Intelâs enthusiast platform puts out the heatâ"here are seven ways to take it off
Intelâs Sandy Bridge-E enthusiast platform brings with it a new Intel socket, and that means new cooler mounting brackets! One nice thing about the X79 chipset: The boards ship with an integrated universal CPU mounting backplate, so no more fiddling around behind the motherboard.
Now, the new chipset hasn't been out that long, so cooler manufacturers are still working on getting their products compatible with the universal backplate. We were able to wrangle up seven LGA2011-compatible CPU coolers, ranging in price from $30 to $120, to see how they cope with the new mounting system and whether theyâre up to the task of cooling an overclocked Sandy Bridge-E CPU.
We tested each cooler on a test bed with an Asus P9X79 Deluxe motherboard, a i7-3960X CPU, 16GB of DDR3, and a Thermaltake Level 10 GT case with its stock fans running at âhigh.â The i7-3960X is the flagship Sandy Bridge-E CPU, and contains six cores and support for 40 PCIe lanes. Its Turbo mode hits 3.9GHz out of the box, but weâve clocked the proc up to 4.2GHz. We tested each cooler after an hour of idling and an hour of full burn at 4.2GHz, using the Sandy Bridge-E version of Intelâs internal thermal testing utility, and measuring the core temperatures with HWMonitor. To further level the playing field, weâre using the same thermal interface material for each cooler, eschewing the included thermal pastes in favor of Arctic Silverâs Lumiere, which weâve used for all our cooler tests for years. In a sidebar, weâll also talk about the effect of thermal paste on CPU performance.
Because very few of us run our CPUs at 100 percent across all cores for long stretches of time, the burn temperatures represent worst-case scenariosâ"this should be the hottest these CPUs ever get.
Cooler Master Hyper 212 Evo
Meet the new boss, same as the old boss
What, this old thing? Cooler Masterâs Hyper 212 Evo is the newâ'andâ'improved version of our standby CPU cooler. Itâs just $35 and offers performance far exceeding other coolers in its price range, so itâs the first thing we reach for when we build a new budget-conscious rig. Given that LGA2011 CPUs donât come with heatsinks, the Evo is the closest thing we have to a stock cooler, and it will be the standard against which all other Sandy Bridge-E coolers are judged.
The plastic fan clips on the Evo are much easier to use than the wire clips that originally came with the first Hyper 212-Plus.
When we reviewed the Hyper 212 Evo in the January 2012 issue of our magazine, we had two major complaints about it: The backplate mounting procedure remained annoying, and the cooler didnât have LGA2011 mounting support. Both of those problems have now been addressed.
The Hyper 212 Evo is a skyscraper-style heatsink, with four direct-contact copper heat pipes rising into an array of aluminum heat-dissipation fins. From the base of the heat exchanger to the tips of the heat pipes, the cooler is 6.3 inches tall, 3.13 inches deep, and 4.7 inches wide with the included 12cm PWM fan attached.
Thanks to the unified backplate on the LGA2011 socket (a crossover from Intelâs Xeon platforms), the Evoâs mounting process is greatly simplified. The LGA2011 X-shaped crossbar now screws directly into the mounting holes on the included backplate. No more accidentally unscrewing mounting bolts when removing the cooler.
In our test bed, the Evo continues to impress, cooling our i7-3960X at 4.2GHz down to 74 C. Itâs not the best cooler in this roundup, but it outperforms coolers three times its price. And although its height may be a concern in narrow cases, the heatsink and fan do not interfere with any RAM slots, unlike many other coolers in this roundup. The Evo remains an inexpensive, inexplicably kick-ass cooler.
| Dimensions HxDxW (inches, with fans) | 6.3x3.13x4.7 | 
| Weight | 1 lb, 4.6 oz | 
| Heat Pipes | 4 copper direct-contact | 
| Stock Fans | 1x 12cm PWM | 
| Addâl Fan Support | 1 (clips included) | 
Powerful, inexpensive performance
Not the freshest design; Mounting annoying on earlier brackets.
