Month: January 2009

The next big thing in liquid cooling?

Normal 0 false false false EN-US X-NONE X-NONE

About a month ago I got an email from one of my press contacts. “Hey Kyle, CoolIT is going to have a new CPU cooler coming down the pipe in the next month or so. Would you be interested in checking one out?” And of course I was. After several emails back and forth with Joe regarding testing the unit out, we decided that I would get first crack, with Joe doing a follow-up if there were holes or inconsistencies in my testing. So here we are. I starting writing up this review three days after the press embargo on the CoolIT Domino was lifted. I would have loved to have the review polished up and posted before the New Year, but I wasn’t able to get all the testing done to my standards in that time frame and of course, with it being the holidays, I got suitably sidetracked. Speaking of testing (which I will go into greater detail on later in the review) I actually had the necessary components kicking around to provide a good emulation of Joe’s Pentium D 805/Asus P5WD2 test rig with my own under-clocked and under-volted Pentium D 820/ Asus P5NSLI. As I said, I’ll go into more specifics later on in the review.

CoolIT is a rather interesting company. They’ve turned one of the most rarely used enthusiast level cooling setups (TEC assisted liquid cooling) into a marketable product that big name OEM’s like Dell use in their most high end setups. To that end, the Calgary, Alberta based firm (go Canada!) has built up its reputation on these TEC-assisted systems. But what happens when CoolIT loses the TEC’s? Does it become a fish out of water trying desperately to compete with big names in the market like Danger Den and Swiftech? The short answer? No. Once again CoolIT is targeting a market that no one has ever made a decent attempt to compete in. What market you ask? Oddly enough, they’ve decided to target the high-end air-cooling market. Now, most of us know that Swiftech and others have been trying to push budget liquid cooling systems that come close to the price and performance of a decent heatsink with a couple of Delta fans. The problem is that they’ve managed to get within maybe $80 of the cost of the heatsink/fan combo (with most of these budget systems going for around $150), but many enthusiasts (including myself) find it hard to justify the extra $80 to shave off another couple of degrees while reducing noise output. This is where the Domino comes in, and all our jaws hit the floor.

THE MSRP OF THE DOMINO IS $79.99 USD

Now you’re probably thinking (just like the rest of us) that the Domino is going to be a piece of junk that can’t outperform a stock heatsink. You’re also probably thinking (just like the rest of us) that the pump is going to whine, the fan is going to roar and the whole unit is going to explode after running it in your rig for two weeks. Well if you thought you couldn’t get a decent liquid cooling system for $80 you are about to be disappointed.

(Insert CoolIT Domino Picture)

(Insert CoolIT Domino Picture (2))

The Domino comes in a relatively small black box. It advertises LGA 775 / 1366 and AM2/AM2+ compatibility, so no worries about upgrading your motherboard/CPU and losing the ability to use the Domino.

(Insert CoolIT Domino Picture (3))

(Insert CoolIT Domino Picture (4))

One of the biggest advantages of a liquid cooling system like this is that unlike a large tower heatsink most of the weight is borne by the chassis itself and not the motherboard. This is great for systems that get moved around often, as you run the risk of cracking your motherboard with some of the largest tower heatsinks out there. Along with this added peace of mind CoolIT also provides a 2-year warranty, and rates the unit for over 50,000 hours of trouble free operation. All of these features make the Domino very attractive to anyone looking to take their first plunge into liquid cooling.

(Insert CoolIT Domino Picture (5))

The Domino is securely packed in a plastic clamshell. The box contains the clamshell and the manual.

(Insert CoolIT Domino Picture (6))

The clamshell itself contains the Domino, plus various bits of mounting hardware.

(Insert CoolIT Domino Picture (7))

The manual is very comprehensive and includes plenty of full color photos depicting the installation procedure.

(Insert CoolIT Domino Picture (8))

(Insert CoolIT Domino Picture (9))

Removing the top portion of the clamshell reveals the Domino’s built-in LCD display. This LCD monitors fan and pump RPM, plus fluid temperature. This LCD will also provide visual alerts if the temperature sensor, fan, or pump fails. These visual alerts are also backed up with audible beeping should the Domino have any problems. You can also see the small button on the front edge of the Domino which allows you to switch between Quiet, Performance and Full modes.

(Insert CoolIT Domino Picture (11))

(Insert Domino Display)

Moving to the rear of the Domino we can see the 120mm cooling fan that is attached to the radiator. Also notice that the fan comes pre-fitted with soft rubber mounts to reduce vibration when the Domino is attached to your chassis (screws are also included if you want more secure mounting)

(Insert CoolIT Domino Picture (12))

Swinging around to the other side of the Domino we can see the waterblock with its included mounting bracket. The tubing that CoolIT has used is a little odd, but was chosen because it offers incredibly low permeability (meaning that liquid vapor cannot easily pass through.) According to CoolIT the only way they could further reduce vapor-loss from the plumbing would be to use metal pipes, which isn’t really an option.

