Plate-form & protocol

Test platform

For this survey we used a Socket 775 type platform. All CPU coolers are at least compatible with this platform. Most of the models tested are also compatible with other Sockets and if not, there is always an equivalent model adapted to the platforms. So you can adapt our test conclusions to other platforms such as the sockets 939, 940 or 478, because the processors have the same nature, geometry and produce pretty much the same effects.

We used the following reference configurations:

• Epox 5EPA+ (Rev. 1.0)
• Intel Pentium 4 Prescott 540 (SL7KL)
• 2x 512MB DDR PC3200 Mushkin Hi Perf LII V2
• Morex CWT 550-ADP power supply

We installed a Windows 2003 Server on an empty hard drive with BurnP6 so as to use the processor to its maximum during the full loading phase.

Previous tests helped us determine the frequency and Vcore used in Overclocking. To establish these measurements, we chose to use the Intel Box radiator, which is apparently one of most constraining for Overclocking. The P4 540 initially clocked at 3200 MHz for a 1.36 V Vcore was finally stable at 3760 MHz with an additional +0.15V Vcore (1.51V). The CPU cooler voltage was at 5V minimum. As we don’t yet have access to the ClockGen for the PLL of the 5EPA+, the Overclocking was made directly from the bios.

Heat thermal dispersal in full load for each of the two configurations is respectively 85W and 125W approximately. The Pentium Thermal Throttling was deactivated to avoid any measure disturbance. The thermal paste used will be plain silicon.


For radiators provided without fans, we used the Papst fan range. Just for your information, these fans are sometimes available when a fan is included with the computer.

Measuring hardware

First of all, to measure the temperature, we used a thermocouple Voltcraft K102 with two pathways simultaneously associated with two K type thermocouples. One is adapted to the air temperature and the second to the IHS temperature measurements.

To introduce the noise/performance concept, it is imperative to also use a sonometer. The digital product used here is a class II Voltcraft IEC 651. It measures the noise level in A and C scales from 30 to 130 dB. Finally, we also used a standard voltmeter as well as a Coolermaster Aerogate III to establish the fan rotation speed according to the voltage.



For the reasons stated above, the CPU temperature will be taken directly from the IHS with a thermocouple. It is important to point out that during all of the measurement tests (that lasted over 1 month), the thermocouple was never moved, to avoid any additional useless errors. As shown in the picture below, the probe cable goes below the Socket flap. The probe is installed so that the solder ball is directly in touch with the IHS side. Once in position, the Socket is locked to block the probe definitively in the right position.

Each CPU cooler is installed at least three times to make sure it is in the right position; an assembling error is always possible (we compared the delta obtained + measure errors). Once in position, it goes through 16 periods of at least 20 minutes altogether, at the end of which we measured the temperature. After these 20 minutes, experience shows that the transitional running is over and that it is possible to collect reliable data.

How to understand the figure reported

In this article, we will publish the measurements obtained with 5V, 7V, 9V and 12V fan Voltage whether through the regulation provided or through an additional, external, one.

Four measures are reported: fan rotating speed, noise level, temperature delta with the processor in use, before Overclocking and then after Overclocking. On the noise level side, results obtained are from 32 and 68 dB(A). 32 dB(A) correspond to a closed tower and 68 dB(A) is…completely unbearable! So at 35 dB(A), it is almost inaudible. At 40 dB(A) the ventilation is still discreet, at 45 dB(A) noise appears to be quite present. At 50 dB(A) it is really present and beyond that, you must be really fond of noisy computers…

The temperature delta represents the difference between the processor’s IHS temperature and the air temperature used to cool it down. Values measured here vary from 12°C to 45 °C.
Since in summer, your room temperature can climb up to 30°C and the tower air to 40°C, it is important to have a low delta. Under these conditions to have a processor below 65°C, it is essential to have a delta inferior to 25°C !

Aerocool – GT1000

Presentation

Aerocool is the first to release a cooler with a golden base and blades. We may wonder what the real interest of this glazing is: clearly, it is only an aesthetic device because technically the glazing won’t improve performances. However, it does have the merit that this cooler will probably turn up as the most luxurious cooler on the market. The GT1000, sold with a 92-mm fan, is compatible with all Sockets.

Installation

The procedure is very easy and avoids any assembling mistakes thanks to the screws’ support, that will be screwed as far as the abutment.

Measures and results

The fan provided with the GT1000 has a maximum functioning range (5 to 12 V):

Appraisal

Without providing fantastic raw performances or being exceptionally silent, the GT1000 is in a good position compared to other models and it is a good compromise between raw performance and silence. Also, it behaves similarly to the Zalman 7000Cu and yet provides slightly better performances.

ASUSTeK – Star Ice

Presentation

This cooler, sold under the Asus brand, is based on Thermaltake components (base, heatpipe, blades, fixing set). Asus chose an 80-mm fan integrated in a plastic bezel that looks very much like a plane turbo reactor. Finishing touches are very good and several accessories are included such as two fan mates.


Why two? To allow the user the possibility of changing the fan speed, either from the front (front, 3"1/2) or from the back (extension board) of the computer.



Now we will have to check if these turbo reactors are as efficient as they are good-looking.

Installation

It starts with the back plate attached with screws and spacers. The Star Ice is then simply placed on the CPU. The assembling ends with the tightening of the plate to the cooler base by means of four screws.

