by Sergey Lepilov
02/02/2012 | 05:11 PM
Almost three years ago we upgraded our testbed built on an LGA 775 Intel Core 2 Extreme Q9650 processor with a new LGA 1366 platform with an Intel Core i7-920 CPU, which another 18 months later was replaced with a much hotter Intel Core i7-980X Extreme Edition. Back then in our review, which has become sort of a symbolic threshold, we tested thirteen highly efficient cooling products. Well, let’s continue with the good old tradition, and now that we have upgraded our platform yet again to the new LGA 2011 with Intel Core i7-3960X Extreme Edition CPU inside, we are going to see how five of the most efficient contemporary coolers will fight for the leadership. Besides that we are going to check out the cooling efficiency and acoustics performance of the default cooler that is bundled with extreme six-core Intel processors.
Some boxed Intel Core i7-3960X Extreme Edition processors were bundled with Intel RTS2011LC Water Cooler system. In fact, it wasn’t really necessary, because the power consumption of the new processors didn’t really increase compared with the power consumption of the previous-generation Extreme Edition modifications. However, in our opinion, they were primarily trying to respond to their eternal competitor – AMD, which was known for offering the users their top AMD FX processors in a bundle with a liquid-cooling system. Let’s meet Intel RTS2011C Water Cooler.
Inside the white cardboard box there is a special internal casing with section for all components of the cooling system:
Among the included accessories there are retention kits for all contemporary Intel platforms (even including LGA 775), installation instructions guide, two packs of thermal paste and plastic ties:
The system itself is built around the Asetek 570LC platform, which our regular readers should already be well familiar with. It is an exact copy of such liquid-cooling systems as Antec KÜHLER H2O 620 or Corsair H50:
The system consists of a copper water block, a pump with a ceramic bearing, two flexible tubes and an aluminum radiator with a plastic fan. Intel RTS2011LC Water Cooler is a very simple and relatively light-weight cooler. We didn’t find anything new about it, that we haven’t yet seen in the similar systems reviewed before.
The only difference is that the pump is marked with the Intel logo, though the logo of the true manufacturer is written right below it, too:
The thermal paste imprint on the base of the water block turned out incomplete and smudged:
It is in part because the CPU heat-spreader has a little bump in the center, and partially because you have to rotate the water block when removing it.
The system comes with one seven-blade fan with blue LEDs:
The fan rotation speed is automatically adjusted using PWM method in the interval from 1000 to 2200 RPM. We didn’t find any other useful information about this fan (the rotor sticker doesn’t have anything else of importance on it).
You can install Intel RTS2011LC Water Cooler water block using the universal retention frame with removable plastic caps in the corners:
The radiator is attached to the back of the system case with the fan installed between the radiator and the system case back panel. Its airflow is directed outside of the case:
The Intel logo and the message on the water block glow blue when the system is up and running:
Although the Austrian Noctua Company has recently announced new cooler models, they still have only one tower super-cooler in their product range - Noctua NH-D14 SE2011, which now also supports LGA 2011 platform. This is the cooler that will participate in our comparative testing today.
There are absolutely no constructive differences between this model and the original Noctua NH-D14:
The only thing I would like to point out at this point is that the cooler fans now support PWM rotation speed control function and can automatically adjust their rotation speed in 300-900 RPM or 300-1300 (1200) RPM intervals.
The well thought-through retention with very secure hold guaranteed impeccable contact with the processor. This is one of the today’s best thermal paste imprints:
This is what the new Noctua NH-D14 SE2011 looks like inside the system case with new configuration:
The old-new Noctua cooler is priced at $89.99. Note that SE2011 version is only compatible with the LGA 2011 platform.
Relatively new Phanteks PH-TC14PE super-cooler has recently been reviewed on our web-site. It demonstrated remarkable efficiency and defeated all previous leaders without exceptions. Massive dual-array heatsink on five 8-mm heatpipes and two 140-mm fans with uniquely shaped blades combined with a very reliable and smart retention mechanism didn’t leave any chance for competitors.
Will Phanteks PH-TC14PE be able to shine in our today’s test session, too? Well, we are about to find out shortly. So far the contact between the cooler base and the processor heat-spreader turned out to be not only pretty far from ideal, but also far not the best among the today’ testing participants:
However, these imprints are definitely not the worst either.
All the coolers are positioned a little bit differently on the new platform, because of the processor socket placement the coolers have been moved away from the back of the system case and closer to the center of the PCB. Phanteks PH-TC14PE is no exception:
We would like to remind you that the silver Phanteks PH-TC14PE is priced at $89. The price of other three color modifications is about $10 higher.
