Closed Loop AIO Liquid Coolers: 14-way Mega Roundup Review
by E. Fylladitakis on February 12, 2014 7:00 AM ESTCooler Master
It would seem that Cooler Master simply wanted to create the best performing AIO coolers available. In that respect, they certainly managed to do so, as their Nepton series coolers are constantly at the top of our high load thermal performance charts. On the other hand, one would have to be at least partially deaf in order to use these coolers with the fans anywhere near their maximum speed. Even with their voltage reduced down to 7V, both kits are rather noisy. It is not possible to run these kits quiet even if you reduce the voltage even further, as the fans are clearly audible from 1 meter away even at just 4.7V, the minimum voltage required to start them. Therefore, we simply cannot recommend them to anyone seeking a quiet cooling solution.
The Seidon 120V however is an entirely different product. Priced at just $49.99, it is more of an alternative to an average air cooler than competition for liquid cooling solutions. It fares relatively well at lower loads but its performance diminishes with thermal loads greater than 150W. With an average thermal resistance slightly above 0.014 °C/W, it will most likely be unable to compete with most extreme performance air coolers, and it has a particularly noisy pump, making it noisier than other 120mm AIO coolers. However, it has other merits, as it does not stress the motherboard with its weight and it requires very little space around the CPU area, making it ideal for special builds and/or systems that are moved around a lot.
Corsair
Corsair has such a vast selection of AIO cooling products that we could make a roundup just for them. Each of the five coolers that they shipped us for this roundup displays entirely different behavior; thus, each of them is suitable for a different type of user.
With the H75 Corsair is offering a compact 120mm AIO cooler but with two 120mm fans for extra performance. The H75 performs relatively well, although it generally does better at thermal loads lower than 150W due to the low capacity of its small radiator. The use of two fans however increases the noise of the system, giving the advantage to Enermax's Liqmax 120S if low-noise operation is the top priority.
The 140mm H90 on the other hand displays great all-around performance and very low noise levels. Even with its fan constantly running at maximum speed, the H90 can be considered fairly discreet and is comfortable for everyday use. If quiet computing is what drives you, the H90 deserves a very strong consideration.
Corsair informed us that the H100i is their most popular AIO cooler and we can see why. Despite its size, the H100i easily competes with coolers using significantly larger radiators. The stock fans have a wide operating range and the USB interface allows the user to adjust the performance/noise ratio to meet his or her exact needs. Furthermore, the size of the radiator makes the H100i compatible with a relatively wide array of cases.
After looking at the performance figures of the H100i, we felt disappointed by the performance of the newly released H105. The H105 hardly performs much better and it only does so when the thermal load is very high. Unfortunately, the thicker radiator can create compatibility issues and it also is more expensive than the H100i, all while lacking the USB interface that the H100i has. It is hard to recommend the H105 over the H100i for the slightly better thermal performance alone, unless of course maximum thermal performance is virtually the only concern of the user and a larger radiator cannot fit into the system.
Finally, the last AIO cooler from Corsair that we have tested, the Hydro H110, possibly stands as the performance winner of this roundup. Although it does not have the best thermal performance, it is very close to the top of the charts and manages to do so while maintaining very low noise levels. However, the size of the radiator limits the compatibility of the cooler with only a handful of cases currently available and, considering that the retail price of the H110 is over $125 at the time of this review, it is a costly thermal solution.
Facebook
Twitter
RSS
139 Comments
View All Comments
Jon-R - Thursday, February 13, 2014 - link
How did you make sure that the noise-floor and not the bottom range of your sound level meter was 30.4dB(A)? Would it be fair to say that the range that SPCR considers quiet(around 13dBA) wouldn't be measurable with the meter and room you used?E.Fyll - Thursday, February 13, 2014 - link
The meter that I am using can read well below 30 dB(A). The accuracy however declines dramatically the further you step away from the bottom limit. I chose the specific setup exactly because it is ideal for basic sound level testing. I would not perform (and do not plan on performing) any tests that I myself consider them invalid in any way. That being said, if I needed a better meter or another setup in order to produce valid results, I would not perform noise testing at all until I could afford the equipment. This is why I do not test and review several other parts as well, such as fans. When I have proper equipment to do so, articles about them will start flowing as well.No, that is wrong. Aside from the fact that the equipment is vastly different and you should not compare their numbers with mine in any way, what SPCR considers quiet would still add to the floor noise level of my room, or of any other room. It is wrong to even consider that the fan is producing 13 dB(A); it just adds up and brings the floor-noise of their setup up to 13 dB(A). If a fan would really produce exactly 13 dB(A), it means that in a zero-dB(A) environment (ISO lab) the meter would read 13 dB(A). If you test the same fan in a 12 dB(A) environment, it will not add 13 dB(A) to it, it is at a different level of a logarithmic scale. Read and try to understand my explanation above.
