We measure the CPU temperature of our test rigs when the CPU is idle (Windows desktop) and under load (by running the smallfft test in Prime95). We use the freeware application CoreTemp to measure the temperature, which reads the data from the DTS (digital thermal sensor) embedded in the core of every modern CPU. The DTS gives a far more accurate reading than any third-party temperature sensor, because its readings are not affected by sloppy BIOS programming or a poorly calibrated external health monitoring chip on the motherboard.
Because we don't have a temperature control chamber, it's more accurate to measure and quote the delta T (difference) between the ambient and CPU temperature. This means HSFs (heatsinks and fans) that were tested on different days, when the ambient temperature may have varied, can be fairly compared to one another. You'll find the temperature results for the reference HSFs in the graphs. The key question to consider when choosing a new HSF is whether it's cooler or quieter, or both, than the reference Intel and AMD HSFs.
We've rebuilt our CPU thermal test systems to cater for our current recommended CPU sockets; AMD's Socket AM3 and Intel's LGA1155. While LGA1366 motherboards are still very popular, they've been around for some time and our recommended cooler, the Thermaltake Frio, is still the best option. As Intel's Core i5-2500K and Core i7-2600K CPUs are just as good as all of Intel's LGA1366 CPUs (bar the Core i7-980X Extreme Edition CPUs in a tiny number of situations), LGA1155 is now our main focus, especially as these CPUs run so much cooler.
Socket AM3 test system:
3.2GHz Phenom II X6 1090T Black Edition CPU overclocked to 3.6GHz with the vcore boosted to 1.6V
Gigabyte GA-MA785GMT-UD2H motherboard.
LGA1155 test system:
3.4GHz Core i7-2600K CPU overclocked to 4.6GHz with a vcore of 1.3V
MSI P67A-GD65 motherboard.
Each test rig is built inside a SilverStone PS03 midi-tower case and includes an Nvidia GeForce 9600 GT graphics card. The temperatures shown are the delta T between ambient and CPU temperature.
Socket LGA1155 CPU delta T
Arranged by 100 per cent CPU load delta T value
Antec Kühler H2O 920 (Extreme Profile)
Thermaltake Frio (high speed)
Be Quiet! Dark Rock Pro
Thermaltake Jing (high speed)
Zalman CNPS9900-MAX Blue (high speed)
Thermaltake Frio (low speed)
Alpenfohn Matterhorn
Antec Kühler H2O 920 (Silent Profile)
Zalman CNPS9900-MAX Blue (low speed)
Be Quiet! Dark Rock Advanced
Thermaltake Jing (low speed)
Corsair H70 (full speed)
Gelid Tranquillo
Antec Kühler Box
Prolimatech Genesis
Thermolab Bada 2010
Xigmatek Dark Knight
Intel Reference Cooler
Nexus VCT-9000
15
46
15
49
15
50
15
51
12
51
13
51
15
52
18
52
14
53
15
53
15
53
16
53
16
54
16
54
17
54
20
59
15
61
18
69
20
75
0
10
20
30
40
50
60
70
80
Delta T (°C) (lower is better)
Idle
Load
NOTE: We've removed the AMD cooling figures, as we've since found an irregularity with the AMD thermal test system. Bit-tech prides itself on being accurate first and foremost so we don't want to publish possibly inaccurate information for a second longer than necessary. We will update the situation as soon as possible.