Recommended Heatsinks

Cooling | Reference|Recommended
Viewing page 1 of 4 pages. 1 2 3 4 Next

Update History

  • Sept 21, 2014 - NZXT Kraken 61 added.
  • Sept 16, 2014 - Big changes - another complete overhaul.
  • Dec 13, 2010 - Minor corrections, added Thermalright Ultra-120 eXtreme Rev C.
  • July 10, 2010 - Minor corrections, added Scythe Samurai ZZ & Grand Kama Cross.
  • May 20, 2010 - Complete overhaul.
  • June 1, 2009 - Prolima Megahalems, Scythe Ninja 2, Thermolab Baram added. A few older models retired.
  • June 24, 2008 - Zerotherm Zen and Noctua NH-C12P added. Zalman Reserator1 removed.
  • June 9, 2008 - Thermalright HR-01 Plus, four Xigmatek models, and Scythe Zipang added.
  • Mar 22, 2008 - Ninja demoted one point.
  • Jan 16, 2008 - Long overdue update, many additions and retirements. New ones on the lists in light blue/violet.
  • Aug 21, 2006 - A handful of new coolers added.
  • Jan 29, 2006 - First update in a year, much needed, many changes.
  • Jan 23, 2005 - Added several new models.
  • Sept 22, 2004 - Added Thermalright XP-120, Scythe NCU-2000
  • Aug 25, 2004 - Complete overhaul of recommended HS tables, many new models added, some retired. Ranking system changed to make use of full 1-10 range.
  • Aug 7, 2003 - Zalman 7000 Cu/AlCu, Scythe Kamakaze & Arctic Cooling SS 4 Pro TC added. All coolers for current CPU combined into one table -- too many multi-platform models made for much repetition.
  • June 17, 2003 - Arctic Cooling SS Pro TC added; Swiftech MCX models downgraded - recent testing shows very good performance with high (read: noisy) airflow but mediocre cooling with low airflow. There are better choices for less money for quiet cooling. Rankings reshuffled a bit based on recent work.
  • Feb 24, 2003 - Thermalright SLK-900U and Zalman ZM80A added
  • Dec 3, 2002 - Zalman 6500al-cu and Thermalright AX478 added; other changes to P4 list based on P4 HS roundup.
  • Oct 29, 2002 - Changes to Socket-A list based on Socket-A HS roundup.
  • July 17, 2002 - First published.

Recommending quiet heatsink/fans for CPUs has become both simpler and more complex in recent years. On the one hand, fierce competition has created a large pool of high performance CPU coolers that need little if any user modifications for effective and near silent operation, and almost any of the top performers that we review will do fine for silent or near-silent computing. On the other hand, the thermal specs (TDP) of current CPUs spans from under 30W to over 140W, so it makes sense to match the cooler to the TDP of the specific CPU, not wasting space and money for a huge heatsink when a smaller one will do fine. There's also a bigger range of cases sizes in wide use, from huge Extended-ATX towers to tiny mini-ITX desktops. To accommodate all this, we established two different CPU test platforms some years ago, one for large cooler and a second one for smaller coolers.

Our approach to apprasing heatsinks is summarized below. For full details, please read the article about SPCR's 2010 Heatsink Testing Platform. For full HSF reviews, please check in the Cooling section. Links to the listed coolers are provided in the tables below.

We consider the heatsink and mounting system together as a unit. The heatsink's intrinsic cooling power is determined mainly by its radiating surface area, the heat transfer coefficient of its materials, the spacing and number of fins, the geometry, the smoothness and flatness of the CPU contact surface, and overall mass. The mounting system is mentioned because it is critical in maintaining the all-important contact between CPU and heatsink. The amount of pressure brought to bear on the interface also affects cooling. Some clips are poor; others are integrated wonderfully into the heatsink. Increasingly, the overall mass is becoming important as HS get bigger, taller and heavier: The cantilever effect in any tower case can put tremendous stress on the vertically mounted motherboard.

The test results summarized here are with the stock fan(s) whenever available, or our reference fan(s) at 550, 700, 900 & 1100 RPM. In many cases, replacing the stock fan can lead to improved performance, and this is usually detailed in the reviews. However, for this list, since most coolers come with their own fan(s), it was decided that performance with the stock fan was most apropos.

A HSF represents only one component of effective CPU cooling. The influences on CPU temperature include the overall heat generated in the case, case ventilation airflow, ambient temperature, software applications, and user habits. The temperature rise above ambient, which tells us how many degrees above the ambient temperature the CPU is allowed to rise by the particular heatsink, is a specification that tries to elminate these other influences.

Heatsinks come and go faster than just about any peripheral PC product. We don't expect to keep up with everything available. Remember, we're only interested in well-designed heatsinks that cool well with low noise.

Summary tables on all comparable coolers are included in every SPCR heatsink review these days, and these tables are the best one-glance, up-to-date survey of silent CPU cooling. Such comparison tables have made this recommended heatsink article less compelling than before, which is one reason it hadn't been updated for years

Our previous recommendations ranked the coolers into large, medium and small (low profile) coolers and assigned a "Q" (quality rank) based each unit's performance, noise, etc. Reducing an entire heatsink's "value" into a single number was no easy task. Given the even greater complexity of the task, we're resorting to a version of the comparison tables used in our reviews.