Only compare SPCR's numbers to their own and mine to mine. Never in between different setups. You cannot compare anything else than measurements taken in zero-dB(A) environments with vast anechoic chambers (far larger than a room) and such equipment is unavailable to common people.
Jon-R - Thursday, February 13, 2014 - link
Interesting. Extechs site lists the HD600s range as beeing from 30dB upwards. Still the fact remains that what you call silent(sub-30dBA) is vastly louder than what SPCR consider silent. Earlier in this thread, you said that you consider that level silent because your equipment can't measure it. Does this then not mean that you're not able to measure the truly quiet fans? Even SPCR can't get measurements from the quietest fans(sub-11dBA) at the slowest speeds, as their noise is below the noise-floor of their chamber.E.Fyll - Thursday, February 13, 2014 - link
I believe that you do not understand much of what I said.At 30.4 dB(A), the conditions are an empty room, in a rural area, at 2:00AM. It does not really get much more silent than that. You cannot possibly discern any sound under such conditions. The point where my ears begin discerning any sound is above 33-34 dB(A). So, comparing my readings to those of SPCR's is a massive mistake. It may be equipment, the conditions or anything else, but the scale is entirely different. And yes, as you said the Extech HD600, which is one of the better SPL meters, has a minimum specified range limit of 30 dB(A). Think to yourself, why one of the best devices specifically made for the measurement of sound levels cannot accurately display anything below 30 dB(A)? Because, in reality, your very room right now has a floor noise that is higher than that figure.
If someone tells you that they cannot take a measurement because the sound level is "too low", then their methodology is flawed or the sound level of the said device indeed is too low to cause any registerable reading. As their equipment is very good, most likely the latter. Even an 1 dB(A) source will add to a 40 dB(A) environment, the magnitude however is so small that it would require equipment capable of displaying a 3rd or 4th decimal point. So let me try and explain it to you once more. If a "sub-11 dB(A)" fan is inserted inside an aechoic chamber and is tested with equipment that has a self-noise of 12 dB(A), the reader should register something higher than 12 dB(A), as the fan will simply add to the self-noise of the microphone. The same goes for an environment with a floor noise of 30.4 dB(A), the "sub-11 dB(A)" fan will simply add to that figure. To be entirely precise, an 11.8 dB(A) fan should make the meter register 30.6 dB(A), which is hardly any different than absolute silence on my setup. If it doesn't cause any change on the reading, it really is far too quiet and the difference is not enough to cause a change on the setup's reading.
Bottom line: At 30.4 dB(A), my setup depicts absolute silence. At 12 dB(A) (I think), SPCR's setup depicts absolute silence. Do not try and compare the readings of the two setups, they are on a different level of the decibel scale and recorded using entirely different equipment. Apples and oranges.
If you cannot understand this, I simply cannot explain it in any simpler way, sorry.
Jon-R - Thursday, February 13, 2014 - link
I do understand what you're saying, I'm simply not agreeing with it.SPCR measured a noise floor of 18 dBA before they built their anechoic chamber. And that's in Vancouver. You say you measure 30.4 dB(A), but you're using a meter that only goes down to 30 dB(A), and has a accuracy of 1.4dB(A). It seems likely that what you're measuring is the lower limit of your meter. I don't understand what has made you so confident that the noise floor of your locale actually is 30.4 dBA, and not lower. Seems awfully coincidental. There's a reason why SPCR never considered a meter like the one you're using to be suitable for their tests, because it simply isn't sensitive enough to be used for the measurement of quiet equipment. Instead they went with a over 2000$ Type 1 mic setup. They did consider going the SML route, but getting similar sensitivity would've cost them northwards of 10 000$.
There has been plenty of articles here at Anandtech where the reviewers have said that the noise measured is below the 30 dBA noise floor, and as such can't be measured. Reviews where ever piece under 30dBA is on the same line of 30dBA. If your locale has a noise floor of 30 dBA, a 12 dBA piece of equipment will not add enough noise to be measurable by your meter. The addition is too small for it to be measurable, because it would require an unrealistically accurate meter. SPCR does this when a fan is so quiet that the noise it produces is drowned out by the background noise of their 11 dBA anechoic chamber. That is what they mean when they say that it's below the noise floor. Once something produces enough noise to give a readable increase over the noise floor, they measure that.
E.Fyll - Friday, February 14, 2014 - link
Let me say this once again, because you still do not understand what I am trying to tell you. Unfortunately, this will be my last time, as I simply do not have more time to spend on such a matter.The HD600 can and will display readings below 30 dB(A). It just lacks any real accuracy within this range and for a good reason. It is far below the level that humans can really sense in normal environments and way too low for any sensor (yes, including microphones) to accurately sense. A really good setup needs constant calibration at sub-30 dB(A) ranges, because it can sense even the slightest vibrations of the air, something you cannot fathom to sense with your ears.
