Testing Methods

We've included results from several CPU coolers in each test. Not all are present in all socket tests because some aren't compatible.

We've dropped in the Swiftech kit as a representation of what true watercooling can do, but it's worth noting that at £200 it's significantly more expensive and requires more effort to install than the Corsair H50.

We've also included the Titan Fenrir, ThermalRight Ultra Extreme 1366 RT and Noctua NH-U12P HSFs as these are the best CPU coolers we've tested recently so are what the H50 needs to beat in order be worth buying. In addition to these coolers, we've also included the reference cooler for each socket, which is a good indicator of what kind of improvement you'd see if you're still using the boxed cooler. For instance, you certainly wouldn't want to buy a third-party cooler that performans worse than the reference cooler.

Where CPU coolers have variable speed fans, we've tested at both the lowest and highest fan speed settings and recorded idle and load delta T results for each setting to give you some idea as to how much performance is affected.

The temperature results you see in the graphs are not the absolute CPU temperature it it the delta T, in other words the difference in °C between the ambient temperature and the recorded temperature of the hardware, in this case the CPU. We used Core Temp 0.99.5 to take temperature readings of the CPU from the Digital Thermal Sensor (DTS) embedded in the core of every modern CPU, at idle and under load while using the smallfft torture test in Prime95.

A multimeter thermal attachment probe was used to take the ambient air temperature three inches away from the case's primary intake location. We use a delta T measurement because it's a more accurate and comparable method of recording temperature because the ambient room temperature fluctuates from day to day.

Socket LGA1366 System Results

For LGA1366 we used a Core i7-965 Extreme Edition overclocked to 3.6GHz with a vcore of 1.3V - the absolute maximum overclock and overvolt the reference Intel cooler can cope with. This is the same test rig we use for all Core i7 coolers, so you can compare the results with the other products we've tested recently. The other components in the test rig include an MSI Eclipse SLI motherboard and a Zotac GeForce GTX 280, all housed inside a Cooler Master Stacker 830 chassis.


The LGA1366 test results for the Corsair H50 are extremely positive - only two degrees off the full watercooling kit (albeit with silent 5V fans). It also pretty much annihilates the Domino ALC when we consider noise and performance together - only at full speed can the Domino ALC keep up, but it's a few degrees off the much, much quieter Corsair H50 that uses a single slow spinning 1,500rpm fan, compared to the Domino's beastly 3,000rpm+ fan.

Corsair claims that mounting the fan so that it pulls cool air from the outside of the case through the radiator yields a lower CPU temperature than sucking air through the case. This is because, the ambient air temperature inside the case will be significantly higher due to it passing over the disk drives, motherboard and expansion cards. However, this goes against the natural airflow dynamics of most cases, which are designed to pull air in at the front and expel in out the back. As a result we tested the Corsair H50 in both scenarios, with one and two fans fitted.

On the graph above, when one fan was used this was the Corsair only, however with two fans we used a Noctua NF-12P in addition to the Corsair fan. An additional fan makes no difference to the performance when setup to draw in air from outside the case. However, when exhausting air the second fan only drops the delta T a few degrees, so there's really no point given the extra noise.