A List of Big Coolers

We have all types of coolers in the mix: Top-down and side-blow towers, single fan and dual fan. The only criteria here are that they are large, best in cases meant to house large CPU coolers, and able to get down to at least 20 dBA @ 1m with no more than 45°C rise over ambient. All-in-one water coolers are in the mix because there's no better place to put them (for now). They're meant to compete with big air coolers, and their radiators usually need extra room around 120mm or larger fan exhaust openings. Some of these WCs will obviously fit fine in small cases where big air coolers cannot.

BIG COOLERS: °C Rise @ SPL
<<Tested on 130W TDP i7-965>>
SPL (dBA@1m)
Type
20~19
18~17
16~15
14~13
12~11
W
34
37
Prolimatech Genesis
(2 ref fans)*
S
36
37
39
S
35
38
40
S
38
41
42
S
38
40
43
Noctua NH-D14 (2 ref fans)†
S
38
41
44
Noctua NH-C14 (2 fans)
T
37
39
41
S
37
40
Thermalright HR-22 (2 ref fans)*
S
38
39
Prolima Megahalems (ref fan)*
S
38
41
44
Noctua NH-D15 (2 fans)
S
38
39
40
S
39
40
43
S
39
40
41
43
S
40
42
45
S
38
40
43
S
38
39
41
W
38
NZXT Havik 140 (2 fans)
S
39
41
41
43
S
40
41
42
S
40
43
S
41
43
46
S
40
42
47
S
41
44
54
S
43
48
S
44
46
S
44
50
S
44
47
S
40
S
42
44
SilverStone HE02 (2 ref fans)*
S
44
46
W
44
44
47
S
45
47
S
45
48
50
52
W
44
46
47
S
45
47
49
SPL (dBA@1m)
Type
20~19
18~17
16~15
14~13
12~11
S = side blow towers, T = top down airflow, W = watercooled
* = reference fan(s) because sample came with only one fan or no fan.
† The Noctua NH-D14 results are with 2 Nexus 120 fans; tests with the stock 14 & 12 cm fans were done only on the previous socket 1366 platform & cannot be fully trusted.

The term "cooler" (which really irks my language logic) is actually apropos here, as both "air only" heatsinks and all-in-one water cooling devices are in the mix.

The table may be a bit difficult to understand at first. Here's an FAQ:

1. What's the data shown? The core information shown is the rise in CPU temperature over ambient air temperature (in °C) at various measured noise levels (sound pressure level or SPL in decibels, A-weighted, at 1m distance) for all of the listed cooling devices. The test platform is a hot overclocked (and over volted) 130W TDP 1366 Intel core i7 running Prime95. The the coolers here were tested with its own single, stock fan.

2. What are the tested noise levels? The SPL levels are divided into five 2-dB range increments, from 20~19 dBA, to 12~11 dBA. The different noise levels are achieved by changing the speed of the fan(s); we have several standard speeds (defined in RPM) that are used for almost every review, though sometimes, hardware/fan limitations prevent this.

Most of you know that 20 dBA@1m is already pretty quiet. A 1 dBA difference is extremely difficult to hear, especially at such low levels, so the difference between 17 and 18 dBA or between 19 and 20 dBA is negligible; hence the 2-dB increments, to keep the table from getting overly complex. As the SPL drops, small differences get increasingly harder to hear, because it approaches the threshold of your hearing sensitivity, and the ambient noise level of your room.

In most cases, anything below ~15 dBA is pretty much the same, subjectively, because it will be masked by ambient noise, — other components in your PC or other sources, such as the fridge in the kitchen, the distant drone of traffic a block away, etc. 11 dBA is the ambient noise level of the anechoic chamber where the coolers are measured; we cannot measure anything quieter (and it is extremely difficult to hear).

3. Why is there no data above 20 dBA? There is in the original reviews, but we set 20 dBA as the upper limit for recommended coolers. You want to know how a cooler performs at low noise levels, not when it's blasting at 49 dBA and 4000 RPM!

4. Why the 45°C rise cutoff? Given 25°C ambient, that means 70°C at the CPU. Many CPUs don't throttle till ~80°C, but the ambient inside a case can easily be 10°C higher than the room. If anything, we might be overly generous; it could have been set at 40°C... but then many reasonably good coolers would have been excluded, and not many use a 130W TDP CPU any more.

5. In what order are the coolers listed? The lower the °C rise at each SPL, the better the cooler. The coolers that test best are listed first. (Best on top.) Keep in mind that the accuracy of temperature measurements is not better than +, -1°C, and a 1 dB difference is virtually inaudible.

6. Is the higher position cooler always the better performer? Usually, but not always. Not all coolers behave the same way as fan speed is reduced. For example, in the table above, the Zalman CNPS10X Quiet has about the same 40~41 °C rise as the Cooler Master Hyper 212 Plus at ~20 dBA. At a super-quiet 11~12 dBA, however, the Zalman achieves 47°C rise, while the Coolermaster runs 7°C hotter with a 54°C rise. In other words, sometimes a cooler that doesn't cool as well as another model at higher fan speed cools better at a lower fan speed.