Thermalright Archon is the good old veteran of the cooling product competitions. In pure efficiency this cooler was the absolute best until products like Phanteks PH-TC14PЕ, NZXT Havik and Zalman CNPS12X came out. And in terms of efficiency-t-noise ratio, it is most likely still the best up until now due to an excellent Thermalright TY-140 fan. The company has recently released the “A” revision of this cooler, which features a 150 mm fan and an LGA 2011 retention kit and it is already on the way to our lab as we speak. In the meanwhile let’s check out the sane good old Archon once again:
The convex cooler base and the convex CPU heat-spreader didn’t allow us to achieve perfect contact between the cooler and the CPU, which you can clearly see from these uneven imprints:
In fact, my colleague had exactly the same problem while working on his Asus P9X79 Deluxe review with a completely different Thermalright Archon unit.
Thermalright Archon is installed onto LGA 2011 mainboards with four small bushes, which can be purchased separately, and is very similar to the installation process for the other coolers:
The MSRP for Thermalright Archon is $79.
Another super-cooler from Thermalright participating in our today's test session is Thermalright HR-02 Macho, which we also reviewed not so long ago. Being a budget reincarnation of the well-known HR-02, Macho proved to be an excellent product and became the least expensive super-cooler available at the MSRP of $39.95.
The best contact between the cooler and the CPU occurs in the very center, as you can clearly see from the following imprints:
When we tried to install Thermalright HR-02 Macho onto our Intel Siler DX79SI mainboard, we discovered that if the retention frame was positioned “correctly” (that is when the airflow is directed towards the back of the system case), the cooler interferes with the heatsinks on the processor voltage regulator components and therefore cannot be screwed on to the PCB. For that reason we had to rotate the cooler and the retention frame by 90 degrees. As a result, this is what we ended up with:
Other than that we experienced no problems with the installation and positioning of the Thermalright HR-02 Macho cooler on our LGA 2011 testbed.
The newest participant in our today’s test session is Zalman CNPS12X. This one-kilogram cooler turned out a real success in many respects: very efficient, quiet and very attractive-looking, though all of that comes at a pretty high price of $99:
The contact between the cooler W-DTH base and the CPU heat-spreader didn’t turn out too good either:
Looks like I will have to lap and polish the heat-spreader of my Intel Core i7-3960X extreme Edition processor after all.
Zalman CNPS12X fit perfectly fine into the new system case and onto our test mainboard. However, you should keep in mind that if you are using memory modules with tall heat-spreaders, you won’t be able to install them into the closest DIMM slots.
In addition to our brief recap of the testing participants’ features, we would like to some them up side by side in the following table (except for the Intel liquid-cooling system):
I have to say a few words about the upgrades of our testbed. First of all, it is now built around Intel Siler DX79SI mainboard with LGA 2011 processor socket based on Intel X79 Express chipset:
Even though the BIOS of this particular mainboard is pretty peculiar, it is fully-functional and allows not only to skillfully “juice” the CPU dry, but also ensure that it remains stable at the highest clock frequencies. We used BIOS version 0308 from November 28, 2011.
The second significant upgrade is, obviously, the CPU. Our Intel Core i7-3960X Extreme Edition processor manufactured using 32 nm process is an engineering sample made in
The typical TDP for this processor at its nominal clock is declared to be 130 W, and the maximum temperature upon exceeding which thermal throttling kicks in is set at 90°C (confirmed empirically). The contact surface of the CPU heat-spreader is 38x38 mm:
According to the following schematics, the processor die is 19.4x14 mm big and it is positioned along the processor socket (parallel to the DIMM slots, in other words):
Therefore, one can assume that for maximum efficiency the cooler should be installed with the heatpipes going across the die, i.e. perpendicular to the memory slots. By the way, the die dimensions on the illustration above taken from an official Intel document do not correspond to the official declared die size of 435 mm2: according to the dimension on that document the die size adds up to 271.6 mm2, not 435.
Nevertheless, we didn’t check out whether this assumption is correct in our today’s test session because in order to correctly estimate the dependence of cooling efficiency on the cooler position on the CPU we need to have perfect contact between the cooler base and the CPU heat-spreader in all positions. As you have already seen, none of the testing participants can actually boast that.
So, we performed all cooler tests inside a closed system case. Here is our testbed configuration:
For initial tests and summary diagrams we overclocked our six-core processor with the clock generator frequency set at 125 MHz, the multiplier set at 35x and “Load-Line Calibration” enabled to 4.375 GHz. The nominal processor Vcore was increased to 1.385 V in the mainboard BIOS. After that we tested all super-coolers at an even higher frequency and voltage settings. Turbo Boost was disabled during this test session, and Hyper-Threading technology was enabled. The memory voltage was at 1.65 V and its frequency was 1.666 GHz with 8-8-8-16_1T timings. All other parameters available in the mainboard BIOS and related to CPU or memory overclocking remained unchanged.
