Sign in with
Sign up | Sign in

Test Settings And Benchmarks

Nine Big Air Coolers For Intel's Haswell CPUs, Reviewed
By
Test System Configuration
CPUIntel Core i7-4770K (Haswell): 3.5 GHz Base Clock Rate, 8 MB Shared L3 Cache, LGA 1150, Overclocked to 4.50 GHz at 1.25 V (1.24 V Set)
MotherboardAsus Z87-Pro Rev 1.02, BIOS 1007 (05/17/2013)
RAMG.Skill F3-17600CL9D-8GBXLD (8 GB) at DDR3-1600 C9 Defaults
GraphicsIntel Integrated HD 4600 Graphics
Hard DriveSamsung 840 Series MZ-7PD256, 256 GB SSD 
SoundIntegrated HD Audio
NetworkIntegrated Gigabit Networking
PowerCorsair AX860i: ATX12V v2.3, EPS12V, 80 PLUS Platinum
Software
OSMicrosoft Windows 8 Professional RTM x64
GraphicsIntel 9.18.10.3071
ChipsetIntel INF 9.4.0.1017
Benchmark Configuration
LinX 0.6.4Default benchmark at 500 repetitions
RealTemp 3.70Maximum Temperature, All Cores Averaged
Galaxy CM-140 SPL MeterTested at 1/4 m, corrected to 1 m (-12 db), dB(A) weighting

Asus’ top-overclocking Z87-Pro lays the groundwork for thermal testing, but not without compromise: knowing that CPU throttling would prevent accurate thermal measurement, I decided to prevent that occurrence by reducing core voltage to 1.25V.

LinX 0.6.4 is the new application that generates the most taxing thermals using Intel's diminutive Haswell die. I also warned cooler manufacturers that we'd have to drop them from contention if the CPU reached 100° Celsius, since throttling would prevent accurate readings above that point. For our part, I kept the room temperature below 23° Celsius. I further prevented the case from affecting thermal readings by leaving the motherboard in open air, hanging the board vertically to replicate its position in a chassis.

Lacking any mechanical support for a hanging graphics card, I used integrated graphics during these tests.

Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 131 comments.
This thread is closed for comments
Top Comments
  • 16 Hide
    Madn3ss795 , July 14, 2013 9:27 PM
    No NH D14?
  • 10 Hide
    s3anister , July 14, 2013 11:30 PM
    I can't even fathom why one would spend this much time to write an article like this and still leave the NH-D14 out.
Other Comments
  • 16 Hide
    Madn3ss795 , July 14, 2013 9:27 PM
    No NH D14?
  • 0 Hide
    016ive , July 14, 2013 10:03 PM
    Thermalright Archon SB-E X2, That's one big as* cooler!
  • 6 Hide
    jk_ventolero , July 14, 2013 10:16 PM
    The CM H212 Evo has been a crowd favorite for quite some time now. :)  I wonder how that might have performed. :D 
  • 1 Hide
    CaptainTom , July 14, 2013 11:27 PM
    I hate to say it but all this article demonstrated to me is that Haswell is in no way better than Ivy Bridge. Even Ivy Bridge used 25% less energy with a 10% boost when compared to Sandy Bridge, and I can still get IB to 4.8 GHz on air.

    Haswell= 10% more power for 10% more energy and 10% less headroom. Oh and it costs more...

    Man I hope AMD can whip them back into shape with Steam Roller...
  • 10 Hide
    s3anister , July 14, 2013 11:30 PM
    I can't even fathom why one would spend this much time to write an article like this and still leave the NH-D14 out.
  • 3 Hide
    iknowhowtofixit , July 15, 2013 12:03 AM
    Would love to see how the H60 (new version) compares as a point of reference.
  • -5 Hide
    spazoid , July 15, 2013 1:10 AM
    It's funny how some of you guys still think Intel made Haswell with desktops in mind. This is a CPU architecture made for mobile. It's quite sufficient for desktops in every regard. Yes, it might not outperform IB by much, and it might run a bit hotter, but in a laptop, there is nothing that compares to Haswell.
  • 8 Hide
    Crashman , July 15, 2013 1:25 AM
    Quote:
    No NH D14?
    Noctua picked the NH-U14S instead of the D14. Please feel free to ask them why.

    Since Noctua chose not to feature the NH-D14 this time, you might want to look at its closest competitor. The article links a comparison of Phanteks TCP14E and NH-D14. Since Phanteks chose to keep its big cooler in the running but Noctua chose otherwise, this was the closest "frame of reference" available.

    Quote:
    I can't even fathom why one would spend this much time to write an article like this and still leave the NH-D14 out.
    I can't even fathom why you would repeat the first post without regard to any response. And speaking of a response, please read above.
  • 1 Hide
    Shneiky , July 15, 2013 1:42 AM
    Really disappointed that there was no be quiet! and Cooler Master representatives. I would enjoy seeing how a Dark Rock Pro 2 or a V8 GTS/V6 GT.
  • -5 Hide
    flong777 , July 15, 2013 1:49 AM
    The comments are correct concerning the omission of the NH-D14 from this testing. It really doesn't matter a damn what "Noctua picked," that is not the issue. The FACT is the D14 is legendary for its cooling and to leave it out of any major test is just plain stupid. Don't forget that it won your last competition for best cooler.

