OVERCLOCKERS ARE ALWAYS LOOKING for something that will help them in their quest for more speed. From high-quality RAM that will run at insane bus speeds to motherboards with every last tweaking option, the list goes on and on. Water-cooling systems, which use water instead of air to cool the processor or other components, are growing in popularity. Such systems typically cool much better than traditional heatsink/fan combos, and run quieter, to boot.
Today we’re looking at one such system, Ahanix’s Iceberg. The Iceberg is sold as a complete CPU water-cooling kit which includes all necessary hoses and other hardware. Some of its more unusual attributes include its price (only $99, pretty cheap for a water-cooling system) and its ability to install to Socket 370, Socket A or Socket 478 systems with the included hardware. We’ll look at the general quality of the kit as well as what’s involved with an installation, and we’ll also examine the performance under some pretty strenuous conditions. Curious? Read on.
Water cooler primer
If you already know how water coolers work, feel free to skip to the next section. This is for the newbies, or for those who want a refresher.
The processors in mainstream PCs are air-cooled. Typically, air cooling involves strapping a large chunk of metal (a heatsink) to the CPU (metal is a very efficient conductor of heat) and ensuring that said chunk of metal has lots of fins to give it plenty of surface area. The heat generated by the CPU is conducted to the heatsink and the into the air, keeping the CPU cool.
Of course, most modern processors generate so much heat that a heatsink alone isn’t enough. Eventually, the heatsink would absorb enough heat that it would be as hot as the processor, at which point it would be doing no good at all. Therefore, a fan is typically mounted on the heatsink blows air over the fins, using that air to absorb heat from the heatsink and cool it off, so it can . . . absorb more heat from the processor. Lather, rinse, repeat ad infinitum (or at least until you shut your system off).
Water-cooling is attractive because water is much more efficient at absorbing heat than airthere’s a reason that nearly all modern cars are water-cooled rather than air-cooled. Of course, water-cooling has its own issues. For example, it’s no problem to have lots of air hanging out around electronic components, waiting to be directed by a fan to cool a heatsink (in fact, the alternative would be pretty damned difficult). Water, on the other hand, doesn’t generally play nice with electronics, so obviously the first order of business is to make sure that the water itself never touches any of the other system components directly. But the water must come close enough to the CPU to cool it properly.
There are other issues, as well. Because air is so abundant, you don’t really need to worry about cooling the air after it’s been heated by the heatsink. You just use a case fan to suck in . . . more air. With a water cooler however, there is a finite amount of water, and therefore you need to somehow cool that water off, or you’ll end up with a repeat of the “heatsink with no fan” scenario outlined above. Finally, you need a place to store the water that isn’t either cooling the CPU or being cooled itself, as well as a way to move the water between these areas.
So let’s go over these issues one at a time. First, you need something to transfer heat from the CPU to the water, and that component is a waterblock. Basically, it is a metal box with two openings that is mounted to the CPU much like a heatsink. Hoses are attached to the openings, and water flows in one and out the other. While in the waterblock, the water absorbs heat from the CPU (which has been absorbed by the metal of the waterblock) and then carries it away from the waterblock.
Now the hot water coming out of the waterblock needs to be cooled off, so it’s directed to the radiator. This device works more or less like the radiator in your car; the hot water pours into a series of small metal tubes with small fins attached to the outside. Many small tubes expose more surface area than one large tube, which cools the water more quickly. Air from a fan is directed through the fins and tubes, which helps the process, much like the fins and fan of an air-cooled heatsink.
Once the water has passed through the radiator and been cooled off, it heads back to the reservoir. The reservoir is basically a watertight tank that holds the majority of the water. The reservoir is also where the pump is located, which is the component that pushes the water through the path outlined abovefrom reservoir to waterblock to radiator and back to reservoir. Hoses connect the various components together and provide a path for the water.
The first thing that struck me when opening up the Iceberg kit was how professional it looked. I must confess that I was half expecting the components to arrive in Ziplock bags bathed in styrofoam peanuts, encased in a plain brown box. Fortunately, I was way off-base. The outer packaging looked very professional, and opened up to reveal the kit’s contents packed in a foam block with cutouts molded to each piece to prevent movement during shipping. In addition to the pieces outlined above (pump, reservoir, radiator and fan, hose and waterblock) the kit also includes various bits of necessary hardware, such as clamps for the hose ends, adhesive pads to secure the reservoir, and mounting hardware for the waterblock. The kit even came with a sticker for the outside of your case to alert people to the fact that it’s water cooled.