Measures and results

The fanmates provided, helped to change the voltage from 5 to 12V:

Appraisal

With this cooler, appearances are just what they seem to be… The fan is extremely noisy: the noisiest of all the models tested here. However, its performances are also very high. Combining the 6 heatpipes with this high airflow rate fan makes of the Star Ice the best cooler in terms of raw performances with a delta Air-IHS ratio of only 19.4°C in full load (o/c). This fan is for performance enthusiasts only, even if it is to the detriment of the noise level. If you are looking for a silent configuration, this cooler is definitively not the right choice.

Arctic-Cooling – Freezer 7

Presentation

The Freezer 7 is one of the rare, cheap, heatpipes (less than US$40). The heatpipes’ specificity is that the evaporation area is halfway between the beginning and the end of the tube and not at the end. So, this system is closer to a four-heatpipe configuration. The copper base looks relatively raw, as do the finishing touches. It is made up of two parts to capture the two tubes. Blades are in aluminium, and they are present all the way to the top of the cooler.


The original (70 mm) ventilation is suspended by some sort of trapeze. The lack of skirt is quite surprising from the point of view of airflow optimization and guidance. Finally, the chosen fixing system is based on Intel’s clips system for box models. That means this one is only compatible with the Socket 775.

Installation

The installation is no more difficult than the radiator Box. Be careful with the cables as the fan blades don’t have any protection.



If it is used in a tower, one can orient the airflow generated either to the power supply or to the outside.

Measures and results

The table below presents some of the measures for 7 and 12 V. The ventilation, even it is low, is still acceptable for 9 and 12V. At 5V, the fan doesn’t start without external help so we did not carry out any measurements with this voltage. Finally, we measured results at 7V except for the full load phase in o/c, which wasn’t at all stable because of an overly high temperature:

Appraisal

The worst of all in terms of performances! Despite average performances in 12V and a moderate noise level, it is incapable of dealing with the load imposed by the Pentium 4 in full load if the voltage is reduced. In Overclocking, we were unable to obtain a result because the system crashes too fast. It is also the only one that makes the system completely unstable.

Performances went down rapidly when the voltage was slightly reduced. Temperatures reached were very high and the rotation speed was dramatically reduced. The low diameter fan and the lack of skirt don’t help at all and they don’t force the air through the blades. This design must be left out, even if it is silent under 9V…

Coolermaster - CI5-9HDPA-01

Presentation

The Coolermaster CI5-9HDPA-01 is the brand’s entry-level product for Socket 775. Standard 92-mm radiator, copper base, aluminium body and blades, it doesn’t feature any particular characteristics. Even the LGA775 fixing system is based on the normal system. Despite the lack of extraordinary aspects, the cooler’s finishing touches are elegant.

Installation

Assembling is standard and is made through four clips to be pushed.

Measures and results

Appraisal

The noise level is too high and it is the last but one for all voltages. It is true that a 92 mm fan at 4500 Rpm makes a lot of noise…We notice that designs in flower shapes provide nice results but require a more sustained ventilation than heatpipe coolers. Thermal performances also fall apart faster than average when the voltage is reduced to lower the noise level. You might as well keep the P4 box cooler: it is better!

Coolermaster - Hyper48

Presentation

Leading product of the Coolermaster range, the Hyper48 is, according to its specifications, capable of reaching very high performances. It includes more than 800g of copper for the base and blades, as well as four heat pipes and excellent finishing touches. A 90-mm fan with a protection grid ensures ventilation.


There is, quite logically, a compatibility with other Sockets (478, 754, 939 and 940) like with most high-end products. It is sold with all the screws and accessories needed to install it.

Installation

I haven’t had the opportunity to test it with platforms other than the socket 775. The installation starts by fixing the metallic support under the dispersal around the cylindrical base with four screws. We then have a plate provided, used at a screw-nut at the back of the motherboard. It is also used to make the system more rigid and allows a good tightening while reducing the motherboard deformation. The installation is finished after tightening the four screws to press against the processor. We found that access to the screw wasn’t very easy and we had to use our screwdriver diagonally, but that is just a detail.

Measures and results

Appraisal

This cooler is the only one that does not increase its noise level from 5 to 9V despite an obviously increased fan rotation speed. So it is much more interesting to use it at 9V, in order to reduce the temperature by 12°C compared to 5V without an increase in noise levels! It is also very discreet at 12V with only 45 dB(A) measured, far from the 70 dB(A) measured with some of the coolers we tested… Its noise level is equivalent to the most silent fans such as the S1. It is a very good product if you are looking for silence as a priority, along with good performances around 9-10V.

Coolermaster – Hyper6+

Presentation

After the Hyper6’ success, Coolermaster is back with the Hyper6+. This cooler features 6 heat pipes and very handsome finishing touches. It is compatible with all Sockets and comes with a transparent fan, a grid and a 4-pin power connector.

Installation

The installation is easy but it does require removing the motherboard from the tower case to position the plate in the back (this is often the case for heavy CPU coolers). The cooler is then placed on the CPU with four screws. The installation ends with the tightening of four screw-nuts at the back of the plate.

Measures and results

We made four measurement series from 5 to 12 Volts:

Appraisal

The hyper6+ provides a really interesting level of performances. At 12V for example, it is the second most efficient of this survey (1°C more than the Star Ice) and is 16 dB(A)D less noisy. In the end, it doesn’t offer such a good compromise as the Thermalright products, but it provides excellent performances, especially at full voltage.