If you look closely you can see that the Domino uses metal barbs with multiple “flares”. I was a little surprised that CoolIT didn’t bother using clamps as well, but these fittings seem to hold extremely well and clamps would clutter up the appearance of the unit somewhat.

(Insert CoolIT Domino Picture (13))

The Domino’s mounting system uses long bolts, springs, and a custom backplate that gets attached to the rear of your motherboard. Notice that CoolIT installed little plastic cups on the bolts so your screwdriver doesn’t slip and punch a hole through your motherboard.

(Insert CoolIT Domino Picture (17))

(Insert CoolIT Domino Picture (18))

The radiator is all aluminum and uses a very effective louvered fin design. While an all-copper design might improve performance a tad (depends on the type of manufacturing being used) the all aluminum unit is no doubt far more inexpensive to produce than a similar copper design.

(Insert CoolIT Domino Picture (19))

The Domino’s pump is truly tiny. I’m not sure what its flow rate or head is, but I can tell you that it uses a ceramic bearing and supposedly produces less than 21 db of noise at 3200 rpm. I can also tell you that it is called the CFF1 (Compact Form Factor Version 1) and was designed in house by CoolIT.

(Insert CoolIT Domino Picture (20))

The waterblock uses a copper micro-channel design and appears to have been nickel plated. The micro-channel design should give excellent performance at the expense of flow rate, so it’s a good thing that CoolIT didn’t throw any other waterblocks into the loop. The bottom of the block comes covered in CoolIT’s own Pro ATC thermal compound. Unfortunately, with the tight timing of the review I was unable to include results with both the Pro ATC and my usual Arctic Silver Ceramique, so this review only includes test results with the Arctic Silver.

(Insert CoolIT Domino Picture (21))

Located at the bottom of the radiator is a small rubber cap. This cap appears to seal the reservoir that is integrated into the radiator. Either that or it is merely a simple filling port.

(Insert CoolIT Domino Picture (22))

(Insert CoolIT Domino Picture (23))

(Insert CoolIT Domino Picture (24))

(Insert CoolIT Domino Picture (25))

(Insert CoolIT Domino Picture (27))

Let’s have one final look at the whole unit before I jump into the thermal testing.

Poor-Man’s Flatness Test

After the stock thermal compound was removed I performed the Poor-Man’s Flatness Test:

(Insert CoolIT Domino Picture (28))

(Insert CoolIT Domino Picture (29))

By the looks of it the Domino’s base is more or less perfectly flat.

Installation

Installing the CoolIT Domino is an extremely easy process. While it does require access to the rear of your motherboard to mount the back plate, the install is simple and doesn’t require fiddling with small parts to get the waterblock mounted. After you mount the backplate on the motherboard and line up the bracket on the waterblock you simply thread in the included spring/bolt units in a cross pattern until you can’t turn them anymore. The bolts bottom out once you reach the appropriate mounting pressure. This is great, as it completely eliminates the possibility of using too much mounting pressure and breaking something. After that simply hold the radiator segment up to a 120mm case fan mount and pull through the rubber mounts from behind. Plug in the 3-pin power to your motherboard and that’s it. Just be sure to turn off any automatic fan control or fan speed reduction so that the Domino gets its required amount of juice.

*PLEASE NOTE*: I do not recommend using the including adhesive strips to attach the motherboard back plate. While the strips remove the irritation of having to hold the back-plate in place while you re-install your motherboard, it is a HUGE pain to remove the back-plate if you use the adhesive. I used one of the two included strips and it took me over an hour with plastic cards and flathead screw drivers to carefully peel the plate off.

The only problem I ran into during the install procedure was that the radiator segment of the Domino would not sit flush against the rear of the chassis, because the Domino is wider than a standard 120mm fan and it was bumping into one of the edges of the chassis. This wasn’t a huge problem though, as the rubber mounts stretch enough that I was able to get the radiator segment mounted with all four anyway. A similar problem was encountered when I installed the Domino in my main rig with the ULTRA m998 chassis. The Domino was too wide for me to attach it to the removable motherboard tray and slide the tray back in. But after some fancy handiwork I managed to get it mounted in the chassis just fine.

(Insert CoolIT Domino Picture (30))

(Insert CoolIT Domino Picture (31))

(Insert CoolIT Domino Picture (32))

(Insert CoolIT Domino Picture (38))

Thermal Testing

Now some of the people on the Overclockers forum have gotten ticked at me over thermal testing before, so this time I have established an accurate and repeatable method for CPU heatsink testing.