The construction of the various components is solid, as well. The reservoir is made of clear plastic and the fittings on the lid are pre-molded, ensuring there are no leaks. A rubber gasket seals the lid to the reservoir. in fact, once the kit was assembled, I held the reservoir upside down for a minute or so without so much as a drop of water leaking out.
The waterblock seems small and thin compared to other waterblocks, but that doesn’t mean it can’t do the jobwe’ll wait for the testing to pass judgment. The interior of the waterblock is actually made of a number of thin passages to ensure uniform water flow; Ahanix has a cutaway picture on their website here.
Some water coolers use automotive transmission or oil coolers as radiators, but the Iceberg’s radiator looks as though it was custom fabricated for this kit. It’s exactly 80mm wide, to take advantage of the fan mounting locations in most cases. The fins are flexible enough that you’ll want to take care during installation to avoid bending them, but of course this is to be expected; if they weren’t so thin, they wouldn’t do their job very well.
Interestingly, there weren’t any directions included with my kit, but a sticker on the box referred me to www.icebergcooler.com for complete installation instructions. I generally found the instructions very clear, but I did see room for improvement, particularly in the order of steps. I improvised in a few areas, and I’ll get into that below.
The first step according to Ahanix is to cut the tubing into four pieces. Right away this made no sense to me; none of the components are mounted yet, and I’m supposed to cut the tubing to run between them? I held off on this step, though I did cut the 4cm length of tubing to run between the pump and the lid of the reservoir.
Once I had attached the pump to the lid and run the pump wiring through its grommet, it was time to attach the lid to the reservoir itself. The instructions don’t go into much detail here, but if you want to get the gasket on properly, turn the lid upside down, mount the gasket around the raised area of the lid, then set the reservoir on top of it and hold it while screwing the two together. There is a groove in the reservoir for the gasket, but you’ll go crazy trying to seat it in there.
Next, you’re supposed to attach the fan to the radiator. The fan attaches using case fan screws, and here I ran into a problem. Because you’re using case fan screws, the screw threads “bite” into both the fan holes and the holes in the radiator casing. This makes mounting the fan flush with the radiator casing very diffcult. I used a 13/64 drill bit to drill out the holes in the back of the fan, which let the threads slip through but not the screw head. This way, you’re not trying to thread through two different pieces at the same time, while simultaneously trying to screw them together. IMPORTANT: If you do this, make sure to only drill out the set of holes on the “intake” side of the fan. Case fan screws are used in the fan holes on the “outlet” side of the fan to secure the fan/radiator combo to the inside of the case. Drill out all the holes, or drill out the wrong ones, and you’re, well, screwed. As I alluded to above, you want the fan mounted so that it pulls air through the radiator and out of the case. Otherwise, you’ll be pushing cool air through the radiator, heating it and then pushing it into your case, which is a bit counterproductive. It’s a good idea to plug the fan in, power up your system, and verify which direction the fan blows, or you might end up doing all this twice.
Incidentally, you might want to measure your case before purchasing to make sure the radiator will fit. The radiator is only slightly wider than a case fan, but lengthwise it extends approximately 3/4″ beyond the fan on one side, and approximately 1 1/2″ beyond the fan on the other side. Also, given the mounting method, you might have issues with cases that use non-standard fan mounts such as the Antec SX series. I experimented with this on my SX1030, and everything fit when the fan was snapped into Antec’s plastic fan mount, but it was pretty tight, which made installing the mount back into the case a little tricky. I also checked the fit in my SX635, and as near as I can tell, there’s just no way the radiator is going in that case. A metal brace runs the length of the case, and is directly in the way of the radiator. The bottom line is: eyeball it and if it looks questionable, measure it. If it looks fine, measure it anyway.
Next, you’ll want to secure the waterblock to the CPU. Make sure to use some heatsink paste to ensure a good seal. This step will vary depending on the type of CPU you have; I installed the system on an Athlon, and I have to say that I’m very impressed with the heatsink clip in the kit. I’ve seen some waterblocks that use screws to clamp the heatsink to the socket; this can make judging the required force difficult. Is it too loose to make a good seal to the CPU? Is it so tight that it crushes the core? The clip with the Iceberg kit is spring-loaded like the clips on most air-cooled heatsinks, so you just clip it on and forget it. A couple of other nice features about the clip: First, it uses all three “teeth” on each side of the socket, and second, it has an easy to reach “handle” that lets you easily install and remove the clip without a screwdriver. Nice.