Cooljag – Sunflower

Presentation

Cooljag product ranges include a desktop cooler, but mostly for servers with low-profile CPU coolers for 1U Rack and other types of products. They sent us the desktop cooler, the Sunflower. It is a radiator with a copper base and blades. With the 92 mm-copper ventilation powered by a four-pin connector, the entire cooler’s weight is over 700g.


The pictures speak for themselves as regards the finishing touches, it is very satisfying. It is only compatible with the socket T, but others versions are available for other platforms. Finally, the installation system includes screws, springs and screw nuts.

Installation

The installation starts with the positioning of the reinforcement plate in the back of the motherboard (the studs, close to screw nut level, facilitate this operation).

Then you just need to place the radiator on the CPU to tighten up the whole system. From that point of view, it is almost impossible to mess up the Cooljag installation. The screws’ abutments rest on the motherboard’s eyelet. This solution offers the possibility of carrying out a uniform tightening, if the technique is correctly applied, and it is with the Sunflower.

Measures and results

Below 9V, the rotation speed is too low to measure any temperatures.

Appraisal

It provides very good overall performances but has a high noise level with almost 60 dB(A) at 12 V. The impossibility of lowering the voltage below 8-9 V doesn’t provide an opportunity to reduce the noise level sufficiently.

Gigabyte - 3D Rocket Pro

Presentation

The 3D Rocket Pro is a four heat pipes’ cooler sold by Gigabyte. The cooler design is quite original and has a 60-mm centrifugal ventilation installed inside it. It is compatible with most Sockets and includes a connector to plug it directly to a power supply Molex. One good point, a 3-pin format cable for rotation speed control is provided. Finally, Gigabyte is one of the few to provide a fan control via a potentiometer installed in front of the tower in a 3"1/2 bay. We also noticed (for those who might be interested) a couple of blue LED installed on the top of the blades.

Installation

Without being irksome, the installation takes a little more time than a clips cooler. The LGA775 support is first screwed (it is important to pay attention to the instructions). The base is then placed on the HIS and the two metal gears in the support notches. You will need to have small fingers or small pliers. Anyway, we advise you to install the 3D Rocket Pro outside the tower case.

Measures and results

As the fan has its own regulation system, it is impossible to add a Fanmate. We only tested the 3D Rocket Pro with the Voltage range allowed by the potentiometer provided: from 6 to 11 V:

Appraisal

The cooler’s compactness is really attractive, because it doesn’t exceed the Socket size. So no incompatibility is expected. Performances are very good and don’t drop too much with a lower voltage. The noise level is unfortunately still present, and the centrifugal fan, which spins at 4000 Rpm, is as noisy as an axial one. It is unfortunately at the bottom of the list in terms of noise level.

Intel – Box

Presentation

Provided with all new Intel CPU, the Box is one of the most common radiators. The one provided with the Pentium4 540 is a standard model with copper base and aluminium blades. An 80-mm fan powered by the new 4 pin connectors ensures the ventilation. The installation system is the same as most of the thermal dispersal’ because it is, of course, Intel’s official solution.

Installation

There are no particular difficulties for the installation. It is done in a few seconds.

Measures and results

The box fans reacts strangely to Voltage modifications. Below 7V, the rotation speed re-increases to reach 1740 RPM at 5 V. 5 and 6V voltages therefore do not offer any particular interest.

Appraisal

The Box is far from being the best cooler in terms of thermal efficiency, but Intel made some serious progresses for the fan choice. This one is quite silent compared to previous generations. At 7V it even has an excellent noise level, but thermal performances are much lower, and so it is unfortunately practically last. We can’t really complain anyway about it because it is given for free with each processor…

Noiseblocker - Cool Scraper

Presentation

The Cool Scraper is a single heat pipe column manufactured by NPowerTek. The Cool Scraper’s specificity is an extraordinary size (more than 150mm high). It is built entirely around a central tube where all the aluminium blades are attached. It is sold with the Noiseblocker SX1 (120-mm) and a set of tools for the installation on 478 and K8 sockets.


The cooler also comes with a potentiometer, a set of electroluminescent LED and several adapters for Molex and three pins connectors. Finally, the fun touch is that, seen from above, the Cool Scraper looks rather like the Batman logo.

Installation

The installation system is relatively simple. For each Socket two plates are provided: one to reinforce the motherboard and the other one to place on the base of the Cool Scraper. The first one is installed with screws and nuts at the back of the motherboard. They are tightened with 4 nuts, just like some water-cooling blocs. It must be very tight because of the very large overhang due to its height.

Measures and results

Measurements were carried out with two different 120-mm fans: the Noiseblocker SX1 sold with the cooler, of course, and the Papst 4412 F/2GL. If some measurements are lacking, it simply means that the rotation speed evaluated for a voltage was too low to be usable.

Appraisal

The Cool Scraper is one of the most silent coolers, thanks to the slow speed of the 120-mm fan. The supertube and the very high blade surface provide excellent performances. It is also the most silent when the Papst 4412 F/2GL Voltage is reduced to 5V. The noise level is very low and the airflow almost reduced to zero. The choice of the fan has very little influence on performances and only the size of the cooler might be a problem for some towers. Finally, it is an excellent choice in terms of performance and silence.

Noiseblocker - Cool Tower

Presentation

The Noiseblocker Cool Scraper was sent along with its little brother, the Cool Tower. It looks very much like the Cool Scraper: technology, installation system, and material. The only difference is its size. It is designed to work with an 80-mm fan. It is sold with the famous S4 (the fastest fan in the 80-mm Noiseblocker range).



