Liquid cooling is becoming increasingly important at data centers all over the world, thanks to the rise of AI and high performance computing. As computers get ever smarter, and their capabilities continue to grow, so too does the amount of heat being generated by the equipment that powers them.
Liquid cooling is now essential to the thermal management of many data centers, and as HPC continues to advance so will the importance of the most powerful and efficient cooling systems. There are several different forms of liquid cooling that data centers can choose from, each of which comes with its own distinct pros and cons.
In this guide, we’ll talk you through direct-to-chip cooling, an advanced cooling technique that uses the high thermal transfer properties of liquid to reduce the heat generated by individual processor chips.
How does liquid cooling work in data centers?
Liquid cooling fulfils a role very similar to that of traditional air cooling systems, but it does so in a way that’s far more efficient… and far more effective.
With liquid cooling, temperatures are reduced using the higher thermal transfer properties of water or other dielectric fluids. This newer method of cooling is 3,000 times more powerful than the air cooling methods data centers have been relying on for decades. It’s also more environmentally friendly, because it uses less power to do the same job.
There are several different forms of liquid cooling that are now on the rise in data centers, but the premise of them all is the same. Liquid is used to reduce the heat being generated by equipment and maintain a consistent temperature no matter how hard systems are asked to work.
Let’s explore the options in more detail.
Cooling for high performance computing (HPC): What are the options?
Data centers now have several options to choose from when considering how best to keep temperatures at optimal levels throughout their facilities.
Liquid cooling is of course a huge talking point at the moment, and there are several different forms of liquid cooling that can be used. Direct-to-chip cooling is becoming ever more popular in HPC, but like anything it does have some drawbacks to be aware of.
Introducing direct-to-chip cooling
In direct-to-chip cooling, liquid coolant is applied directly to the processors and other components that generate excess heat. This direct application is behind the efficiency improvements seen in liquid coolings vs the likes of air cooling.
Liquid is incredibly effective at reducing the temperatures of heat-generating components, allowing them to work at full capacity with no concerns about temperature-induced outages.
How do direct-to-chip cooling systems work?
Direct-to-chip liquid cooling involves the use of a number of core elements. The systems work in a way that’s deceptively simple. Using liquid coolant, heat generated by CPUs or other components is moved to a cold plate and then carried away to be released elsewhere.
Cold plates are fitted with internal channels for coolant to move through. The plates fit to the chips, allowing them to absorb heat directly. In such systems, liquid circulates in a continuous loop, removing heat as it does so.
Heat absorbed by the coolant liquid is then transferred by another key part of the system: the heat exchanger. The heat exchanger is used to take that heat to an external source, where it can be removed. Once that process is complete, the cooled liquid is ready to be used again, and so the cycle continues.
There’s another, slightly different form of direct-to-chip cooling to be aware of too: two-phase cooling. Both single and two-phase cooling offer efficient ways of absorbing heat, but they work in slightly different ways. Let’s take a closer look at the two types.
Single-phase direct-to-chip cooling
In single phase direct-to-chip cooling, a cold plate is used to move heat away from the components that generate it, such as CPUs and GPUs. The coolant absorbs heat and moves through the cooling system, until it reaches the heat exchanger. This takes heat to an external device, where it can be released.
Two-phase direct-to-chip cooling
If two-phase direct-to-chip cooling is used, liquid moves to evaporators, where it’s boiled using the heat generated by equipment. Excess heat can then be removed using evaporation, with the gas being taken away to be released away from any working equipment.
Why choose direct-to-chip cooling
There are many advantages to direct-to-chip cooling, particularly when it’s compared to traditional methods of cooling like air cooling. Here are some of the key reasons why data center operators are investing in direct-to-chip cooling at the moment.
Efficiency: Direct-to-chip cooling is one of the most efficient methods of cooling. In data centers that house HPC equipment generating vast amounts of heat, this efficiency is essential.
Energy consumption: Direct-to-chip cooling uses less energy than air cooling, which translates to a reduced carbon footprint and lower running costs.
No wastage: This form of cooling targets the components that need to be cooled directly, reducing the wastage seen in other forms of cooling system. It’s far more effective than air cooling, for example, where much of the cooled air is wasted as it moves through the data center.
Performance and reliability: If components are cooled via direct-to-chip cooling, there’s a far lower chance of overheating. This means reliability and performance benefits. It can even result in lower maintenance bills as components are always kept at optimum temperatures.
Space-saving: Air cooling systems take up huge amounts of space in data centers. Direct-to-chip cooling, on the other hand, is far more practical in terms of the amount of space needed to accommodate it. Given the fact that space is now at a premium in the vast majority of data centers, this is a key driving force in the current rise of direct-to-chip cooling.
