Test Setup and Temperature Benchmarks
The open test bench in action
An open test bench was used for all benchmarks, which really helped with switching between the coolers in this review. Both the cooling and noise levels are likely to be different inside of a case, but this depends on airflow.
Both an Intel Haswell (i7 4770K) system and an Ivy Bridge-E system (i7 4930K) were used for the benchmarks, running each CPU at both stock frequencies and 4.50 GHz overclocks. The benchmarks were performed with a constant 18 C room temperature and Prime95 was used to generate load temps, with the maximum heat/stress torture test running for a minimum of 10 minutes before each temp reading was made with RealTemp.
One quick note on idle temps between the systems: You might notice slightly lower idle temps with the Ivy Bridge-E system with a few of the coolers, which seems erroneous given the high thermal load of the hex-core 4930K. I isolated the reason for this during testing, and it was simply due to the default settings on the Haswell motherboard (ASUS Maximus VI GENE) keeping the 4770K from down-clocking at idle, which was consistent through all 4770K testing.
For all tests Noctua NT-H1 thermal compound was used, and the processors and coolers were properly cleaned before each installation to ensure consistency.
|Intel Haswell System||Intel Ivy Bridge-E System|
|Processor||Intel Core i7 4770K||Intel Core i7 4930K|
|Motherboard||ASUS Maximus VI GENE||ASUS Rampage IV GENE|
|Memory||G.Skill Sniper 8GB 2133MHz||G.Skill Sniper 16GB 2133MHz|
|Graphics Card||NVIDIA GeForce GTX 770 (OEM)||NVIDIA GeForce GTX 770 (OEM)|
|Storage||OCZ Vertex 460 120GB SSD||Samsung 840 Pro 128GB SSD|
|Case||Dimastech Bench/Test Table Easy v2.5||Dimastech Bench/Test Table Easy v2.5|
|Power Supply||Corsair CX 750W Modular PSU||Corsair CX 750W Modular PSU|
|OS||Windows 7 Professional SP1 64-bit||Windows 7 Professional SP1 64-bit|
|Corsair H105 Extreme Performance Liquid CPU Cooler|
|Corsair H100 Extreme Performance Liquid CPU Cooler|
|Cooler Master Seidon 240M|
|Corsair H75 High Performance Liquid CPU Cooler|
|Corsair H60 High Performance Liquid CPU Cooler|
|Noctua NH-D14 SE2011 (with NM-i115x mounting kit for LGA1150 installation)|
|Noctua NT-H1 thermal compound used for all tests|
First we'll look at the Haswell system results.
A strong first showing here for the H105, coming in just behind the H100 set to 'high' fans. This is more impressive considering the H105 is using a much quieter auto fan speed with its PWM fans. The Noctua air cooler is actually the best here, and it excels at mitigating heat at these stock settings.
The overclock of 4.50GHz on the 4770K was actually running at a full 4.5 GHz on each core at load (HT enabled for a full 8 threads), and required 1.312 volts for stability. Here the H105 really impresses, and the strength of the thicker radiator can be seen. You'll notice that Cooler Master's 240mm Seidon 240M was able to cool just as efficiently at load here, and actually ended up on top due to better average temps at idle.
Next are the Ivy Bridge-E results, beginning with the 4930K at stock speeds.
The venerable H100 takes first place here at both ‘high’ and ‘low’ fan settings. The H105 is still excellent, with a very low idle temp of 24 C and just 52 C maximum load temp - and without any increase in fan noise. Once again the Noctua NH-D14 performs very well, as it seems to be a great choice for very quiet cooling at stock speeds.
The overclocked 4930K was the most severe test of the day, and as before with the 4770K, the turbo frequency was linked in the motherboard setup to force 4.50 GHz on each of the 6 cores (and 12 total threads) of this Ivy Bridge-E processor. It required a setting of 1.408V on the core to retain stability under full load, and generated a lot of heat!
The Corsair H105, which was doing very well with some very good noise levels to this point, hit it out of the park with this most challenging test. As you can see it was a full 2 C cooler at load than an H100 at max fan speeds, and this was consistent through long Prime95 runs. The H75 was also very impressive here, coming in at just 78 C in this worst-case scenario for CPU thermals. Surprisingly the Seidon 240M, which had done very well in testing on the Haswell system, was unable to keep the CPU under 80 C even with some jet-like volume levels!
A quick note on the H60’s result: it did pass the minimum 10-min Prime95 run, but at 90 C I was not willing to push it (or the 4930K) any harder. Then again, when would a budget cooler be paired with a $560 processor? Not likely!
Next, we'll see how the coolers stacked up after the noise tests.