Once you’ve attached the reservoir to the bottom of the case with the included pads of double-sided tape, it’s time to route the pump wiring. An obvious problem here is that the plug runs on AC power, and AC plugs are too big to fit through any of the holes on the back of a typical computer case. The Iceberg kit solves this by using a plug that attaches to the pump wiring with nothing more than a screwdriver. This lets you route the wiring through any number of holes in the back of the case, then attach the plug. One note here: if you unscrew the cable housing at the back of the plug, flip the housing over, and screw it back down, it will clamp down on the wiring and provide strain relief. Don’t take this as an excuse to be a lazy idiot and unplug the pump by yanking on the cord.
Once that’s done, fill the reservoir with water, plug in the pump, and watch the show. Hopefully, there won’t be much of a show, because that would mean you have a leak, and leaks are bad. If you do have a leak, aren’t you glad you didn’t power up the PC yet? Let the thing run for a few minutes to be absolutely sure nothing is leaking, and you’re ready to go.
Our testing methods
Ahanix’s web site has a graph comparing the operating temperature of the Iceberg to a number of air-cooled heatsinks with a “simulated CPU.” However, their graph only goes for ten minutes, and apparently doesn’t even use a real processor. I decided to give the Iceberg a bit more of a workout than that.
For my tests, I used an unlocked Athlon XP 2000+ (1667MHz stock) overclocked to 1750MHz (10.5X multiplier on a 166MHz bus) on a Soyo SY-KT400 DRAGON Ultra motherboard. I chose the 10.5X multiplier because the chip wasn’t stable at 11X. The CPU voltage was set to 1.8V. Essentially, this configuration was chosen to produce as much heat as possible. For comparison purposes, I used a Taisol CGK760172 heatsink. This is a high-end air-cooler heatsink that has been approved by AMD for processor models as high as the Athlon XP 2600+.
The test was conducted as follows: The system was powered up and given five minutes to finish booting. At that point, an initial temperature reading (both CPU and case temperature) was taken using Motherboard Monitor to read the sensors on the motherboard (no program was available to read the thermal diode in the CPU). At the same time, the stress test component of Prime95 was used to bring the CPU utilization to 100%. Thereafter, temperature readings were taken every five minutes until an hour had elapsed from the time of the first reading. At the 60 minute mark, Prime95 was stopped, allowing the CPU to drop into an idle state. Thereafter, temperature readings were taken once every minute until fifteen minutes had elapsed from the time that Prime95 was stopped.
The purpose of this test regimen was to observe the cooling characteristics of both the Iceberg and the Taisol over a relatively long period of high CPU use. The test should reveal how both cooling solutions respond to a high amount of heat over time. It should also show how quickly (and how far) the coolers lower the CPU temperature once the period of high utilization is over.
I also used a sound level meter to measure the decibel levels of each configuration. I could have measured the sound level of just the pump and heatsink, but to me the more important question is what effect, if any, the water cooler has on noise levels in the context of an actual system.
For this reason, the sound level readings were taken from outside the case, with everything powered up. That means that in each configuration, there are such fun noisemakers as a spinning hard drive, video card heatsink fan, north bridge heatsink fan, power supply fan, and case fan. Keep that in mind when looking at the sound levels. Three readings were taken for each configuration, one each from the front, side and back of the case. The front and side readings were taken from 3 1/2″ away. Because of the fans exhausting out the back of the case, the back reading was taken from 14″ away so the airflow from the fans wouldn’t interfere with the readings.
Now that the explanations are out of the way, let’s talk about the results. We’ll start with the sound level readings.
|Front measurement||48.5 dB||49.6 dB|
|Side measurement||50.8 dB||53.7 dB|
|Rear measurement||51.8 dB||54.5 dB|
According to the numbers, the Iceberg configuration is quieter than the Taisol configuration, though I wouldn’t call the difference a dramatic one. My subjective impressions are about the same: the Iceberg configuration seems a little bit quieter, but with all the other noise emanating from the case, it’s difficult to be sure.
Now let’s move on to the heart of the matter, cooling performance.
There are actually a few interesting trends here. Perhaps most importantly, the Taisol and the Iceberg both had the same maximum CPU temperature: 134 degrees Fahrenheit. This is a disappointing outcome for the Iceberg, and certainly a much different result than the graph on their web site. Incidentally, I checked for a ceiling effect here; by unplugging the fan cooling the Taisol, I was able to get the motherboard sensor to read as high as 138 degrees, so it wasn’t maxed out on these tests.