All tests were performed under Windows 7 Ultimate x64 SP1 operating system. We used the following software during our test session:
So, the complete screenshot during the test session looks as follows:
The CPU was loaded with two consecutive LinX test runs with the settings as indicated above. The stabilization period for the CPU temperature between the two test cycles was about 8-10 minutes. We took the maximum temperature of the hottest CPU core for the results charts. Moreover, we will also provide a table with the temperature readings for all cores including their average values. The ambient temperature was checked next to the system case with an electronic thermometer with 0.1 °C precision that allows hourly monitoring of the temperature changes over the past 6 hours. The room temperature during our test session varied between 24.0-24.5°C.
The noise level of each cooler was measured between 1:00 and 3:00 AM in a closed room about 20 m2 big using CENTER-321 electronic noise meter. The noise level for each cooler was tested outside the system case when the only noise sources in the lab were the cooler and its fan. The noise meter was installed on a tripod and was always at a 150 mm distance from the cooler fan rotor. The tested cooling systems were placed at the edge of the desk on a sheet of polyurethane foam. The lowest noise reading our noise meter device can register is 29.8 dBA and the subjectively comfortable noise level in these testing conditions was around 36 dBA (do not mix it up with low noise level). The fan(s) rotation speed was adjusted in the entire supported range using our in-house controller by changing the voltage with 0.5 V increment.
Besides their default fans, all coolers were also tested with two Scythe Slip Stream 140 fans at 800/1000/1300 and 1750 RPM speeds. The Intel RTS2011LC Water Cooler system was tested with two Thermalright TR-FDB-2000 fans. The rotation speed of all fans was controlled suing the same controller with ±10 RPM precision.
Well, it is time to check out the obtained results.
The results of our cooling efficiency tests are summed up in the diagram and table below:
Even at 1620 RPM fan seed Intel RTS2011LC Water Cooler system quite predictably was the last in the ranks allowing the CPU temperature to reach the critical 90°C. AT the maximum speed of 2170 RPM this system was as efficient as Noctua NH-D14 SE2011 at 800 RPM or Thermalright Archon at 1000 RPM. However, it is much more interesting to follow the competition between the air coolers here.
Phanteks PH-TC14PE took the lead here, and his efficiency on the new platform was so superb that even at 800 RPM speed of its two 140-mm fans it yielded only to one single cooler. And it wasn’t Thermalright Archon or Noctua NH-D14 SE2011, as you may have expected, but Zalman CNPS12X, which was the second best in this test, according to the overall scores. As for Archon and NH-D14 SE2011, they go neck and neck at the maximum fan speed as well as at the medium speed of 1000 RPM. At the minimal 800 RPM Archon yields 2°C to NH-D14 SE2011, although we should also keep in mind that the Thermalright cooler has only one fan against Noctua’s two. Nevertheless, we will split the winning third place between these two coolers without hesitation. The fourth cooler in this race is the most inexpensive of all – Thermalright HR-02 Macho. While it lost only 1°C to Archon at 800 RPM, 2°C – at 1000 RPM and 3°C – at 1290 RPM, this cooler once again proved that it indeed boasts unprecedented combination of efficiency and price (and noise, too).
As soon as we overclocked our processor to 4.5 GHz at 1.405 V Vcore, some of the testing participants suffered inevitable losses. I am sure Intel RTS2011LC Water Cooler failing in this test is not a surprise at all. However, Thermalright HR-02 Macho and Archon failed this test at 800 and 1000 RPM, and Zalman CNPS12X couldn’t cope with the overclocked processor at 800 RPM fan speed. All this is clear evidence of how greatly the heat dissipation of an overclocked processor increases. Take a look at the new results table and diagram:
The leader remains the same – Phanteks PH-TC14PE. Zalman CNPS12X also looks pretty good, although it loses to Phanteks 5°C at maximum fan speeds and 4°C at medium fan speeds. Thermalright Archon is not too far behind Zalman CNPS12X (as you remember, Archon has another trick up its sleeve: the ability to take on a second fan, while the Zalman cooler has already exhausted all optimization resources). Noctua NH-D14 SE2011 does pretty well against Archon and Macho due to its two fans, that is why it can cool the CPU properly even at 800 RPM fan speed. However, Phanteks PH-TC14PE, which design is very similar to that of Noctua NH-D14, is as much as 7-8°C more efficient than its Austrian brother. This is a very serious advantage in the super-cooler segment, where the competitors fight for every degree.
In conclusion to this part of our test session where all coolers were tested in their default configurations we can state that Phanteks PH-TC14PE and Zalman CNPS12X managed to cope with an even higher overclocked processors working at 4.625 MHz frequency with 1.455 V and 1.445 V voltages for the Phanteks and Zalman coolers, respectively. At the same time, despite higher voltage Phanteks managed to cool our overclocked processor 2°C better:
Phanteks PH-TC14PE (1230 RPM x2)
Zalman CNPS12X (1190 RPM x3)
Unfortunately, the remaining three tower-coolers failed during this overclocking test.
