Water in a computer? Sounds crazy, but enthusiasts often do crazy things in the name of performance. When it comes to moving heat, density is king, and water beats air by several orders of magnitude in volumetric heat capacity. Now, with that physics-jargon-laden sentence out of the way, let’s talk a little more about how water cooling works.
Anyone in the cooling business should be familiar with the benefits of using liquids as an exchange medium. This is especially true in the automotive industry, where fluid-cooled engines are far more common now than air-cooled varieties. Liquid cooling is also picking up steam (ha!) in the HVAC industry, where geothermal systems utilize water’s excellent thermal characteristics to harness the constant temperature of the Earth. Thanks to the use of a superb heat mover like water and a heatsink as large as the Earth itself, common heat pumps and blowers can manage the temperatures of entire buildings with huge increases in efficiency over traditional furnaces and air exchangers.
In a computer, the principles are all the same, but the scale is a little smaller. Sure, some people go nuts and build systems large enough to cool a motorhome, but even most hardcore enthusiasts will be happy with anything that enables an overclock that makes their forum buddies jealous. The quest for higher overclocks drove gutsy early tinkerers to cobble together water cooling systems from equipment normally used in all kinds of different areas. Popular components included small car radiators, laboratory-grade tubing, and water pumps meant for small ponds and fish tanks. These folks still needed to transfer heat away from the CPU, and since few had the means to mill custom water blocks, it’s no surprise that some manufacturers picked up the slack, offering designs of their own.
Water cooling gained popularity as more PC-specific components entered the market, but folks were still piecing together systems using parts from different manufacturers. To make things a little easier, cooling companies starting bundling all the parts one might need into complete kits. These became quite popular, as they also guaranteed component compatibility. Assembly was still required, though. That doesn’t usually faze enthusiasts, but when you’re dealing with conductive liquid flowing around the internals of an electrically charged PC, securing tubing and checking for leaks can be a little harrowing.
In recent years, some manufacturers have begun dabbling with factory-sealed kits that take much of the assemblyand paranoiaout of water cooling. These units are often just as easy to install as traditional air coolers, and they don’t cost all that much than some high-end heatsink towers. We had our first real look at such a design when we pitted CoolIT’s Domino ALC against a range of traditional air coolers back in June. The Domino cooled well and was easy to use, but its fan noise put us off a little. Today we’ve corralled a similar factory-sealed unit from Corsair, the new Hydro Series H50, to see how it compares.
So what’s all in an all-in-one water cooler?
Corsair’s H50 isn’t quite as all-in-one as the Domino, but only because the included 120-mm fan doesn’t come pre-attached to the radiator. Other than that, the H50’s components require no assembly.
Instead of putting the pump right next to the radiator, Corsair embeds it inside the water block assembly. While this makes the radiator lighter and therefore easier to install, it does burden the motherboard with a little extra weight. Thankfully, the mounting hardware includes back plates to alleviate any extra stresses on the motherboard. Mounting hardware is only provided for Intel LGA775 and LGA1366 sockets, though. AMD-compatible brackets (shown on the left in the picture below) are sold separately and cost $3 online. Lynnfield-compatible mounting hardware will apparently be available soon on the Corsair website at a similar price.
Each retention package comes with a motherboard back plate, retention ring, and four spring-tensioned bolts to hold everything together. Four long screws, complete with washers, secure the fan and radiator to an enclosure’s rear 120-mm fan mount. The washers are basic metal ones, which is somewhat surprising considering the popularity of vibration-dampening rubber washers in not only cooling products, but cases, as well.
The H50’s all-copper water block surface is large enough to provide plenty of contact area for modern processors. It’s also coated with a putty-like thermal compound, further simplifying the installation process. Factory-sealed and semi-flexible tubing leads right out of the top of the water block assembly, stretching a full 11 inches before you get to the radiator.
Installing the Corsair cooler is a two-stage process, but it really doesn’t matter which step you do first. To secure the water block assembly, I found it easiest to half-tighten the retention ring to the back plate before orienting the block to slip through the ring’s circle of teeth. Once past the teeth, the entire water block assembly can be rotated to anchor it under the ring, after which tightening the bolts locks everything into place. Compared to other locking mechanisms I’ve worked with, Corsair’s approach with the H50 is both effective and simple. You do have to remove the motherboard to secure the retention bracket’s back plate, though.