Xigmatek Gaia
A challenger appears!
Sometimes when we use the Hyper 212 Plus in a build we get comments to the effect of, âWhy donât you use Xigmatekâs Gaia? Itâs just as good and just as cheap!â Just as cheap? Definitely. Just as good? Weâll see!
Stop us if youâve heard this before: The Gaia is a skyscraper-style stack of aluminum cooling fins on top of three direct-contact copper heat pipes. The Gaia is 6.5 inches high by 2.9 inches thick (with the fan) and 4.9 inches wide. At one pound, 4.7 ounces, itâs practically the same weight as the Hyper 212 Evo. Aside from the slightly narrower cooling fins and the fact that it has three heat pipes rather than four, and its 12cm PWM fan is held on by rubber pegs rather than a plastic clip, the Gaia looks a lot like the Evo.
On LGA2011, the Gaiaâs install process is pretty easy. Four double-sided thumbscrews screw into the universal backplate, a mounting plate screws onto either side of the heatsink and mounts onto those thumbscrews, and then four nuts secure the mounting bracket. There are no screws or any other way of putting pressure on the heat exchanger other than tightening the mounting nuts as far as possible. The rubber fan mounts pop into slots cut into the sides of the cooling fins, which is great for damping vibrations but not as easy to attach or remove as plastic or wire clips.
We just realized that the Xigmatek Gaiaâs cooling fins resemble the letter X.
The Xigmatek Gaia cooled our overclocked i7-3960X to 75.7 C in our burn test, within two degrees Celsius of the Hyper 212 Evo. We still prefer the Evoâs installation procedure, at least for LGA2011, and donât mind the $5 price premium, but the Gaia is nearly as good as the Hyper 212 Evo, and even less expensive.
| Dimensions HxDxW (inches, with fans) | 6.5x2.9x4.9 | 
| Weight | 1 lb, 4.7 oz | 
| Heat Pipes | 3 copper direct-contact | 
| Stock Fans | 1x 12cm PWM | 
| Addâl Fan Support | 1 (rubber mounts included) | 
 
Almost as good as the Evo; even cheaper.
Annoying fan mounts; cooler mount doesnât apply much pressure
$30, www.xigmatek.com
NZXT Havik 120
NZXTâs second air cooler, and they still canât spell âhavocâ
NZXT didn't enter the CPU cooling game until quite recently. We reviewed its first cooler, the skyscraper Havik 140, in December 2011. The Havik 140âs dual 14cm fans helped it power to the top of our air-cooling charts, though the slightly cheap-feeling mounting bracket kept it from Kick Ass Award status. NZXTâs second air cooler is the smaller, less expensive Havik 120.
The NZXT Havik 120 isnât even close to the largest cooler in our roundup.
The Havik 120 stands 6.4 inches tall, 4.5 inches deep, and 5 inches wide with both 12cm fans strapped onâ"just about a quarter of an inch shallower and narrower than the 140. The Havik weighs more than two pounds, and has four fat heat pipes to the 140âs six. The Havik 120âs mounting mechanism uses the bars-and-crossbeam model from the 140, and installation is easy. The coolerâs two 13-bladed, 12cm fans have 3-pin connectors, and connect to one of two Y-splitters (one with a speed-limiting resistor). The fans attach to the cooling fins with rubber straps, which is not the most convenient mounting mechanism, especially when the motherboard is already mounted in the case.
The Havik 120 performed second-best of the five air coolers we tested, dropping the overclocked i7-3960X to 73.7 C at full burn. Using the included resistor splitter cable increased CPU temperatures by a few degrees but reduced noise to quite low levels.