Moving beyond the max temps, we can see the differing characteristics of using air or water to cool a CPU. First off, note that the air-cooled Taisol reaches its maximum temperature much more quickly, hitting 134 in half the time of the Iceberg. This effect goes both ways, however. Once the CPU drops back to idle, the Taisol cools off more quickly than the Iceberg, and it winds up five degrees cooler than the Iceberg by the end of the test.
To understand what’s likely happening here, we need to look at the case temperatures. The Iceberg does slightly better than the Taisol in this department, which makes sense. The Iceberg is channeling nearly all of the heat from the CPU out of the case, while the Taisol is channeling that heat directly into the case. The lower temperature of the Iceberg, as well as its slower climb towards its maximum temperature, shows the effect of removing the CPU as a heat source. The slower climb is brought about by heat from other sources, such as the graphics card, the actively-cooled north bridge, and the hard drive.
With the case temperatures in mind, let’s think about exactly how the Iceberg is working. The water is cooled by a radiator, which is in turn cooled by a fan that pulls air through the radiator and out of the case. But where does that air come from? Inside the case, of course, and this is the root of the problem. As the case temperature rises, the radiator becomes less effective, because the air flowing through it is already heated to over ninety degrees Farenheit.
Over time, the water in the reservoir gets hotter and hotter, which decreases the temperature differential between the water and the CPU, further hampering performance. After the test concluded, I opened the case and put my hand on the reservoir, and it was quite warm to the touch, confirming my suspicions. Incidentally, this is also the reason that the Iceberg configuration doesn’t “recover” very quickly after the CPU returns to idle. Water tends to hold heat for a long time, which means a slow drop in CPU temperatures once the load test is concluded.
Needless to say, I wasn’t terribly impressed by the Iceberg’s performance. Marginal differences in case temperature aside, the Taisol performed equally well and recovered more quickly. However, looking at the results, it’s easy to see that the stock installation is crippled by the placement of the radiator. I started wondering how much better the system might perform if the radiator wasn’t having warm case air pulled through it.
Ahanix even alludes to this problem in the appendix of their web site, suggesting an external installation of the radiator “for higher performance.” I decided it to try the external radiator configuration, partly to see if the performance would improve and partly to test my theory about why the thing was performing so badly.
Fortunately, Ahanix gives you plenty of plastic tubing, so I had enough left over to swap in some longer pieces that I ran out the back of the case through one of the expansion card openings. The radiator/fan piece was attached to the hoses, and I spent five minutes creating a really ugly bracket to hold it in place (even MacGyver would wince). I wound up with this:
With this, err, configuration, I repeated my earlier test, the results of which are summarized in the graph below. The new results here are the ones labeled “Ext. Rad.” while the “Int. Rad.” and Taisol results are the same ones included in the previous graph.
Ahh, that’s better. Clearly, the placement of the radiator was the limiting factor before. The external configuration’s maximum temperature is eleven degrees lower than the internal configuration and case temps are down six degrees. This gives the Iceberg a substantial lead on the Taisol, and put the Iceberg much closer to what one would expect from a $99 cooling solution.
Obviously, there are issues with mounting the radiator externally. Mounting hardware of some sort (hopefully significantly better than mine) would have to be fabricated, since none is included in the kit. Having the radiator outside the case would make it more prone to damage, as wellcertainly an important point if you’re into LAN parties. However, given the substantial performance benefits from mounting the radiator externally, if you’re going to buy this kit, it’s in your best interest to mount the radiator outside your case.
My opinion on the Iceberg kit has been all over the map during the course of preparing this review. I went from being extremely impressed with the build quality of the kit, to (initially at least) extremely disappointed with the performance relative to an air-cooled heatsink. At the end of it all, however, my opinion sits somewhere in between. The price is certainly right, and the build quality is exceptional. Still, in a product like this, performance has to come above price and build quality, and testing reveals that performance (in the stock configuration at least) isn’t very impressive. Those looking for a great cooling solution that resides entirely in their PC should look elsewhere. Of course, the performance deficit is easily remedied, provided you’re willing to live with an external radiator. Accept this fact, and you wind up with a solution that cools significantly better than a very capable air-cooled heatsink. In this context, with this limitation, I feel comfortable recommending the Iceberg to anyone looking to get their feet wet (bah-dump-bump) in the area of water-cooling.