To fasten the H50’s radiator, the four long screws are run first through the case and fan before anchoring in the radiator and securing everything in place. Corsair recommends that the fan be oriented as an intake to keep the coolest air flowing over the radiator fins, which makes a certain amount of sense. The amount of heat transferred from one medium to another depends on the difference in temperature between them, and the air outside your case is likely to be cooler than what’s circulating inside. Bear in mind that if you’re making changes to the orientation of any case fan, it might be necessary to adjust the orientation of others to maintain a good system airflow pattern. If you switch a filtered intake to exhaust air from the case, you might be able to move its filter up to the H50’s fan mount.
With the H50’s two main components in place, all one needs to do is connect the power. Instead using a control module, Corsair feeds power to the H50’s pump and fan separately. The four-pin fan lead is designed to plug into a motherboard header, and it will play nicely with temperature-based automatic fan speed controls. However, the pump’s three-pin connector must be connected to a power source that provides a constant 12 volts. The pump’s power cord is only 7″ long, so you might run into trouble reaching an appropriate header on your motherboard. There’s more reach with the fan’s power connector, which measures a lengthy 11 inches.
Before getting into our test results, I should take a moment to point out that the H50 doesn’t take up all that much space inside a system. Modern motherboards usually keep the socket area relatively clean, knowing that rear exhaust fans are common. Many of today’s larger air coolers still manage to crowd the socket area, coming quite close to the side panels in some cases. That shouldn’t be a problem with the H50, whose footprint is really quite modest. The fact that users can easily change the water block’s orientation also adds a measure of flexibility, since the orientation determines where the 11 inches of tubing starts.
To gauge the H50’s performance, we’re going to compare it with a similar all-in-one liquid cooler in CoolIT’s Domino ALC and a high-end air tower in Noctua’s NH-U12P. If you’re unfamiliar with either, I suggest checking out our recent air versus water cooling showdown, which details both coolers.
We tested each cooler with a Phenom II X4 940 Black Edition running at its native 3GHz and overclocked to 3.5GHz with an extra 100 millivolts flowing to the CPU. The H50 was configured with its pump running on a constant 12V and its fan plugged into the motherboard’s temperature-controlled CPU fan header. With the Domino, users are given the choice between low, medium, and performance options, with medium automatically adjusting fan speeds based on coolant temperature. The medium setting worked for us at stock speeds, but we had to crank the Domino up to performance mode at 3.5GHz to achieve stability. Our NH-U12P air cooler runs nice and quietly even on 12V, which is how we configured the cooler for testing.
Our test system hasn’t changed since that showdown, but ambient temperatures have, so we’ve conducted a new round of testing for this review. We built everything up in a Thermaltake Spedo enclosure because it’s easy to work in and represents the kind of larger enclosure you’d expect from enthusiasts interested in high-end or liquid-based cooling. All of the Spedo’s case fans were hooked up to a 7V adapter, allowing them to move plenty of air at significantly reduced noise levels that should allow us to better hear the CPU coolers we’re testing. I did remove the Spedo’s massive 220-mm side fan, though. It’s simply too noisy, even on only seven volts, so any differences in sound levels coming from the CPU coolers should be easily heard. The large 220-mm side fan has been removed, as it creates too much noise even at seven volts.
|Processor||AMD Phenom II X4 940 Black Edition 3GHz|
|System Bus||HT 3.6 GT/s (1.8GHz)|
|Motherboard||Asus M3A32 MVP Deluxe|
|Memory Size||2GB (2 DIMMs)|
|Memory Type||Corsair CM2X1024 DDR2 SDRAM|
|Hard drive||Maxtor 200GB SATA|
|Graphics||XFX GeForce 8800 GTS 512 PCIe with GeForce driver version 190.62|
|Power Supply||Enermax MODU 82+ EMD625AWT 625 Watts|
|OS||Windows Vista Ultimate x64 Edition|
|OS updates||Service Pack 1, DirectX redist update August 2008|
We used the following versions of our test applications:
The tests and methods we employ are usually publicly available and reproducible, but as is always the case with cooling and overclocking, individual results will vary. We’ve locked down as many variables as we can to focus on the performance differences between these three coolers.
CPU Temperatures at 3GHz
To compare the performance of these three coolers, we graphed a log of Everest’s CPU temperature probe over a computing session. Before starting the log, I let the system warm up to a stable idle temperature. Next, I let the system idle for 15 minutes before turning on Prime 95’s maximum heat test for all four cores. The Prime95 load ran for 20 minuteslong enough for load temperatures to stabilize with each cooler. Finally, I halted the Prime95 test, returning the system to idle, and let Everest log CPU temperatures for another 25 minutes.