We like that the Havik 120âs fin stack is high enough so the fans donât interfere with most RAM DIMMs, and that it performs well without being too loud. Itâs neither as quiet nor as effective as the 140, but itâs smaller. At $55, itâs a good midpoint between the $30 coolers in our roundup and those in the $80-100 range.
| Dimensions HxDxW (inches, with fans) | 6.4x4.5x5 | 
| Weight | 2 lbs, 1.1 oz | 
| Heat Pipes | 4 direct-contact | 
| Stock Fans | 2x 12cm 3-pin (w/ splitter / resistor cables) | 
| Addâl Fan Support | No | 
 
Good performance; two stock fans; easy install.
Rubber fan attachments can be tricky; fans can get loud at full blast.
$55, www.nzxt.com
Noctua NH-D14 SE2011
Cooling prowess as massive as its dual heatsinks.
When we reviewed the first-edition Noctua NH-D14 back in 2010, we praised its quiet performance, but our then-current test bed didnât put out enough heat to best showcase its cooling chops. Fortunately, our new one does. This coincides nicely with Noctuaâs release of the NH-D14 SE2011, which includes (gasp) LGA2011 support and updates the D14âs two fans to include PWM, or pulse-width modulation.
If you own an earlier Noctua cooler, the company will send you the LGA2011 mounts for free.
The new NH-D14 is just as massive as the old one. With both fans attached, the heatsink stands 6.5 inches high, 6.2 inches deep, and 5.5 inches wide and weighs more than two pounds, 12 ounces. The coolerâs six heat pipes rise into two sets of cooling fins, each the size of a lesser coolerâs entire allotment. A 12cm PWM fan clips to the front and a 14cm fan clips between the two sets of fins, both using Noctuaâs wire clips, which are easy to install and remove. Installation is easy; just remove the center fan, mount four double-sided thumbscrews to the unified backplate, install two mounting bars over those, secure them with thumbscrews, and tighten the two spring-screws on the D14 onto the posts on the mounting bracket.
Because itâs so massive, the NH-D14 blocks some RAM slots; we had to use low-profile RAM in order to fit the 12cm fan onto the cooler. But once that was accomplished, the NH-D14 achieved the lowest CPU burn temps of any air cooler in our roundup, at just over 72 C. When we added the resistor cables, the temperature increased to 76 C, but fan noise became almost imperceptibleâ"certainly quieter than the case fans.
Of the air coolers weâve tested on LGA2011 thus far, the NH-D14 takes the crown. If it werenât quite so massive, itâd be even easier to recommend, but it still garners our firm approval.
| Dimensions HxDxW (inches, with fans) | 6.5x6.2x5.5 | 
| Weight | 2 lbs, 12.8 oz | 
| Heat Pipes | 6 | 
| Stock Fans | 1x 12cm, 1x 14cm, PWM | 
| Addâl Fan Support | 1 | 
 
Massive performance; quiet; easy installation.
Massive footprint; front fan can interfere with RAM heat spreaders.
$85, www.noctua.at
Zalman CNPS12X
The latest Zalman heatsink looks cooler than it is
The CNPS12X might be Zalmanâs most eye-catching cooler, with two arrays of black-nickel-coated cooling fins and three 12cm fans to push air through them. And it is massive. Itâs 6.1 inches tall, 5.25 deep, and more than 6 inches wide, and weighs two pounds, four ounces. Itâs so big it overhangs the inner four RAM slots on our Asus P9X79 Deluxe test motherboard, requiring the use of RAM without tall heat spreaders. The six direct-contact heat pipes rise into two sets of cooling fins, with the front and rear fans nestled into their respective fins, and the middle fan in between the two sets. All three fans are controlled via a single 3-pin power connector.
Installation of the CNPS12X is a pain. Zalman has adopted the common mounting system of two parallel mounting bars with a perpendicular crossbar that bolts to each mounting bar and keeps the cooler in place. Alas, Zalmanâs system is flimsy and uses the same skinny hex bolts we complained about in our 9900MAX review, which require a special (included) hex wrench. The bolts are hard to access thanks to their position directly below the center fan, and donât seem to be able to apply much pressure to the heatsink
Whether because of this or simply the nickel finish on the entire heat exchanger, the CNPS12X performed poorly on our overclocked system, getting the CPU to just below 78 C. Lest we base the score on an isolated incident, we installed the cooler several times over several days and got the same performance.