Test methods:

• Before installation, mounting surfaces (heatsink base and CPU) are cleaned using Arctic Silver’s ArctiClean product.
• Arctic Silver Ceramique is applied to the CPU according to Arctic Silver’s instructions
• Heatsink or Waterblock is mounted according manufacturer’s instructions
• System is installed in the heatsink testing chassis configured as follows:

CPU:                            Pentium D 820 under-clocked and under-volted to 2.77 GHz at a

V-core of 1.22v at full load producing 98 watts of heat (nearly identical to the output of the Pentium D 805)

Motherboard:             Asus P5NSLI

Chassis:                       Sunbeam Tuniq 3

PSU:                             Ultra X-Finity 800 watt

Cooling fans:               Front 120mm (~50 cfm), rear 120mm (~90 cfm)

Hard Disk:                   Western Digital Caviar 7200rpm w/ XP Pro x64

• System is placed under full CPU load using Orthos (multi-threaded Prime95) for a minimum of 25 hours with a minimum of six well distributed 15-minute cool down periods to ensure that the Ceramique compound has cured per Arctic Cooling’s instructions. In this case six cool downs were conducted with a little over 27 hours of full load testing.
• To measure ambient temperature a thermal probe is installed approximately 0.3 inches above the CPU fan intake, which follows the Intel recommended testing procedure. In the case of liquid cooling systems the probe is installed with 0.3 inches between it and the face of the radiator on the intake side of the rad.
• To attain the final temperature measurements the system is loaded with Orthos for two hours. During the last 20 minutes ambient temperature is recorded every 5 minutes, and the 5 results are averaged. After the two hours is up Die-temp is recorded using SpeedFan. This follows Intel’s recommended thermal testing instructions. The Die-temp is monitored for several minutes and any variations noted. If a range is noted the tester will determine if a median or mode result is a more appropriate representation of the results. For example if the die-temp is recorded as being between 56-58 degrees and it is observed that the CPU spends a roughly equal amount of time at each temperature level then a median result is used. On the other hand if it is observed that the CPU spends almost all it’s time at 56, but occasionally jumps to 58 then the mode method is used. No matter what, the tester will always indicate the observed temperature range and what method was judged to be most accurate.
• To measure the audible spectrum of the heatsink or water-cooling system in question all case fans are first unplugged. Then an Omni-directional studio microphone is held up to the heatsink or water-cooling system’s primary noise source. Should the cooling device include more than one audible component (a separate pump in a water-cooling system for example) the microphone will be placed an equal distance from the separate audible components. An audio recording approximately 10 seconds in length is made of the noise source. Finally a noise spectrum analysis is run on the audio sample.

*Please note that the spectrum analysis will NOT provide accurate absolute acoustic measure of noise. Rather a common noise source (in this case an Intel provided LGA 775 heatsink) is provided for an accurate RELATIVE comparison. The spectrum analysis provides us with the ability to isolate not just how loud the cooling device is, but also the TYPE of noise it produces. In this way we can establish if a cooling device produces an annoying loud shriek or rather if it tends to produce a far less annoying loud whoosh. This type of audio testing allows us to factor in the human element.

Performance:

In order to “zero” this testing rig to the one Joe uses I tested both the Domino and the Intel Retail heatsink. If our stock heatsink results are similar you can be fairly sure that our test setups are similar enough to be roughly interchangeable. One of the biggest problems with testing the Domino is that Joe measures C/W in the following way:

Rating Joe’s way = (Case temp – ambient temp) / 95 watts

The Case temp in his case is the thermal probe that is epoxied to his modified Pentium D 805. Unfortunately in my case I was unable to modify my Pentium D in the same way, so I measured my results in the following way:

Rating my way = (Die temp – ambient temp) / 98 watts

Thankfully this isn’t a huge deal because Joe includes Die Temp measurements with all his tests anyway. But for the sake of providing an approximate C/W that is comparable to his method I measured my results in this way as well:

Rating Joe’s way (test emulation) = (Approximate case temp – ambient temp) / 98 watts

How did I attain an approximate case temp you ask? Well I went through all of Joe’s heatsink reviews and measured the delta between case and die temp (for a total of 26 results). These results were averaged to attain the approximate delta of 9.8 degrees Celsius. This delta was subtracted from my die temp measurement to obtain the approximate case temp. Of course this result is not precise, but it should give you a good idea of where the heatsink falls on Joe’s scale.