Now let’s see how the coolers will do with two identical Scythe Slip Stream 140 fans. As we have already mentioned earlier, we used Thermalright TR-FDB-2000 fans for Intel RTS2011LC Water Cooler. For obvious reasons Zalman CNPS12X participated the way it is. Again, let’s check out the results table and the diagram:
As we see, Phanteks PH-TC14PE demonstrates the same results as before except for one test mode at the maximum rotation speed of the two Scythe Slip Stream 140 fans. The results of the dual-tower Noctua NH-D14 SE2011 also remained almost unchanged. However, Thermalright Archon and Thermalright HR-02 Macho have become noticeably more efficient when we replaced their default fan with two alternative ones. The coolers ranked the same in all four speed modes: Phanteks PH-TC14PE being number one, Thermalright Archon following 3-4°C behind, Thermalright HR-02 Macho falling another 1-3°C behind, and Noctua NH-D14 SE2011 as the last cooler losing 1-3°C to Thermalright Macho.
Although we couldn’t modify Zalman CNPS12X in any way, this cooler looks very confident in this test even with its default fans, while all other testing participants received an efficiency “booster” in the form of two high-speed Scythe Slip Stream 140 fans. As for the compact liquid-cooling system, Intel RTS2011LC Water Cooler tested with two Thermalright fans attached to its aluminum radiator, it doesn’t look too impressive against the background of the tested super-coolers, even though it coped with our CPU overclocked to 4.5 GHz (even at 1000 RPM fan speed).
Now that we have transitioned to the new testbed, we will start feeding data into our summary table and diagram all over again. These results are obtained from coolers tested in their default configurations, in quite mode and at maximum fan speed with the processor overclocked to 4.375 GHz and Vcore set at 1.385 V:
* - The peak temperature of the hottest CPU core is posted on the diagram
taking into account the difference with the current ambient temperature
and is reduced to 25°C.
We will also rate cooling systems according to the maximum CPU overclocking that they allow. At this time this chart includes only five coolers and one liquid-cooling system:
The only thing left to address at this point is the acoustic performance of the tested cooling products.
We measured the noise from our testing participants in the entire supported speed range of their fans following the methodology described above. The results are summed up on the following graph:
The most efficient Phanteks PH-TC-14PE cooler also turned out the noisiest one among five air coolers tested today. Zalman CNPS12X and Noctua NH-D14 SE2011 are significantly quieter than the Phanteks cooler, but Noctua’s fans seem subjectively softer than Zalman’s. And the quietest ones in our today’s test session are again Thermalright Archon HR-02 Macho equipped with identical TY-140 fans. Our regular readers should already be used to this overall picture and it doesn’t depend on the platform type and configuration. As for the liquid-cooling system from Intel, although it is not present on the graph, we have to point out that it has a rather noisy fan, which is louder than any other in our today’s test session, and a quiet pump.
I can’t say that the super-cooler tests in the new platform with an LGA 2011 processor produced any unexpected results or drastically changed the alignment of forces among them. The obtained results correlate well with what we have obtained previously on our LGA 1366 platform. But at the same time, we can’t say that nothing has changed. Larger processor heat-spreader sets stricter requirements to the size and finish of the cooler base plate, contact quality and retention pressure. At the same time increased processor heat dissipation during overclocking produced bigger difference in temperature not only between super-coolers, but also between the best of the best. Taking into account all of these factors we hope that the new platform will make it possible to perform even more objective tests of the cooling systems in all price segments.
As for the super-coolers tested in our today’s review, Phanteks PH-TC14PE is indisputably the best one in pure efficiency, but it has a very serious drawback – it is the noisiest cooler of the five tested today. Ideally, Phanteks PH-TC14PE should be equipped with two quiet Thermalright TY-140 fans, then it will have absolutely no competitors among air coolers of any kind. The second most efficient cooler is Zalman CNPS12X, and here we don’t have any complaints about the noise. As you remember, we managed to overclock our test processor to the same maximum under Zalman CNPS12X, as we did under Phanteks PH-TC14PE. No doubt, this Korean cooler is a true success.
The third place goes to Thermalright Archon and HR-02 Macho. Although Archon proved 1-3°C more efficient than Macho, it sells for only half of Archon’s price, while their noise level is the same. This allows us to rank this cooler as high as Archon. The fourth place is taken by Noctua NH-D14 SE2011, but this isn’t the outsider (there are no outsiders in this test session, except for the Intel RTS2011LC Water Cooler, of course). Unlike the same Archon, this Austrian cooler is equipped with two quiet fans by default (now with PWM control), which make it just as efficient as Thermalright Archon with one default fan.
Now all we have to do is wait for the new super-coolers to hit the streets, so that we could check them out on the new Intel Core i7-3960X extreme Edition platform.