As for the "If your locale has a noise floor of 30 dBA, a 12 dBA piece of equipment will not add enough noise to be measurable by your meter", yes, it will, and I can even calculate exactly how much it will add. I actually did that calculation above for you but you obviously did not even bother to read it thoroughly, as I suggested. Furthermore, the fan does not really produce 12 dB(A); the meter reads 12 db(A) as the result of the fan's noise plus the self-noise of the sensor. The equipment should always read the self-noise of the sensor and any other source would add to that. Unless of course if we are talking about an ISO certification lab with a 0 dB(A) acoustic chamber and specialized equipment. If such sensitive equipment has a self-noise of 8 dB(A) and the device adds nothing to it, it does not mean that the device is producing lower than 8 dB(A) but that it produced no noise at all. So, the readings you see at other sites, whichever site that might be, they are what the device adds to the environment, or to the self-noise of the instrument if the environmental noise is too low. So...there is no "12 dB(A) fan", unless you took that reading in a lab.
Instead of believing whatever you read simply because you want to or whatever someone is trying to convince you online, go find a reliable source (not just any website) and check a few facts for yourself. For instance, instead of looking at noise graphs here and there, go have a look at the chart in the last page of my meter's manual. You will see that a whisper over a distance of 5" is registered at above 30 dB(A). A residential area at night is above 40 dB(A). A household is nearly 50 dB(A). You cannot easily create sub-30 dB(A) environments in the real world even if you try very hard. You mislead yourself by believing that a device that allegedly is producing 20 dB(A) inside an anechoic chamber is loud, when in reality 20 dB(A) are absolutely nothing on their own. If however that device is inserted to the 35 dB(A) environment of your quiet room, it will bring it up to 38 dB(A) (random number, I did no calculations here) and you will notice it.
And no, they would not have gotten "similar sensitivity" for $10.000+. There is a very good reason why the price multiplies manyfold and I simply cannot even try and explain it to you. Do not take that as an insult, but someone who tells me that my meter "has an accuracy of 1.4 dB(A)" would never understand the complexity of such a text, as he obviously has no knowledge about measurement systems whatsoever. The accuracy that manufacturers list is the lowest possible and refers to the top of the meter's range (that goes for all kinds of meters, not just SPL meters). So, indeed, the accuracy of my meter is +/- 1.4 dB(A), when it is reading the maximum value of the set range. The lower the value, the more accurate the reading becomes. In the case of SPL meters (including microphones), they become rapidly inaccurate below 30 dB because the sound pressure level is far too low to generate a proper electric signal and can be affected even by the slightest vibration. These are the utmost basics when it comes to measurement systems and equipment.
E.Fyll - Wednesday, February 12, 2014 - link
My apologies, I forgot.I simply cannot concern myself with what other reviewers/people think and I do not want to impose any of my thoughts towards other people as well. Whether they (SPCR or anyone else) like AIO coolers or not, I find no reason to comment upon it. I never said if I like them or not either. What I like and what I do not like are my personal, subjective preferences. I cannot impose my preferences upon other people. What I can do is test the products, log my data, present them to people, comment on their quality/bundle/value, take pictures and perhaps provide some recommendations. It is up to every reader to decide whether they like a product or not, be it aesthetically or otherwise. There are people that like AIO coolers just because they leave the RAM slots easily accessible, there are people that hate them just because they use liquid. Likewise, there are people who will find even the slightest humming noise annoying, while others would tolerate a Delta fan for that extra 50 MHz out of their CPU. Each person can decide on his/her own. Some like to call this "free will", I prefer the term "critical thinking".
Subyman - Wednesday, February 12, 2014 - link
Thank you for the great write-up. It can be hard to find decent comparisons for AIO coolers. I don't personally use them, but my friends and clients seem to love them so I need to be informed. I loved your introduction! I remember brazing barb fittings on a heater core I picked up at the local car parts shop to cool my PC. I used hardware store tubing, an aquarium pump, and made a fan shroud out of fiberglass. That was a very fun summer!Sivar - Wednesday, February 12, 2014 - link
Silent = No sound whatsoeverQuiet = Barely audible or inaudible
No cooler with moving parts is entirely silent even when its fan's voltage is reduced down to 7 Volts.
DLoweinc - Wednesday, February 12, 2014 - link
Nitpicking, but I had a Coolit systems ECO cooler in 2010, so I don't think 2012 was the year these 'appeared'.That cooler, by the way, leaked after 2 years and destroyed my mobo/processor. The coolant was mineral based and caused enough damage, who knows if a water based coolant would have left the system components undamaged? The leak was at the tubing where it went into the pump. This was in a 24/7 system that did not get touched until I noticed one day it was offline.