Scaling of Cooling Performance

The Thermalright Ultra 120 Extreme has been the top performer so far in CPU cooling at stock speeds. As overclocks are raised, the Ultra 120 with the Scythe SFLEX fan maintains its cooling advantage. The Scythe Ninja Plus B is average in cooling efficiency with the stock fan. However, a swap to the specified 14 dB-A/72 CFM SilenX fan improved cooling to the levels of the top tier heatpipe towers in our reviews.

At 2.93GHz the retail HSF is running at 41C, compared to 31C with the Scythe Ninja Plus B with the stock fan.. This is a delta of 10C. The SilenX drops that temperature to 27C, or a delta of 14C. The delta with the stock fan remains about the same as the overclock increases. At 3.73GHz the idle with the retail fan is 56C compared to the Ninja stock at 47C - a delta of 9C. Performance improvements are much greater with the SilenX fan on the Ninja Plus B. At 3.73GHZ the SilenX cools to 36C, while at the highest stable overclock of 3.90GHz the Ninja/SilenX is at 42C. This almost flat line at idle at the top of our testing is an indication the SilenX is effectively cooling at idle all the way to the top speed that could be reached in benchmarking.

Cooling efficiency of the Ninja Plus B, with stock and SilenX fans, was next tested under load conditions. Results are compared to the retail HSF and other recently tested CPU coolers. The Scythe Ninja Plus B with the SilenX fan generated results almost identical to the Scythe Infinity with Push-Pull fans and reached the upper tier of overclocking performance. Results with the stock fan were average.

By 3.90 GHz - the highest overclock the Tuniq, Thermalright Ultra 120, OCZ Vindicator, and push-pull Scythe Infinity could reach with stability - the Ninja Plus B is at 62C with the SilenX fan. This is still worse cooling performance than many of other top coolers at this speed. Push-pull SilenX fans lower this only a couple of degrees more - to 60C. The Tuniq and Ultra 120 are both at 51C, which is some 9C to 12C lower than the Ninja Plus B. However, the Scythe was completely stable at 3.90GHz with the SilenX fan even though it ran a bit warmer. The single stock fan could only manage 3.83GHz so the SilenX or dual Sythe low-noise fans in push-pull configuration are much better fan setups for overclocking the Ninja Plus B.

As stated many times, the overclocking abilities of the CPU will vary at the top, depending on the CPU. This particular CPU does higher FSB speeds than any X6800 we have tested, but the 3.9GHz top speed with the top tier heatpipe towers is pretty average among the X6800 processors we have tested with Tuniq cooling. A few of the other processors tested with the best air coolers reach just over 4 GHz, but the range has been 3.8 to 4.0GHz. Stock cooling generally tops out 200 to 400 MHz lower, depending on the CPU, on the processors tested in our lab.

Cooling at Stock Speed Overclocking and Noise


View All Comments

  • Pirks - Wednesday, April 11, 2007 - link

    I'm going to get myself a monster of a heatsink called Cooler Master GeminII. It has two 120mm fans on top that blow right down on the mobo. On the other hand I keep reading all these news here coming from Gary, Wesley, Jarred and the gang, that tower coolers like Infinity or Thermalright kick ass and yield only to waterblocks. Now, the question is - are you guys really missing this important part of cooling mosfets, power circuitry, some northbridge, memory etc? Or is it really unimportant?

    I mean, what's the point in having such an imbalanced cooling solution where the CPU is cool but evetything else is almost melting? I recall my experience with Athlon XP in a tight poorly ventilated case - I put Zalman on it instead of a little tower HSF and you you know what? The power circuitry temperature went down 16 degrees right there, and I stopped getting Windows freezing on me. Stability became just like on those pesky Macs (j/k) :)

    So what is it? A vertical tower? Or a horizontal cooler that blows on mosfets? Or a mix of both with tower and additional coolers around it to cool mosfets?

    At least a link to some systematic research won't hurt, I can only see opinionated forums where half of the population says towers suck because they don't cool mosfets and the other half says towers rock precisely because they don't cool mosfets :) Wesley? Gary? Any other AT overclocking guru? What's your opinion and arguments for it (or against it)?
  • LaGUNaMAN - Thursday, April 12, 2007 - link

    Speaking of the Cooler Master GeminII, requesting to have that reviewed as well. Great job on the article BTW Bozz Wesley. Can't wait for the 120mm fan roundup. (^^,) Reply
  • Wesley Fink - Friday, April 13, 2007 - link

    The CoolerMaster Gemini II is in the lab for review. I agree it looks really promising with a pair of low-noise high-output fans. We hope to get to it within the next couple of weeks. Reply
  • Kaleid - Saturday, April 14, 2007 - link

    Could you try placing one fan in the middle of that cooler? Does it really perform better with two fans or would one silent fan do the trick? Reply
  • Stele - Wednesday, April 11, 2007 - link


    I mean, what's the point in having such an imbalanced cooling solution where the CPU is cool but evetything else is almost melting?

    That's a very good question, one which I've often asked myself when looking for coolers. Granted, a straight-through airflow right out the back of the casing helps prevent a build-up of warm air inside it, but that would leave a weak level of airflow around the other hotspots on the motherboard... especially with the likes of a 680i SLI overclocked.

    It would be terrific if Anandtech could deploy a few thermocouple probes or even an IR thermometer and measure the temperatures of those hotspots - PWM heatsink (or PWM MOSFET if no heatsink present), northbridge, and if time allows, perhaps even the side of a RAM DIMM and graphics card nearest the CPU socket area. That would certainly help differentiate the coolers good at just cooling the CPU from those that can do better overall. :)
  • noobzter - Wednesday, April 11, 2007 - link

    1. How does the CPU temp as reported in NTune compare to TAT's?
    2. What was the ambient temp during the test?
  • noobzter - Thursday, April 12, 2007 - link

    Sweet, thanks! Reply
  • Wesley Fink - Thursday, April 12, 2007 - link

    TAT workload testing with the stock Scythe fan on the Ninja Plus B show test results virtually the same at TAT 80% CPU load as our 30 minute gaming tests. Since you are likely familiar with TAT results that info may be useful. TAT does stress both cores in testing, which our current game tests do not do. Reply
  • Wesley Fink - Thursday, April 12, 2007 - link

    1. TAT (Thermal Analysis Tool) is an Intel program, so it did not occur to us to use it for verification on an nVidia chipset. However, it is CPU centric, and we found it DOES work on the 680i chipset. Comparing Idle temps to nVidia Monitor results at 3.73GHz overclocked showed reported Monitor and TAT temps within 0-2C of each other. We will test more with TAT using internal load tests in the future.

    2. Ambient room temperature during tests is 70F (21C).
  • iluvdeal - Wednesday, April 11, 2007 - link

    I am looking for a quiet HSF which allows good OCing for a Core2Duo and the Ninja fits that bill, however the one thing that's constantly repeated about it is it's difficult to securely mount with a S775. That's what scared me away from it. Reply

Log in

Don't have an account? Sign up now