Without the included resistor cable, the CNPS12X is quiet. Itâs very quiet with the resistor cable, and so if youâre not overclocking, the CNPS12X is an appealing and beautiful cooler. But given the flimsiness and frustration of the install process, we canât recommend it for overclocked LGA2011 cooling.
| Dimensions HxDxW (inches, with fans) | 6.1x5.25x6 | 
| Weight | 2 lbs 3.8 oz | 
| Heat Pipes | 6 direct-contact | 
| Stock Fans | 3x 12cm, 1x 3-pin connector | 
| Addâl Fan Support | No | 
 
Beautiful looks; quiet fans.
Frustrating install; lousy retention mechanism.
$100, www.zalman.com
Intel RTS2011LC
Underwhelming in the extreme
Intel isnât shipping stock heatsinks with Sandy Bridge-E CPUs, perhaps assuming that consumers of $600 and $1,000 CPUs are going to want performance a little better than Intelâs stock coolers have typically provided. Instead, Intel is offering a branded liquid-cooling loop as an optional accessory. The Asetek-built cooling loop features a glowing blue Intel logo and a bright blue LED on its single 12cm fan, but otherwise looks nearly identical to the Antec Kuhler 620, which was also built by Asetek (which also built AMDâs Bulldozer-branded liquid cooling loop).
Yet another single-fan all-in-one liquid CPU-cooling loop.
The RTS2011LC uses Asetekâs low-profile pump unit/heat exchanger and a 120mm radiator thatâs less than an inch thick and mounts in place of a stock case fan (typically the rear exhaust fan). The heat exchanger mounts to the LGA2011 socket using Asetekâs standard barb-and-ledge mounting system, although with big solid posts instead of the flimsy screws and plastic clips Asetek used to use. The PWM fan plugs into the pump, which plugs into the CPU_FAN header.
Alas, the coolerâs single fan and radiator were no match for any of the air coolers in our roundup save one. The RTS2011LC tied with the Zalman CNPS12X for last place, and its 12cm fan was irritatingly loud at full speed. For stock-clocked chips, the Intel cooler is fine. Itâs quiet when not fully ramped up, and it doesnât block any RAM slots, unlike the Noctua and Zalman coolers. But for an overclocked Sandy Bridge-E CPU, the Intel liquid cooler is louder and hotter than the Cooler Master, Noctua, Xigmatek, and NZXT coolers, and more expensive than all but the Noctua, which is just $5 more. If you want a great liquid-cooling solution for an overclocked LGA2011 chip, keep looking.
| Dimensions HxDxW (inches, with fans) | 6x2.6x4.75 | 
| Weight | 1.25x2.5x2.5 (circular) | 
| Heat pipes | 1lb, 12.5 oz (total) | 
| Stock Fans | 1x 12cm PWM | 
| Addâl Fan Support | 1x | 
 
Easy install; good build quality; quiet at low temperatures.
Loud at high temps; doesnât cool as well as a $30 air cooler.
$80, www.intel.com
Corsair H100
Double your radiator space, double your fun
Another all-in-one liquid-cooling loop! Hooray! Corsairâs H100 is its fifth liquid cooler; after two with Asetek, the company has put out three with CoolIT. The H60 is your standard 120mm radiator-with-single-fan, the H80 is the double-thick double-fan version, and the H100, the first Corsair liquid-cooler to support LGA2011, is its first cooler with a 240mm radiator.