At stock speeds, all three coolers perform almost identically, with the Domino managing a CPU temperature one degree lower than the rest under load. More interesting is the shape of each temperature curve around when we start and stop our load. The NH-U12P and Domino cool our test system’s processor at a quicker pace than the H50. This suggests that the H50 could have more overall mass than the Domino or that it’s simply not whisking heat away from the CPU as effectively.
Separating our data into idle, load, and cooldown temperatures, we get another look at just how closely matched these three are. The fact that all can keep a full-loaded Phenom X4 running around 43 degrees is quite impressive. How about the other part of the equation, though? These days, a heatsink doesn’t just have to keep a processor coolit has to do so quietly.
To get a feel for how much noise each setup makes, we took sound level measurements with each cooler from three locations: 12″ away from the front, top, and left side of the case.
Now that’s a quiet fan: even running on a full 12V, the Noctua makes less noise at idle than either of our two water coolers. The H50 is the quieter of the liquid-fueled units, and it emits a less obnoxious whine than the Domino.
Even under the strain of a fully loaded quad-core processor, the H50 doesn’t run any louder than it does at idle.
Upping the Ante
You’re probably not too interested in dropping over 50 greenbacks to run your processor at its stock speed, so we bumped our Phenom’s multiplier to 17.5X, raised its core voltage by 100 millivolts, and conducted another round of tests at 3.5GHz. The Domino only made it about halfway through our load test before the CPU overheated, crashing the system. We had to abandon the cooler’s medium setting in favor of performance mode, which was perfectly stable when overclocked under load.
The H50 holds a slightly higher temperature both at idle and under load. And again, it cools off more slowly than the others.
Corsair runs the CPU a degree or two warmer than its rivals here. The temperatures are still impressively low considering that we’re overclocking a 125-watt processor by 17%. Plus, there’s also the matter of noise levels:
With the Domino forced to run in performance mode to maintain system stability, the H50 ends up being significantly quieter at idle. In fact, it’s no louder at 3.5GHz than it was at 3GHz. What about under load?
This is the first time we’ve seen the H50’s fan speed ramp up, but noise levels only increase by two decibels. Of course, the Noctua air tower doesn’t get any louder at all.
The H50 is a good example of the evolution of all-in-one PC water cooling kits. Corsair’s cooling division has created an easy-to-install unit that offers great performance and commendably low noise levels. With a street price around $80 online, though, it’s not the cheapest way to get this sort of heat-dissipating power. Still, the H50 did fare quite well against the similar Domino ALC, achieving comparable CPU temperatures with consistently lower noise levels. The Domino is a little cheaper at $65, but it’s not as easy to install or adapt to different cases and custom builds.
Now, just because I’d recommend the H50 over the Domino doesn’t mean that the Corsair model is the best way to cool your processor. Air-cooled CPU heatsinks have had a lot longer to mature than their factory-sealed, liquid-fueled counterparts, and the Noctua maintained comparable temperatures to its liquid-cooled peers while making less noise than even the H50. I suspect the Corsair unit would have fared better with a higher flow rate, larger-diameter tubing, and more coolant. That said, I can still think of a couple of reasons why one might want to spring for the H50 over a high-end air cooler.
The first and most obvious reason is case clearance. Standing only 2.5″ tall, the H50’s water block and pump assembly is significantly shorter than aptly named tower heatsinks. Even with the associated tubing, the H50 requires only 3.5-4″ of clearance above the CPU; the Noctua, by comparison, needs roughly six inches of vertical clearance. Of course, the H50 still requires a case with a 120-mm fan mount, and those enclosures can typically accommodate taller air towers. Enterprising modders should be able to add a 120-mm fan mount to smaller cases and squeeze in the H50’s relatively slim radiator, though.
Another reason to opt for an H50 over an air tower relates to how often you move your system. Hanging a heavy, 6″ metal heatsink and a 120-mm fan off of a motherboard back plate works for plenty of folks, but if you often take your PC to LAN parties or otherwise move it around a lot, you might appreciate the fact that most of the H50’s weight sits lower to the motherboard or is anchored directly to the case. I’ve never had a problem with larger tower heatsinks damaging motherboards, but I’ve heard enough horror stories to prefer lower-profile coolers for more mobile builds.
The H50’s simple design should also appeal to folks who are specifically looking to dabble in water cooling but don’t want to go through an involved installation. As we’ve seen, though, the right big-honkin’ heatsink can give you lower temperatures and noise levels.