    To say compare it's performance with the Plantek's performance is lame. Use some common sense when you do these reviews.
  • 3 Hide
    Crashman , July 15, 2013 1:52 AM
    Quote:
    Really disappointed that there was no be quiet! and Cooler Master representatives. I would enjoy seeing how a Dark Rock Pro 2 or a V8 GTS/V6 GT.
    Since I've never heard from anyone at Be Quiet, I've sent a message. Thanks for the reminder!

    Cooler Master chose not to participate this time.

    Quote:
    omission of the NH-D14..doesn't matter a damn what "Noctua picked,"...FACT is ...test is just plain stupid..Use some common sense
    Troll much? The FACT is that Noctua said use this. That's the end of the Noctua story, anything you'd like to add after that is counter-factual.

    You'd be placing the blame wrongly, if there were any blame to place. Now if you want some clue as to how it should perform, read the noc-vs-phan review. It's linked. Good luck.

  • 1 Hide
    Shneiky , July 15, 2013 1:59 AM
    Quote:

    Crashman said:
    Since I've never heard from anyone at Be Quiet, I've sent a message. Thanks for the reminder!

    Cooler Master chose not to participate this time.


    Hope it works out. Really interested to see those models. I am quite certain they will make a bit of a difference in your performance charts. Good luck.
  • 0 Hide
    Dinko Mat , July 15, 2013 2:34 AM
    What about to improve the thermal performance of Haswell by removing Integrated Heat Spreader Removal and Replacing IHS Thermal Paste. You can get 10 or more degree C less temperature.
  • 2 Hide
    Dinko Mat , July 15, 2013 2:36 AM
    http://www.youtube.com/watch?v=XXs0I5kuoX4
  • -2 Hide
    Dinko Mat , July 15, 2013 2:38 AM
    http://www.youtube.com/watch?v=Bg9ybeB4DDw
  • -4 Hide
    Dinko Mat , July 15, 2013 2:40 AM
    http://www.youtube.com/watch?v=Bg9ybeB4DDw
  • 1 Hide
    ingtar33 , July 15, 2013 2:57 AM
    the true disapointment with haswell is the lack of significant improvement with delidding. While, like IB you can see a solid 10C drop in temps, haswell has a far more agressive heat profile then IB. With minimal bumps in voltage resulting in massive jumps in heat, meaning even with a delidding you could see very little improvement in your overclock, as the chip will still rapidly produce massive ammounts of heat, very fast.

    While i understand Intel has their own formula for determaning TDP for their chips, i've said it before, by moving the VRM onto the chip, they invalidated their calculation. The Haswell performs and benches like a much hotter chip then it's advertised TDP. The first warning to benchers this was the case, was when benchers were getting thermal throttling from turbo mode on the stock intel heatsink, and then those same benchers hitting the thermal ceiling with a very small voltage bump and moderate overclock (4.2) with a solid air cooler the hyper evo. People can overclock 125W amd chips pretty aggressively with a lot of vcore on a hyper evo... meanwhile they were thermal throttling with a very modest bump in vcore on a tiny overclock.

    This raises a question about the formula intel used to calculate the TDP of their haswell chip, as cpu coolers rated for far higher tdps were struggling with haswell. This article is a bit of an eye opener, as most of those cpu coolers are not jokes. The noctua isn't all that different from the old cpu cooler i used to use to keep my PhII x4 965be cooled. It managed to keep a PhII at a 3.7ghz overclock in a 35C room under 55C while running prime. So that's no joke of a cpu cooler. That it struggles so badly with haswell says something significant to me.

    Simply put, you need a water system with a serious rad/fan combo to make a serious stab at a haswell overclock. Air isn't enough anymore.

    Quote:
    What about to improve the thermal performance of Haswell by removing Integrated Heat Spreader Removal and Replacing IHS Thermal Paste. You can get 10 or more degree C less temperature.


    yes... it will give you 10+C better temps. the problem is haswell basically heats up exponentially with more voltage. So this rarely gives you a lot more overclock headroom then you had pre-delidding. Ignoring those NO2 guys or the ones using some peltier or other exotic cooling solution, Haswell doesn't really benefit as much as IB does with delidding. another 0.1ghz or 0.2... at best. and that might be a bit optimistic. The real trick with haswell is getting the right batch... apparently some of the chips from certain asian foundries are atrocious overclockers with almost crippling heat problems...
  • 0 Hide
    chesteracorgi , July 15, 2013 5:00 AM
    I'd like to see how CLCs compare. Nice article.
  • 2 Hide
    vertexx , July 15, 2013 5:15 AM
    Nice article. Would also like to see a follow-up w/CLCs.
Display more comments