Cooler Testing Revisited

Until recently we haven't been very aggressive in testing CPU cooling methods. I'd been busy with notebooks, desktops, cases, and peripherals, but good and consistent writers are hard to come by in this industry, and eventually I couldn't say no. Tentatively, I gave it the old college try, starting with two radiator fan roundups and then doing multiple liquid cooling roundups. Interestingly, it was when more conventional air cooling popped up on my radar that things got complicated.

The cooling testbed was and is solid. We use a 200mm BitFenix Spectre Pro that's throttled to 5V as an intake, and that's in the front of a BitFenix Shinobi XL enclosure, a case which is almost perfect for our needs. The low speed on the Spectre Pro allows for intake of cool air without negatively affecting noise testing, and for closed loop liquid cooling, this is fine. Where things get more dicey is when you introduce an air cooler into the testbed, because as a couple of you rightfully pointed out, without an exhaust fan to direct air, air coolers suffer tremendously performance wise. To be certain I took our original testbed, added a 140mm Noctua fan with a low noise adaptor, and mounted it to the rear exhaust of the Shinobi XL. Even with a minimally powered exhaust fan, the differences in performance were pronounced. Since this is the situation air coolers will typically find themselves in, I'm now using that exhaust fan for testing air coolers. Closed loop coolers continue to do without.

There was one other wrinkle with the existing testbed: our motherboard just wasn't especially stable, and if it crashed, it was difficult to get it to post again. Recently, this became easy to remedy: the micro-ATX board I was originally using for case testing was retired in favor of a full ATX board. Switching over to the Gigabyte GA-Z68MX-UD2H-B3 meant having a more reliable and more fearsome overclock on the Intel Core i7-2700K. Now the chip runs hotter, the socket lines up correctly with the hole in the case's motherboard tray, and it's more stable overall.

Of course all of these changes mean one important thing: a lot of coolers needed to be retested under these more stressful (but also sometimes more ideal) conditions. To be sure, the previous roundups are still useful for comparing products in their individual categories, but now liquid coolers aren't the juggernauts they used to be. Out of the coolers we've already tested, I selected ten to be revisited: five air coolers and five closed loop liquid coolers.

Coinciding with these revisions is our evaluation of Cooler Master's Seidon 240M closed loop cooler along with two new tower coolers from Noctua, the NH-U12S and NH-U14S, both of which were designed to create clearance for memory modules with tall heatsinks. The Seidon 240M is noteworthy because it's not directly sourced from Asetek or CoolIT Systems as many of these products are, and uses a proprietary waterblock design that theoretically improves overall cooling performance.

The Cooler Master Seidon 240M
POST A COMMENT

59 Comments

View All Comments

  • spidey81 - Friday, April 26, 2013 - link

    This article was to test the performance of the coolers at a certain clock speed. However, I think you would find that when pushing the overclock higher you would see some coolers wouldn't be able to keep the temps in check while others could. There's a difference between efficiently transferring the heat and the amount of heat capable of being transferred. That's why the D14 and TRUE are popular among overclockers who push higher than what we see in this test. To that end, the 240mm or 280mm closed loop coolers will also be able to handle higher overclocks.

    In this (http://www.overclock3d.net/reviews/cases_cooling/s... review of the H220, you can see that on the socket 2011 cpu the air coolers weren't able to maintain the extreme overclocks like the closed loop coolers could.
    Reply
  • spidey81 - Friday, April 26, 2013 - link

    sorry, link broke. http://tinyurl.com/cu24bwk That should work better :) Reply
  • epoon2 - Friday, April 26, 2013 - link

    You meant that as the load/watt/heat produced by CPU increases, the order in this graph http://images.anandtech.com/graphs/graph6916/54390... changes?

    but in the above test, the fans are at constant, max speed already ... I know what you mean but I just can't figure out the science/reason behind why your scenario would happen.
    Reply
  • BrightCandle - Friday, April 26, 2013 - link

    Two key factors are at play:
    1) the amount of watts of heat the solution can dissipate at a given delta temperature to the air.
    2) the thermal interface performance to the CPU itself and hence what temperature the CPU is running at.

    Water is cooling the water block but then water itself is above ambient temperature. This often means the ambient temp CPU effectively sees is higher compared to what it was with air. But water warms and cools much faster than air by volume, and it can hold a lot more energy. Combining water with a substantially larger area to exchange its heat to air and you get a solution that cools better given a large number of watts of heat but does less well on low wattage where air cooling fits and is sufficient.

    A 2600k pulls about 170W over clocked. A 3930k can be over 300W. 170W is about the point to consider water but you need an overspecced loop to show any advantage at all. Passed 200 watts however and water takes a clear lead and air stops being able to do the job.
    Reply
  • epoon2 - Saturday, April 27, 2013 - link

    Assuming the water pipes dissipates negligible heat, the role of the water system is equivalent to that of the thermal compounds: to transfer energy and heat to the sink and radiator.

    The difference in rate of heat exchange between the metal pieces and the external environment becomes the key.

    In an open air test, the rankings should remain the same as amount of heat generated by the CPU increases. In the real world , the performance of air coolers depends on the effectiveness of the case's heat exchange system..
    Reply
  • A5 - Friday, April 26, 2013 - link

    For future reviews, I wouldn't mind seeing the Xigmatek SD1283 tested. A lot of people recommend it for people who want something better than the CM Evo but don't want to spend Noctua money. It typically runs ~$50 so it fits that niche price-wise as well. Reply
  • A5 - Friday, April 26, 2013 - link

    I guess I meant the silly-named "Dark Knight 2" when I wrote this. Didn't know they had 2 different SD1283s. Guess I wouldn't mind seeing the cheaper, non-coated "Gaia" as well. Reply
  • Dr_b_ - Friday, April 26, 2013 - link

    Does the U12 have socket 2011 mounting capability? Reply
  • epoon2 - Friday, April 26, 2013 - link

    http://www.noctua.at/main.php?show=productview&... Reply
  • Dr_b_ - Friday, April 26, 2013 - link

    thanks 4 link. NH-D14 is really too big, blocks a socket in my x79-UP4 gigabyte. U12 looks like it might work here. Reply

Log in

Don't have an account? Sign up now