The fan is installed with two piano strings attached to the blades. One can also use two on each side, so as to increase performances.

Installation

The installation process is identical to the Cool Scraper, there is nothing to add.

Measures and results

We tested the Cool Tower with three different 80-mm fans (1 Noiseblocker and 2 Papst). Like its bigger brother, some of the voltages were too low and it was impossible to gauge the temperature. :

Appraisal

With the number of fans tested with the Cool Tower, we clearly show the connection between a silent fan and the performances provided. The slower it spins, the more silent it is. The airflow also diminishes with the speed as well, naturally, as its performance. The point is to find the best compromise. The S-4 provided is the most efficient. It spins faster, even with a lower voltage. It provides very good performances with a moderate noise level, but is less efficient than the Papst for comparable dB(A). There is a considerable 10°C gap at 12V in full load between the most powerful and the weaker fan. So you can see that the choice of fan has important repercussions.
Like its larger brother, the supertube provides very good performances and this Cool Tower is also a very good choice. It is not quite as good as the Cool Scraper, but it is also less bulky and is only 80 mm high!

Noiseless Cooler - Calmera KS10

Presentation

A couple of months ago the German company Noiseless-Cooler released a range of coolers based on heat pipes used in thermosiphon and developed by LG ThermoTechnologies. The whole cooler is made of copper except for the aluminum blades. If you shake it, you can clearly hear the fluid moving inside, to keep the evaporator immersed (the empty area in the bottom). It is sold with a 92-mm fan easily interchangeable and that requires approximately 2.5 Watts. The LGA775 comes with a plate and clips identical to the standard Box installation system.


As shown in the pictures, the angle between the base and the heatpipe isn’t 90°. The gravity effect is crucial in this technology. It is imperative to keep the condenser (blades) above the base. If it isn’t, the heatpipe would run "on empty" (the liquid will go in the pipe) and won’t provide the required results. Heat would only be transmitted through the partitions of the central tube, in other words, through almost nothing.

Installation

The installation is just a pure formality. However, you will need to check that the condenser is always higher than the base. If this seems obvious when you install the cooler on a motherboard in a horizontal position, it isn’t when it is in a tower. With a traditional tower (power supply above the motherboard), the manufacturer suggest orientating the Calmera in the power supply’s aspiration direction. It seems to be the best solution. The base is round so it is possible to adjust the position accurately before tightening the system onto the processor.

Measures and results

As well as the results with the fan provided, we also included measurements with the Papst fan:

Appraisal

It is probably the most original cooler in this survey. It provides good to average performances with a low noise level. Another good mark: performances aren’t really sensitive to ventilation and the performances gap is low when the voltage is reduced. In a manner of speaking, the thermosiphon is self-regulating. As long as the steam doesn’t condense, it hardly has an impact on the evaporator’s performances when it is placed on the processor (internal pressure increases slightly). A base with better finishing touches and not ridged would be an improvement to slightly increase its performances.

A thermosiphon development is always delicate because of the difficult choice of internal pressures. If it is chosen so that liquids boil at 30°C, performances will be good because the temperature won’t be increased much further (constant fluid temperature despite a change of state). If the ambient air temperature is warm, like in summer, when the temperature reaches 35 °C, an important part of the fluid will turn into steam. So the boiling point will be reached when the computer is stopped and we will have to wait for a new liquid/steam balance to occur in order to attain a higher boiling point than the ambient air, if possible, or else it will become 100% steam and that could be catastrophic. The hotter it is outside, the more dispersal capacity and utilizable fluid volume will be lost. To avoid this situation, you need to adjust the device with a boiling temperature higher than the worst situation you could envisage. However, that will restrict the range of uses.

Scythe – Kamakiri

Presentation

Six heat pipes characterize the Scythe Kamakiri and the 80-mm fan imprisoned in the blades. A plastic skirt with the company logo is also included to improve airflow guiding. The Kamikiri is compatible with all Sockets thanks to several installation sets.


There is a standard Molex input to ensure the power supply, as well as a potentiometer and a 3-pin connector to regulate the rotation speed. This potentiometer offers a possibility of adjusting the Voltage from 7 to 12V, and it is placed on an extension so as to be accessible from the back of the tower case.

Installation

The installation on the LGA775 platform is relatively easy. It starts with the support installation on the motherboard with the back plate, the two gears and screws. A tightening plate is used as the junction between the gears (notches) and the Kamakiri (screws).

Measures and results

As the potentiometer offers the possibility to adjust the voltage from 7 to 12 V, our tables include measurements for 7, 9 and 12V:

Appraisal

Good performances at full rotation speed but that fall apart below a voltage of +12 °C between 12 and 7V. This gap is huge compared to other coolers! We must also emphasize that it is one of the noisiest with almost 37 dB(A) at full speed… As the fan is stuck in the middle of the cooler, it disturbs the airflow and causes a lot more turbulence on account of the radiator blades being close to the fan blades on both sides. This phenomenon isn’t present on other coolers. To be avoided…

Swiftech - MCX775-V

Presentation

The Swiftech MCX775-V is one of the few cooler sold without ventilation. That isn’t because it can work without one, but in order to leave the choice up to the user. The MCX775-V is designed to work with an 80-mm fan, there are holes in each corners of the radiator angle so as to install one. This cooler stands out from other ones by a higher manufacturing quality.