The H100âs radiator is around an inch thick and 10.8 inches long and fits in any case that can accommodate a 240mm radiator, though some cases may not have the vertical clearance to mount the fans inside the case. The pump/heat exchange unit is square, and very slightly taller than Asetekâs. It contains four 4-pin PWM headers to control the radiator fans, as well as a connector for Corsairâs Link system control software/hardware combo (sold separately). Thereâs also a three-speed fan-control button on top of the pump. The pump unit itself has a 3-pin motherboard fan connector and a 2-pin Molex for power. The cooler unit mount is simple; four double-sided thumbscrews mount to the unified backplate, the brackets at the corners of the pump unit slide onto those, and more thumbscrews secure them in place.
The H100âs 240mm radiator supports up to four 12cm fans, and the pump unit
We mounted the radiator to the top panel of our Thermaltake Level 10 GT test bed, with the fans on the inside pushing air through the radiator out of the case. In this configuration, with the H100âs fan speed set to medium, the Corsair cooler lowered our CPUâs burn temperature to 69.2 Câ"the best of any cooler in our roundup. Even with the fans at their lowest setting, the H100 was a match for the air coolers in the roundup.
Its 240mm radiator wonât fit in every case, and itâs the most expensive in our roundup, but the H100 is a competent, quiet liquid cooler and the best LGA2011 cooler weâve yet tested.
| Dimensions HxDxW (inches, with fans) | 10.8x2x4.7 | 
| Weight | 1.6x2.6x2.6 (square) | 
| Heat pipes | 2 lbs, 2.5 oz | 
| Stock Fans | 2x 12cm, 3-pin | 
| Addâl Fan Support | 2x | 
 
Quiet; excellent performance; easy install; supports up to four fans.
Wonât fit in every case; need to open case to adjust fan speed; expensive.
$110, www.corsair.com
Keep Your Cool
This is the biggest cooler roundup weâve done in years, and it was definitely enlightening. First, we really appreciate Intel finally bringing a unified backplate to a consumer-level CPU socket, and hope the trend continues, as it makes cooler installation much less irritating. However, some mounting systems still manage to be frustrating. Zalmanâs CNPS12X mounting bracket is likely responsible for its poor cooling performance.
The most interesting thing about our test results is that price and performance arenât strongly correlated. The cooler with the fourth-best performance, the Hyper 212 Evo, is just $35, and only 5 C hotter than the best cooler, which costs four times as much. Heck, itâs better than the Zalman, which is $70 more expensive, and the Intel liquid cooler, which is $45 more. Both the Cooler Master and Xigmatek budget coolers earn our recommendation for low-to-medium overclocks.
We got the best performance in this roundup from Corsairâs 240mm-radiator H100, and if you have the room in your case, youâll find it a quiet, effective solution that doesnât block any RAM slots. Adding a couple more fans will only make it better.
NZXTâs Havik 120 and Noctuaâs DH-14 are both great coolers, as well; the Noctua is much larger and quieter (though youâll need low-profile RAM), while the Havik is more compact and cheaper and shouldnât conflict with your RAM. We bet it hits the sweet spot for a lot of users.
Weâd recommend either the Havik 120 or the DH-14 over the Intel liquid cooler, or any single-fan, single-radiator liquid cooler. Theyâre cooler and quieter and they fail better; an all-in-one with a broken pump isnât a cooler, but a massive stack of cooling fins with a broken fan is still a massive stack of cooling fins. And in the end, isnât that what matters?
| CM Hyper 212 Evo | Xigmatek Gaia | NZXT Havik 120 | Noctua DH-14 | Zalman CNPS12X | Intel RTS2011LC | Corsair H100 | |
|---|---|---|---|---|---|---|---|
| Ambient Air | 23.8 | 25.1 | 25 | 25.2 | 24.2 | 25 | 25 | 
| Idle Temperature | 36.2 | 35.5 | 34.8 | 34.1 | 35.2 | 35.2 | 33.2 | 
| Burn Temperature | 74 | 75.7 | 73.7 | 72.3 | 77.8 | 77.8 | 69.2 | 
All temperatures in degrees Celsius. Best scores bolded. All tests performed using an Intel Core i7-3960 at 4.2GHz, on an Asus P9X79 Deluxe motherboard with 16GB DDR3/1600, in a Thermaltake Level 10 GT with stock fans set to High.
