scale quickly and safely. OCP recommends PG25 because it solves the problems previously identified in this blog post.
How often should I test coolant in a data center CDU loop?
This question gets to cooling chemistry, which does affect liquid cooling reliability. If your fluid performance starts to degrade, it may lead to issues including biological growth, corrosion, scale formation, and more – all of which can ultimately lead to IT equipment problems. Testing can help you detect fluid failures and correct them before algae has a chance to grow.
Think about testing with a tiered approach. For a brand new, just-commissioned system, you want to test daily or even multiple times per day. The idea is to catch any initial changes in fluid chemistry that may have a detrimental effect, such as a drop in the PG concentration. It’s not uncommon for concentrations to change significantly as the fluid makes its first pass through the CDU as it mixes with residual flushing DI water that may not have been removed fully. It’s not difficult to boost the concentration, such as by adding a PG50 solution, but you need to test to determine if that’s required.
After a week or so, or whenever your testing shows the fluid concentration has stabilized, you can shift to weekly sampling. Now you’re starting to run jobs on your IT equipment, so you need to monitor how fluid concentrations change under load.
Here again, once you see stability, you can shift to a less-frequent schedule, perhaps monthly. The exact frequency depends on the facility and the balance you want to strike between risk and cost.
When should I use propylene glycol versus ethylene glycol?
The short answer here is, in the vast majority of cases, use PG25. Ethylene glycol (EG) has somewhat better thermal performance and lower viscosity than PG, but it’s far more toxic to humans. For that reason, it’s seldom used in data centers. It’s more often used for freeze protection in automotive applications, such as your car’s cooling system, or in other closed-loop HVAC applications where humans generally don’t need to touch it. In an IT environment, you will need to disconnect the fluid loops each time you swap out a server blade, which happens frequently during server maintenance. During servicing, the quick disconnects (QDs) often have a slightly wetted disconnect rather than a 100% dry break. PG25 is perfectly safe in that situation; EG is not.
Coolant selection for data center liquid cooling: From selection to management
As AI pushes rack densities and chip power higher, liquid cooling strategies are becoming increasingly important, and choosing the right coolant is a key part of that strategy. The right fluid, combined with regular testing, proactive management, and the right architecture, can help improve cooling performance and reliability while protecting AI infrastructure and major investments over the long term.
Visit our liquid cooling resource page to learn more about optimizing liquid cooling performance and developing a proactive fluid management strategy, or feel free to contact us.
Add a comment