The imposing 12-mm thick copper base is an example of this quality. No need of additional polishing, the base has an irreproachable flatness (0.007-mm tolerance) thanks to a very complete industrial surfacing/polishing industrial process. The air exchange isn’t entrusted to blades but to aluminum cylinders placed directly on the base. These cylinders are striated in a helicoid way to increase the exchange surface. The installation is made with a nylon screw set and springs. As is clear from the product’s name, it is only compatible with the socket 775, but Swiftech has released MCX775-V replicas for all current platforms.

Installation

The assembling starts with the installation of four nuts in the Socket corners. Then you only need to tighten the whole system by means of four nylon screws. This solution ensures accurate tightening. We just find the lack of a back plate unfortunate. This could avoid an overly important buckling of the motherboard when tightening.

Measures and results

Here are the results obtained with the two Papst used for the test:

Appraisal

The Swiftech solution doesn’t provide performances as good as coolers based on heat-pipes technologies, but still provides good to average performances according to the chosen fan. It requires sustained ventilation to provide great performances and then it becomes really good. Some people don’t hesitate to use 6000 Rpm fans with this monitor (watch your fingers…) Unfortunately, you almost always have to choose between noise or performance. The fact that the center of the radiator is quite dense and right in the dead area of the fan doesn’t help at all. No heat pipes transfer the heat to areas where the airflow is the most important (on the side).

Thermalright - XP90

Presentation

Developed by Thermalright, the XP90 has been very successful from the start. It features four heat pipes, a copper and nickel base and is sold without a fan, as is usual with this brand. It is, as the name indicates, designed to work with a 92-mm fan. The fan has to be installed on the top of the cooler by means of a set of piano strings. Initially, the XP90 wasn’t designed to work with the LGA775, but Thermalright sells a set to allow its installation on the new Intel platform. However, it is immediately compatible with the Sockets 478 and K8.

Installation

The LGA775 set includes one plate for the back of the motherboard, two plastic gears and a set of screws. The gears are installed on the motherboard thanks to the four screws. The only difficulty of the installation is the tightening of the two XP90 tongues to the gear notches. Two of these tongues are located under the blades, so it doesn’t make their access any easier.

Measures and results

We made a series of three measures: from 7 to 12 Watts with the Papst 3412 N/2GLE:

Appraisal

The XP90 intelligently uses the heat-pipes to transfer the heat to where it is easily dispelled (the side of the fan blades). The convection at this level is the best possible and improves condensation in the heat pipes. Aluminum blades are no longer a problem as the heat doesn’t have to travel far before being transferred to the air. It provides very good performances and at only 7V it turns in equivalent performances to the Swiftech S4 at 12V, and with a lower noise level! With the SE2, the XP90 becomes the first in terms of noise level and only loses 1°C. It is even better than the astonishing Papst 4412F/2GL at 5V. In the end it is an excellent cooler , very light, not too expensive but sold without a fan.

Thermalright - XP90C

Presentation

Released a few days ago, the 100 % copper Thermalright XP90 is now available. Except for the blade material and the lack of nickel in the base, the XP90C is strictly identical to the XP90 as regards compatibility, dimension, installation …As always with Thermalright the manufacturing quality is very good.

Installation

Like the XP90, for the installation on a socket T platform, the LGA775 set is required. For more information on the installation take a look at the XP90 installation paragraph.

Measures and results

Here are three measurement series from 7 to 12V made with the Papst 3412 N/2GLE :

Appraisal

Despite identical behaviour and a very similar design to the aluminium blade version, the XP90C is in fact more efficient with an identical ventilation for high voltages (9 to 12V). For lower voltages, the effect of the copper blades is less noticeable. The XP90C is probably the best compromise as regards performance and silence, because it is even ahead of the XP120.

Thermalright - XP120

Presentation

The Thermalright XP120 is the bigger brother of the XP90. Also sold without ventilation, it is designed to work with a 120-mm fan. Except for the additional heat pipe, the XP120 is identical to the XP90 whether in materials, installation system or compatibility. It is necessary to purchase a LGA775 (sold separately) to install it with a socket T.

Installation

The installation process is identical to the XP90. It is, however, a little more difficult to install because two of the four installation grooves are located under the blades.

Measures and results

Here are the results for the XP120 with the Papst 4412 F/2GL :

Appraisal

Same heatpipe process as the XP90 but one step above. The 120 is in the top 3 of the best radiators and since the Cool Scraper sensitivity to ventilation variation is low, you can even reduce the voltage to lower the noise level. Compared to the XP90, it is approximately 2°C better at full load for the whole range of uses with the best fan, but provides comparable performances to the XP90C. It is also a little bulkier.

Thermaltake – Big Typhoon

Presentation

Thermaltake recently released the Big Typhoon, which is the high end of the range with 6 heat pipes and the 120-mm wide fan. Designed for high performances, the Big Typhoon has, on paper, most of the assets required, it is compatible with all Sockets and features excellent finishing touches.

Installation

It is comparable to the Asus Star Ice (they have the same dispersal and installation systems): back and front plate, screws and nuts. The installation is, however, easier than the Star Ice thanks to wide clearings under the blades of the Big Typhoon.

Measures and results

We made four measures series from 5 to 12V:

Appraisal

From the performance point of view, the Big Typhoon is one of the three most efficient with a reasonable noise level (more than 25dB(A) less than the Star Ice). The Big Typhoon performance/noise ratio is excellent and it is even better than the XP90C for noise level over 39 dB(A). With the Big Typhoon, Thermaltek provides a very good solution and an interesting alternative to the XP90C.