Hence, you need to apply your own judgement: If you think you'll be fine with 20 dBA, the obvious cooler is the cheapest one which hits 20 dBA while keeping temperature rise below ~45°C. (Assuming the room temperature is ~25°C, this keeps our CPU below 70°C, above which it throttles.) But a further consideration is the TDP of your CPU; your target temperature rise should be lower or higher depending on whether your CPU runs hotter or cooler than ours, and by how much.

7. How about if I just want the best? Check the clearance (ie, will it fit with regard to your motherboard, video card, RAM height, etc) and get the cooler positioned highest on the list.

8. What about the missing data points? All coolers are tested at full fan speed as well as at 1100, 900, 700 and 550 RPM (whenever possible). This usually means that the cooling performance at maximum speed is not shown on this table, because most fans on most heatsinks run louder than 20 dBA@1m at full speed. The missing data points exist because the other tested fan RPMs did not fall into those SPL slots. Once you hone in on a particular cooler or two, it's best to study the original review for full details.

9. Can these coolers' performance be improved? Many of these coolers can be fitted with a second fan in a push-pull setup, which usually does provide a bit of performance benefit, sometimes with no or very little increase in noise. Decreasing ambient temperature naturally helps.

10. Are all the coolers still available? Market conditions change rapidly; we just can't track comprehensively around the world, so you're best to check on your own. Some models will have been modified with mounting hardware for newer CPUs; others with new fans. Let us know if you find a cooler is completely discontinued & unavailable.

11. Why not call it a Tower cooler instead of the odd Side-blow? There's only one term for Top-blow coolers, which starts with T.

A List of Smaller Coolers

As with the big cooler list, we're not differentiating between top-down and blow-across tower types, nor between single fan and dual fan. The only criteria here are that they are not huge, fit in smaller cases, and they can get down to at least 20 dBA @ 1m. These are actually ALL the heatsinks tested on our cooler small heatsink test platform. It's a small enough number so the list is not onerous. No maximum temperature rise has been imposed, as some of these are truly tiny heatsinks, for use in systems where nothing else will fit.

SMALLER COOLERS: °C rise @ SPL
<<Tested on overclocked, overcolted Core i5-2400>>
SPL (dBA@1m)
Type
20
19
18
17
16
15
14
13
12
Noctua NH-L12
(2 stock fans)
T
33
-
-
34
-
35
-
36
S
-
32
-
-
-
34
36
38
S
-
33
-
34
-
-
35
38
S
-
-
-
-
-
35
-
42
-
S
-
-
-
-
38
-
-
-
43
T
-
39
-
-
-
43
-
48
T
-
-
-
-
42
-
-
47
-
T
-
-
45
-
-
46
-
52
T
43
-
-
-
-
46
-
61
T
-
-
-
53
56
-
64
T
-
-
56
-
-
61
-
-
-
T
-
-
57
-
-
62
-
65
T
-
-
67
-
-
-
69
-
-
T
-
66
-
-
-
-
77
-
F
S = side blow towers; T = top down airflow ; W = water cooled
F = fail

Fanless Coolers

This is a special category of very few coolers, tested without any fans mounted on them, and with no forced airflow around the open bench test platform (on which the cooler is mounted). In all cases but one, the CPU became too hot and throttled (at around 78°C), which indicates failure. So the metric here is time: The number of minutes before overheating occurred.

FANLESS COOLERS
Minutes to Overheat
NoFan CR-95C Copper
62°C
Thermalright HR-22
13m
SilverStone Heligon HE02
9m
Thermalright HR-02 Macho
8m

Only the NoFan CR-9C Copper can be considered a truly successful fanless heatsink. It managed to keep our Core i7-965 CPU at 62°C (in 23°C ambient) indefinitely under Prime95 stress load without any forced airflow anywhere near the test platform. Inside the walls of a case, the air might heat up enough over time to make our CPU overheat even with this cooler, but if you can keep the internal temperature of your system under ~35°C, you have a good chance of running complete fanless (without any other big heat sources like a gaming graphics card).

Tested on the cooler (95W TDP) smaller HS test platform, the NoFan stabilized a thermal rise of 54°C, which is not bad considering it's making no noise at all. See the review for full details.

The inability to keep our 130W TDP i7-965 CPU from overheating should not be considered a complete failure. Each of these other heatsinks is likely to provide good enough CPU cooling when combined with a good low noise case airflow setup. However, keep in mind that VRMs and controller chips on motherboards (PCH) can overheat without adequate airflow and/or if the CPU is allowed to run very hot indefinitely. And yes, these coolers appear on the Big Coolers list because they were also tested with fans.

* * *

We haven't tested any VGA coolers in quite some time; the previous recommendations will be considered and an update will be made in due course.

The remaining pages are previous recommendations (last updated mid-2010, which will still be useful for some purposes.)

Discuss this article in the SPCR Forums.



1 2 3 4 Next

Reference|Recommended - Article Index
Help support this site, buy the Cooling Fans/Heat Sinks from one of our affiliate retailers!