WHY SHOULD YOU LIQUID COOL?
What’s the point of all this in 2024?
GOOD QUESTION. As you probably spotted, liquid cooling was first and foremost developed as an antidote to the excessive temperatures and noise generated by overclocking CPUs and GPUs. The more you increase the voltages and clock speeds going into those parts, the greater the excess heat produced. To go beyond the limits of our air coolers of the day, the logic was simple: greater cooling to facilitate higher clock speeds, at least until the chip itself became unstable.
In our own experience over the years, liquid cooling has allowed us to do that and more. We predominantly used it to expand our overclocking capabilities, particularly when paired with delidding CPUs (the act of removing the integrated heat spreader and replacing the standard thermal paste with liquid metal or something similar). Alternatively, if you were more focused on the audible sound of your PC at idle or ingame, you could undervolt your hardware instead, dropping temperatures further, and by increasing the surface area available through those radiators, significantly shift temperatures lower and have your fan-speeds reduced, too.
The biggest advantage, however, particularly in 2024, is just how far we’ve come in terms of auto-overclocking solutions. Almost every single Intel 14th gen chip can now quite comfortably sit at 100 Celsius for any length of time during heavier workloads. In fact, their ability to fluctuate clock speeds like that is now so dynamic that you’ll see performance significantly change, depending on what kind of cooling the chip has to hand. Our 14700K build performed 10 percent slower under its air cooler in the ITX Hydra build than it did as standard under a 360mm AIO. The same goes for GPUs and AMD’s Ryzen lines, too. The firmware, software, and BIOS are now so advanced, and the competition so fierce that edging out any temperature you can allows your CPU to really let rip at higher speeds for longer. Expanding that capacity with the help of more surface area and higher thermal capacity in your fluid should lead to significantly greater performance over longer periods of time.
HOW DOES LIQUID COOLING WORK?
Bizarrely, liquid cooling operates in a very similar manner to what you’d find in a combustion engine in a car. But before our motorized analogies, let’s start with the basics. All cooling is based on the same principles—our aim is to transfer the heat generated from the component in question, and move it away and out of the chassis.
Regardless of whether you use an air tower, all-in-one liquid cooler, or full custom loop, the basic process is always going to be the same. You have a coolant or a vapor that’s cool, it makes contact with the heat spreader or contact on the CPU or GPU, absorbs the heat from that component, and then, either via the power of convection or a pump, moves towards a radiator or fin stack of some description, where cool air (usually drawn over the fins by a fan) then pulls the heat out of the radiator and into the surrounding environment (like your office). In fact, this is why most data centers and server rooms have incredibly aggressive air conditioning.
In the case of an air tower (or any cooler or heatsink that utilizes heat pipes), there’s a special liquid inside each pipe that has a significantly lower boiling point than water. Once it makes contact with the heat at the bottom of the pipe closest to the component, it heats up, evaporates into a gas, rises to the top of the heat pipe, where that cool air is passing over those fins, cools down, condenses back into a liquid and then falls back down to the CPU or GPU to start the procedure again. This is an incredibly rapid procedure that takes very little time to occur and uses convection as its basis, making it exceptionally reliable. The downside, however, is that heat pipes like this can only be built so large, and the maximum thermal capacity of the liquid inside is relatively minor in comparison to that of any solution found in a custom loop or AIO.
Speaking of, custom loops and AIOs operate in a similar manner to that of your car. The heat source (in this case, an engine or CPU) generates that heat, and then a series of pumps move a coolant
(typically glycol based) through the system to a radiator with fans attached that expel that heat into the air. The major difference between air towers, AIOs, and custom loops is the volume and capacity available to them. Regardless of what style you use, it’s all about capacity in the fluid, and surface area on your fin stack or radiator. No matter how dense an air tower is, it cannot compete with a significantly larger AIO, particularly when we have radiators that are 360mm, 480mm, and even 520mm in diameter, with four fans of varying sizes. Couple that with the increased thermal capacity and volume of fluid available (disregarding any specialist
coolants), and what you’re left with is a far more efficient system.
IT’S ALL ABOUT THE BOTTLENECKS
There are limits to that formula, however, and if you’re not careful you can get roadblocked pretty hard and fast by one bottleneck or another. In the case of a custom loop (and even an AIO), this bottleneck typically occurs between the CPU, its IHS (integrated heat spreader, the metal plate situated on top of the silicon itself), the TIM (thermal interface material sat between the CPU and its IHS), the water block, and the thermal paste connecting the two.
Transfering the heat between these elements has an effective maximum bandwidth, and regardless of radiator size or fluid in your loop, there comes a point where the heat can’t be transferred fast enough. It’s a case of diminishing returns, where the more radiators you add won’t necessarily reduce temperatures or improve clock speeds further.
There are ways around this; delidding has become increasingly popular over the years. Depending on the processor, the TIM connecting the silicon to the IHS may be a basic thermal paste, rather than soldered directly on. Although cheaper to manufacture and produce, thermal paste chips have a lower thermal conductivity than direct metal solder. Delidding is the act of removing the IHS from the CPU, usually with a specialized tool, separating the glue from the IHS, removing the thermal paste, and then replacing that paste with a liquidmetal substitute.
In our own testing, we’ve seen temperatures on older processors drop by upwards of 10 to 18 C under load by doing this. There are a number of manufacturers out there, Der8auer being the most famous, that have the tools necessary to delid your CPU. Typically, liquid metal is applied, and the IHS then re-glued back into position during the process. However, EKWB and a number of others have created directdie waterblocks and AIOs, specifically built to be used with a delidded CPU without an IHS at all, removing that bottleneck entirely.
Again, this is entirely dependent on the CPU in question. Many of AMD’s and Intel’s processors have been soldered in the past, and trying to remove the IHS in these conditions can significantly damage the chip if you try to do so. It’s always worth double-checking online first.