Thermaltake – Sonic Tower

Presentation

Initially designed as a passive solution (without fan), the Thermaltake Sonic Tower is a good product for those who prefer silent computers. Even if we had included a passive cooler (which would provide mediocre performances but would be very silent), we decided to measure all temperatures with the optional 120-mm fan (identical to the one used with the Big Typhoon). To have an idea of passive performances, you just need to look at results obtained with low voltages (5 to 7V). The product’s finishing touches are very good: the available panel of Sockets is perfect.

Installation

The installation is identical to the Big Typhoon.

Measures and results

Appraisal

The Sonic Tower’s performances with the optional Thermaltake fan are pretty good. The U disposition of the blades increases performances when used as a passive solution. Adding the fan increased performances, but the blade position doesn’t offer the possibility of being as good as the Big Typhoon that was designed to work with this fan. It is then less efficient at 9 and 12V. With almost the same noise level, consequently the Sonic Tower is lagging behind the Big Typhoon.

For the passive use of such a cooler, we have to keep in mind that to obtain great performances, you will need an important fan tower ventilation. A weak ventilation coupled with a less important fan ventilation also provides a good alternative.

TTIC - NPH775-1

Presentation

TTIC, the column heatpipe manufacturer, sent me two coolers including the NPH775-1 available in the weeks to come. It is a product entirely made of copper with 100-mm high column and designed to work with an 80-mm fan. TTIC specifies that the coolers are available with or without the fan and that it is possible to install up to two fans. The NPH775-1 included in this survey was sent with an 80 mm Evercool fan. The NPH775-1 is only compatible with the socket T, but equivalent products also exist for the sockets A, 478 and K8.

Installation

The installation is made with a spring/screws system. First you need to place the plate at the back of the Socket and then tighten the screws down at each corner of the plate. Screws are designed to have optimum tightening when completely screwed down. So it avoids the problem of the NPH775-1 which was either too tight or not sufficiently so.

Measures and results

We made additional measurements to the fan provided with the Papst 8412 N/2GMLE :

Appraisal

With a supertube and copper blades, TTIC has a very good cooler here. It provides great performances and is amongst the first with the Evercool fan at 12V without being as good as the Star Ice in terms of noise level. It is, however, not quite as good as the XP120 with lower voltage. This is probably due to the lack of guidance for the low airflow through the blades...The possibility remaining of installing two silent fans on each side surely helps to improve performances. Indeed, we noticed that the deltaT slope in relation to the temperature increased rapidly. So an additional fan will also provide an important increase in temperature. This doesn’t occur with all coolers. It is also an excellent choice thanks to its compactness.

TTIC - SF775-2

Presentation

The SF775-2 stands out from the two other TTIC products because of the absence of heat pipes. It is in fact a standard 100% copper radiator. A 7 W 92-mm fan ensures ventilation, which is relatively more important when compared to the usual fans. Like with the NPH775-1, the installation system is only compatible with the T socket.

Installation

The installation system is identical to the NPH775-1.

Measures and results

The table below includes measurements with the SF775-2 from 5 to 12V.

Appraisal

This flower with very thin blades and totally made of copper, provides better performances than other flowers with copper cores and aluminum blades like the Intel Box. It is one of the best with radiators, and can compete with the XP120 and the other TTIC with columns. The offside of the coin is that it is also one of the noisiest with 66 dB(A) at 12V. This result is quite logical as the 92-mm fan spins at 4000 Rpm. Even after a strong voltage reduction to 7V, it is still unfortunately one of the noisiest. Unlike the TTIC NPH775-1, an additional fan won’t improve performances as the deltaT slope is quite flat around 12 V.

Zalman - CNPS7000-AlCu

Presentation

The first Zalman product tested was the CNPS-7000AlCu. It is an original radiator as the base is actually made up by tightening the blades together. The blades are then spread out to obtain a round cooler. In order to have a flat base, it is subjected to surfacing. So we cannot speak here of a copper base and aluminium blade coolers, since part of the blades (as well as the base, of course,) is made of aluminium and the other part of copper.


Ventilation is ensured by a 92 mm fan embedded in the cooler. The CNPS7000-AlCu isn’t initially compatible with the T socket. However, thanks to a set sold separately, it will be possible for the users who migrated from a socket 478 platform to an LGA775 platform to keep the same cooler. It is also adaptable to socket A and others K8 platforms. In addition to all the parts required for the radiator installation, Zalman also provides some thermal paste and a Fanmate for voltage adjustment from 5 to 12V.

Installation

As explained above, the socket T installation requires an additional set of spare parts. This set includes two metal plates and a set of screws, nuts and washers. The first plate needs to be placed at the back of the motherboard to strengthen it, and the second one on top for the radiator support. Both are tightened thanks to the screws and nuts. The CNPS7000-AlCu is then installed by means of the two tongues.

Measures and results

The table below includes measurements made with the CNPS7000-AlCu from 5 to 12V:

Appraisal

This is hardly surprising: it is the least efficient of the four Zalman tested here. Aluminium is at the origin of the weakness of its thermal conductivity, since there is not much copper in contact with the Pentium 4. Also, like its three other brothers, since it isn’t very sensitive to ventilation, increasing the fan Voltage from 7 to 12V only increases the noise level. It is possible to reduce the voltage to reduce the noise level (quite important around 12 V, to 7-8 V.