Noise and disruption: Traditional cooling systems have a reputation for being noisy, distracting and creating a difficult working environment for data center personnel. Liquid cooling, by contrast, is almost silent. There are far fewer fans involved, and vibration too is cut dramatically. If noise levels are causing issues in your data center, liquid cooling could be the answer.
Direct-to-chip cooling challenges: Four things to watch out for
There are a few drawbacks to using direct-to-chip cooling as the main form of cooling in a data center, along with a number of challenges for operators to overcome. Let’s take a look.
Upfront costs
While the efficiency improvements seen in direct-to-chip cooling equate to lower running costs once installed, the upfront costs prior to installation are far higher than other cooling options. The components needed are more expensive, and the installation process can be complex.
Risk of leaks
There’s liquid in these systems, and while a good system shouldn’t leak, there’s no denying that leaks can and do happen.
IT equipment should be safe in such situations, because the liquid itself isn’t conductive, however, any leaks will result in problems for the cooling system – not to mention the loss of expensive coolant liquid.
Immersion cooling comes with a lower risk of leakages than direct-to-chip cooling, which is something to consider before making your choice.
Pollution
Liquid cooling is known as an eco-friendly option, and it is in terms of power consumption. However, the liquid used in these systems is far from environmentally friendly, unfortunately.
The coolant liquid, if accidentally released into waterways, can mean pollution risks. And that’s not something many companies would want to be associated with.
Overheating of other components
Direct-to-chip cooling is brilliantly effective at cooling the components it’s designed to target, but this direct method of cooling can result in other components being overlooked. There is a chance that parts of a server not targeted, like hard disks, could therefore come under threat from rising temperatures before operators became aware.
Immersion cooling, by contrast, keeps every part of a system at a consistent temperature, so is a better option in some cases.
Direct-to-chip cooling integration: what to watch out for
There’s plenty to consider before making any decisions about data center cooling. If you’re weighing up the pros and cons of direct-to-chip cooling, make sure you’re aware of the following:
Compatibility concerns
It might sound obvious, but you’d be surprised how many fall at this first hurdle! Not all systems are compatible with liquid cooling, and not all layouts will work with these cooling options.
Verify that cold plates will attach to chips properly, and that your current layout can facilitate the required plumbing (or be adjusted to allow it to do so).
Choosing your coolant
There are several different types of coolant that can be used in liquid cooling, so take the time to read up on the options before you decide on the best one for the job.
The appropriate coolant will have high thermal conductivity and low electrical conductivity, while also being non-corrosive. The cost of such liquids can vary, so it’s worth looking into the options to make sure you’re not overpaying for a liquid when a more affordable one would work just as well.
Flow rate
The best pumps for liquid cooling systems will ensure their reliability for many years to come.
When choosing a pump, make sure it’s capable of maintaining that continuous circulation that you need. Avoid those that require more maintenance than others, as that tends to result in more frequent and expensive repairs.
Choosing your heat exchanger
A heat exchanger is only effective if it’s the correct size for the system it’s working with. By that, we mean that it needs to be able to handle the temperatures you’re moving through it in a reliable and efficient way.
The capabilities of heat exchangers vary, so make sure you’ve got one that fits your system.
Identifying leaks
Leaks will be a concern in even the most high-tech liquid cooling systems, so it’s always worth investing in both the detection and prevention of leaks.
It’s worth paying more for the best quality seals and fittings, to avoid leaks further down the line. Similarly, good quality sensors are another worthwhile investment. These will draw attention to leaks immediately, allowing data center staff to resolve the issue before any damage is done.
Scheduling maintenance and repairs
Liquid cooling systems don’t require huge amounts of maintenance, but we recommend you set a schedule for monitoring and maintenance from day one. Staff will then be able to keep on top of maintenance of the cooling system, keeping it in great condition and extending its life as far as possible.
You still need a backup
Never put all your eggs in one basket, that’s all we’re saying! Even the highest quality liquid cooling systems require some form of backup, should they go wrong.
Consider using backup pumps or another form of cooling should your liquid cooling system encounter any issues. If you do so, you’ll be able to maintain the temperature of equipment and stay online even in the unlikely event that your new cooling system does let you down.
Health and safety
All cooling systems must be checked to ensure they’re compatible with existing health and safety regulations. There shouldn’t be any concerns here, but new systems are required to be compliant with a number of rules and regulations.
It’s also worth looking into the environmental impact of your liquid cooling system, and your chosen liquid in particular, to reduce your company’s carbon footprint as far as possible.
—
If you’d like to learn more about direct-to-chip cooling, or for more information about any other types of liquid cooling, get in touch with our team today.
We’re here to advise you on the best option to suit your needs, whether that’s a form of liquid cooling or a more traditional method like air cooling.
Looking for colocation?
For an unparalleled colocation experience, trust our expert team with three generations of experience