INTEL STOCK HEATSINK:

We’ll start with the C/W according to my scale:

Then we’ll throw in Joe’s original results

Then we’ll throw in my results if I used the approximate case temp calculated above

Finally we’ll throw in Joe’s results if my method was used

By recalculating Joe’s original results using my method we get the following error between our results:

0.02 c/w = +2 degree error at 100 watt load

By recalculating my original results using Joe’s method we get the following error between our results:

0.004 c/w = -0.4 degree error at 100 watt load

In this particular case recalculating my results gives a lower error. Later results will be stated with and without this observed error

CoolIT Domino

Full:

*At this setting Die Temp varied between 50 and 51 degrees Celsius. According to the results 50 degrees was chosen because the CPU spent far more time at this temperature

Using my method:

Emulating Joe’s method:

Performance:

*At this setting Die Temp varied between 52 and 54 degrees Celsius. According to the results 53 degrees was chosen because the CPU spent equal time at all temperatures.

Using my method:

Emulating Joe’s method:

Silent:

*At this setting Die Temp varied between 56 and 58 degrees Celsius. According to the results 57 degrees was chosen because the CPU spent equal time at all temperatures.

Using my method:

Emulating Joe’s method:

Noise Spectrum Analysis

Magenta = Intel Stock Heatsink

Green = Domino on Silent mode

Purple = Domino on Performance mode

Yellow = Domino on Full mode

(Insert Domino Frequency Analysis)

Note that:

• The Decibel scale is logarithmic; therefore a noise delta of 10 db represents one sound being 10 times louder than another, while a noise delta of 20 db represents one sound being 100 times louder than another.
• The generally agreed upon range of human hearing is between 20hz and 20000hz
• The Domino on Silent mode and the Intel retail heatsink peak at almost the same decibel level, but the Intel heatsink peaks at around twice the frequency. This higher frequency translates to a sound that it subjectively much louder, as it is not drowned out by background noise.
• Performance mode peaks at a similar decibel level as Full mode, but is not consistently higher across all frequencies. So the Domino in performance mode produces more of a loud whoosh, while the Domino on Full mode produces a far more general loud roar.
• On Full the Domino is about 10 – 100 times louder than the Intel Retail heatsink. This translates to a quite audible roar that can be heard from down the hall. It’s loud, but not ludicrously loud.

Does the Domino accomplish what it set out to do?

To come up with an answer to that I looked at Joe’s test results for the Thermalright Ultra 120 heatsink with a 120mm Delta fan blasting at 2500 rpm. It’s difficult to imagine a heatsink/fan combo that would perform noticeably better than this. To ensure that an accurate comparison could be made I recalculated the Ultra 120’s C/W using my method, take a look:

Thermalright Ultra 120 results according to my method

CoolIT Domino Full mode results according to my method

If we ignore the temperature error the Domino is outperformed by the Ultra 120, but the difference would only be about 2 degrees Celsius at 100 watts. On the other hand if we include the temperature error I observed then the two cooling solutions yield almost identical results. Either way the Domino is obviously capable of going toe to toe with high-end air cooling. It won’t outperform high-end air cooling, but that isn’t what its intended purpose is. CoolIT wanted the Domino to be a reasonable alternative to high-end air cooling and that’s exactly what it is.

Summary

Just as the marketing material I received from CoolIT indicated, the Domino is positioned to make a splash in the consumer water-cooling market. Coming in at around half the price of every competing product in the budget performance water-cooling market, the Domino doesn’t sacrifice performance or build quality, and it’s built in LCD display and multiple operating modes give it a leg up over comparable heatsinks and water-cooling solutions. When you price out a comparable heatsink/fan combo plus a fan controller/temperature monitoring solution you will likely rapidly match or exceed the price of the Domino. The fact is that the Domino represents a fantastic value for your money. Just as an FYI, I also installed the Domino in my main rig and it’s keeping my e2180 nice and stable at 3.2 ghz.

Pros:

• Excellent build quality
• Solid performance allows it to go head to head with high-end air cooling
• Built in LCD monitors fan speed, pump speed, and liquid temperature
• Three different modes of operation between minimum noise and maximum performance
• Rubber mounts help to minimize vibration noise from the pump and fan
• With an MSRP of $79.99 USD the Domino is priced to sell….. fast
• 2 year warranty

Cons:

• The Domino is laterally larger than a 120mm fan. This may cause problems in cases with extremely tight fan spacing. Although the use of flexible rubber mounts should still allow mounting even if the unit ends up a tad cockeyed.
• Built in LCD is kind of useless if your case doesn’t have a window
• You’ll have to crack your case open to change the Domino’s cooling modes
• If you happen to use the included adhesive strips to attach the motherboard back plate removing the back plate later is almost impossible

Conclusion

The CoolIT Domino is something truly innovative in the computer cooling market. It takes performance liquid cooling to a place it has never been before, into a price bracket that makes it truly appealing to people on a tight budget. If you’re looking for an excellent CPU cooling solution for a very reasonable price the CoolIT Domino is right up your alley.