Zalman - CNPS7000-Cu

Presentation

A high-end product from the Zalman 7000 range, the CNPS-7000Cu is in fact the copper replica of the preceding cooler. Same design dimensions, ventilation and even installation system. Only the price, weight and materials differ. To use it with the socket T, you will have to buy the same adapter.

Installation

The installation process is strictly identical to the CNPS7000-AlCu.

Measures and results

Appraisal

Compared to the other AlCu cooler, the copper version is 2°C better for the whole ventilation range at full load. It isn’t surprising and this result is due to the copper. Thermal and sound behaviour are exactly the same as the AlCu version since the design and the fan are identical. We will also see a little further that this cooler is also far more efficient than the 120 mm AlCu (clearly bulkier and at the same price).

Zalman - CNPS7700-AlCu

Presentation

Zalman recently released a new cooler range. The 7700 series is very similar to the previous range but features a 120-mm fan and is consequently bigger. The design is identical. In terms of compatibility, the Zalman isn’t compatible with the socket A (end of life) but is initially compatible with the socket T additionally to the Socket 478 and several Athlon 64 / Opteron Sockets. Just like the 7000 series, the 7700 comes in two flavours: one copper/aluminum and one 100% copper. Let’s start with the CNPS7700-AlCu.

Installation

Very similar to the LGA775 adapter in the 7000 series, the CNPS7700 installation system uses a back plate and a metal support. Even though it was much bulkier, we didn’t have any compatibility issue with our test motherboard. However, we advise you to take a look at the Zalman compatibility list to make sure your motherboard is compatible.

Measures and results

Appraisal

This 120-mm AlCu is the useless version amongst the four tested. And the reason is that the 92 mm tested previously provides the same performances for an equivalent price, is less bulky and is less noisy for the whole range with almost 4 dB(A) less compared to the 120 mm! The 120-mm version performances don’t vary much with the ventilation, just like other products of the family. That means it is possible to considerably reduce the ventilation to obtain practically the same global performance.

Zalman - CNPS7700-Cu

Presentation

The Zalman CNPS7700-Cu is the completely copper version of the CNPS-7700-AlCu. It is totally identical and is approximately 70 % heavier (almost one kilo!). On average, you will have to spend an additional $10 to buy it.

Installation

The installation process is strictly identical to the copper/aluminum version.

Measures and results

Appraisal

This whole copper version is the champion of the Zalman family and is one of the five most efficient coolers at 12V. It is even in first position at 5V, way in font of the XP120. It is also less noisy than the XP120 with the Papst ! It is 2.5°C better than the 7000 Cu version and almost 5°C compared to the 7700 AlCu at full load for the whole ventilation range. This result is really excellent! The voltage isn’t the only aspect, and the XP90C + Papst duet provides better performances at 7V with a lower noise level. In the end, this cooler is an excellent choice for silent computer enthusiasts who are also seeking excellent air cooling performances.

Noise level

Noise level comparison

Previous pages showing detailed presentation of the 25 cooler tested here will be the basis of the following analyses and conclusions. In terms of noise levels, the rotation speed is only an indication to draw pertinent conclusion.

Measurements of noise level are, however, essential and are really representative of nuisances perceived by the user. This analysis is based on results obtained with these measurements.
The graph below presents the noise level for each of the 31 possible configurations and for the four main fans voltages. Data was then selected at 12V from the least noisy (above) to the noisiest (below). Each 150-mm step corresponds to 28 dB(A) of background noise.



Important :This classification is the one given for 12V. According to the fan, rotation speeds aren’t systematically proportionate to the voltage. Obviously, this classification would be different with a voltage of 5, 7 or 9 V. That means that it is therefore impossible to have an objective classification for each voltage. The next pages, however, introduce the concept of performance/noise ratio to emphasize the efficiency of each model to evacuate the most heat with the lowest noise level possible. So here you have to make up your own mind and compare all the coolers with the data provided above.

Raw thermal performances

Raw thermal performances(o/c@FULL)

This page is dedicated to pure raw thermal performances. The configuration, which dispels the most heat, is included to show the gap from one product to another. The Overclocked configuration features the following characteristics: 3760MHz (235x16) with 0.15 V of additional Vcore (1.51 V.). The FULL is generated by a 20 minute BurnP6 session.

Just like with previous graphs, the list is selected at 12V and the lowest is the best.



The leader in the classification isn’t surprising. Heat pipe products take first place. Except for the star Ice (the exception), the best ones feature coolers with wide fans. Star Ice’s first place, despite its modest 80-mm fan, is due to the high rotation speed that leads to high performances and airflow (but also to a very high noise level).

It is quite clear from this table that the classification would have been different with another voltage. That is due to the differences between the various fans tested here: some have linear functioning; others are big with low rotation speed, etc.

Performance index : Introduction

Performance index : Introduction

So far, measurements helped us to establish several independent classifications: noise level, pure thermal performances with the initial frequency in IDLE and with an Overclocked configuration in FULL. We could find other ways of selecting these coolers, but it is preferable to find a way of accurately classifying them.

So we created an index based on the level of performances and the noise level. Some of the users will seek silence (often to the prejudice of performances) at all costs, and others will look for high performances whatever the noise level, and finally the last ones will want to find the best compromise. This analysis is, in fact, the reason why I chose to make all these measurements in order to define a noise level and a temperature delta each time.

With all this data, it is possible to define the cooler behaviour law defined by: Delta T = f(Noise level). This index is then accessible with a graph to compare all the coolers.