Does Thermal Paste Matter?
Thermal paste is important. Itâs what fills all the microscopic gaps between your CPU and cooler so that as much heat as possible can escape from your CPU. But does a particular thermal paste matter? While our Sandy Bridge-E test bed was occupied testing coolers, we put an X58 system to work on thermal pastes.
We took a Core i7-975 CPU and overclocked it to 3.9GHz on an Asus P6X58D Premium board in an open-air test bench. We tested 17 thermal pastes with a Thermaltake Frio OCK with its fans set to maximum, first at idle and then after several hours of full burn. This test bed wound up being far hotter than the more-efficient Core i7-3960X in our Sandy Bridge-E machine, and the heat really brought out the differences between the thermal pastesâ"to the tune of a 10 C delta between the best and worst pastes in our roundup.
The difference between the thermal pastes was less pronounced on our Sandy Bridge-E test bed. We took the Noctua NH-D14 SE2011 and tested it first with the Lumiere, as we did in the roundup, and next with the NT-H1 thermal paste that came with it, which performed very well in our thermal paste roundup. We saw an 8 C difference between the Lumiere and the NT-H1 with the Frio OCK on our X58 test bed, but on the X79 system with the Noctua cooler the difference was negligibleâ"well within the 2 C margin of error. This could be due to the lower overall temperatures of our X79 system, the Noctuaâs mounting system putting more pressure on the CPU heat exchanger, or airflow within the case itself.
So does thermal paste matter? It depends on how hot your CPU getsâ"the cooler the CPU, the lower the difference between pastes. Itâs only at the high end that the differences get pronounced. In that case, youâd be better off with the 11 pastes that get our Geek Tested & Approved stamp than with the pastes that donât. For the full report on our thermal paste roundup, click here.
| Geek Tested & Approved | Idle Temperatures | Burn Temperatures | |
|---|---|---|---|
| Tuniq TX-4 | Y | 40 | 79.5 | 
| Tuniq TX-2 | Y | 41 | 80 | 
| ShinEtsu MicroSI X23-7783d | Y | 40 | 80.25 | 
| Prolimatech PK-1 | Y | 41.25 | 80.5 | 
| Arctic Cooling MX-4 | Y | 41 | 81 | 
| Arctic MX-2 | Y | 40.75 | 81 | 
| Noctua NT-H1 | Y | 41 | 81.25 | 
| Xigmatek PTI-G4512 | Y | 40 | 81.25 | 
| ZeroTherm ZT-100 | Y | 41 | 81.5 | 
| CM Thermalfusion 400 | Y | 41 | 81.75 | 
| Arctic Silver 5 | Y | 41.5 | 82.5 | 
| Xigmatek PTI-G3606 | N | 42.25 | 84.5 | 
| Rosewill RCX-TC090 Pro | N | 41.75 | 85.5 | 
| Arctic Silver Arctic Alumina | N | 42.75 | 85.5 | 
| Arctic Silver Lumiere (zero-point) | N | 43.75 | 89 | 
| BioStar Nano Diamond | N | 43.25 | 89 | 
| Zalman ZMSTG1 | N | 43.25 | 89.25 | 
Best scores bolded. All tests performed on an overclocked Core i7-975 @3.9GHz (burn) and 2.1GHz (idle) on an Asus P6X58D Premium motherboard with 6GB Corsair XMS3 DDR3, a Radeon HD 5850, and 850W Antec TruePower PSU. The CPU cooler used was a Thermaltake Frio OCK with its fans set to maximum speed. Temperatures recorded after one hour at idle and after three hours of full-burn testing using Intelâs internal Nehalem stress-testing utility. We use HWMonitor to determine core temperatures and TMonitor to keep an eye on clock speeds.
NOTE: This was taken from the April issue of the magazine.
Comment