The first noticeable thing is that it is possible to divide the graph in three areas A, B and C. This separation is in fact quite natural. After analyzing each area, we realized that each of them represents different functioning in terms of thermal or noise performances. The A area corresponds to products with a low temperature delta and a minimum noise level (the best coolers). The B area is the silent products’ range with a lower temperature than the A (less stable Overclocking temperature for example. And, finally, the C area corresponds to high noise level coolers (temperatures obtained are relatively proportionate to their noise level).

Performance index : areas

Performance index : areas

The A area includes all coolers combining performances and silence. It also corresponds to the most efficient area in terms of thermal dispersal. Technologies used in the C area, even if they offer a possibility of obtaining high performance levels, are soon powerless compared to the A area. The reason is that this technology has reached it own limits and that a wide fan won’t make any difference.

This is the reason why the graph shows the most interesting coolers in the market for raw performances as well as those with the best performance/noise level ratio.



Two products clearly stand out, the XP90C and the Big Typhoon. They share the first place (according to the noise level taken into considerations). If we totally take this behavior law into account, the XP90C is rather more efficient, even though the Big Typhoon provides the best performances at full load for an equivalent noise level. Some of the graphs have a tendency to increase (mathematically decrease) with low noise levels. That is due to the products’ difficulties to evacuate the heat when the airflow is too low: that is the case for the l’Hyper6+, the two Thermaltake and the NPH775-1 with the Papst.

Amongst all the coolers representing the highest-end Aircooling products today, we noticed the important position of heat pipe technology and the decrease in standard radiator products. As far as the Hyper48 is concerned, it is difficult to place it in the A or B area. If we only consider its functioning at full speed (9 or 12V), its noise level is inferior to most of its competitors’ and it provides very interesting performances. So the A area would be appropriate. However, if we examine, like with all the other coolers, its functioning from 5 to 12V, we notice average performances at 5 and 7 V. This clearly means that it requires important ventilation.

In the B area are the coolers that give preference above all to the noise level. The difference with the A area is that products are less efficient for evacuating heat with these low noise levels. Of course, this silence is obtained by relatively low rotation speed and even at 12V. Zone B products all feature 92 mm, or above, fans. As shown in the A area, an interesting solution to obtain a high airflow and a correct noise level would be to use wide fans (120 mm).

The following graph includes coolers with a noise level from 34 to 44 dB(A). The temperature gaps between the IHS and the ambient air varies between 28 to 44°C, according to the temperature.



The Hyper48 stands out, because, even if it has some difficulties in evacuating heat with low voltage, it is the best in this range of coolers. However, we find it unfortunate that Coolermaster hasn’t chosen a fan with a wider rotating speed range. This would have given an opportunity to users seeking performances to have a more efficient cooler at full speed.

The Calmera KS10, the only cooler to use a thermosiphon, also finishes amongst the first. We noticed the very linear temperature delta behaviour as regards the noise level. It provides very good performances for a low noise level (~ 34 dB (A)) compared to the rest of the coolers.
Finally, coolers working with the S1 are left behind because of a low airflow. The ventirad Box also provides the same level of performance, but isn’t the worst one as the Freezer 7 is very discreet in terms of noise level as well as in terms of performances…



The last area is the C, with noisier models than in the A or B areas. We know that thermal performances are proportionate to noise level, and the graph below shows this perfectly with the almost linear curves that have relatively similar slant. Products here are entry level products or more evolved products with a wide range of ventilation. These aren’t necessarily the least efficient but are simply adapted to high performances, as well as high noise levels a lot of the time.

The Star Ice provides the best performances (if we only stick to the thermal aspects) but it is penalized by the high noise level. This graph perfectly illustrates the impossibility for an extremely efficient product to compete with products with a considerably lower noise level.

Conclusion

Conclusion

Measurements and previous analyses helped us emphasize the main characteristics of each product. So you can make your own choice according to your own criteria. Some products do stand out, however.


Thermalright confirms its leadership with the XP90, 90C and 120. Amongst the three products, the XP90C is the one that provides the best performances but its counterpart is a high price at US$60 without fan with an optional LGA775 set. The XP120, at the same price, is slightly less efficient and is only a little bit lighter. But if you choose a product with lower performances it is best to take the standard XP90.

The Thermalright competitors aren’t left behind and we noticed Thermaltake’s comeback with the imposing SonicTower, as well as with the Big Typhoon, which is better than the XP90C and that also features lower noise levels (39 dB(A). Those who prefer performances to noise level and are looking for a reasonable product also have a wide choice. We only find its weight and the low availability in several countries unfortunate.

IMG0012609,R,0]Other coolers such as the CoolerMaster Hyper6+ also reduces the temperature delta between the ambient air and the processor temperature even if it has a higher noise level (still far from the completely delirious level of the ASUS Star Ice).

It is important to remind you that if the test platform of this survey was the LGA775, this article’s objective was to test most of the coolers on the market. Only a few of them are designed to work with a single socket. Most of them also exist in similar versions for other platforms.

Finally, we can say that the Aircooling is now back on top, thanks to the Heat Pipe technology, and that the manufacturers’ efforts to innovate and develop efficient coolers, keeping in mind the "silence" challenge, is the basis for the excellent performances attained by the best products. For the time being, it seems difficult to reach higher performances without a considerable cost or bulk increase. However, the release of dual core processors will push the manufacturers to devise even better solutions…