Athlon64 HSFs

A few years ago, the only practical way to cool an overclocked CPU was to use water cooling, simply because most high-performance HSFs were too damn loud. For example, the two award-winning air coolers from our first HSF Labs test, the Vantec VP4-C7040 and the Swiftech MCX478, both featured fans with speeds in excess of 4,000rpm.

But while water-cooling hardware has only improved incrementally, HSF technology has advanced quite significantly. The use of heatpipes means that it's now possible to get the same level of cooling from a HSF that you'd get from a water-cooling kit, at a similar, or sometimes lower, noise level. This doesn't mean water cooling isn't worth considering anymore, but it's now only worth it for more extreme overclocks or multi-waterblock systems comprising CPU, chipset and GPU waterblocks.

For this Labs test, we rounded up 21 of the latest CPU HSFs, all of which feature some form of heatpipe cooling technology. Out of all the HSFs in last year's Labs test, only ten had heatpipes, and in the test before that only one had heatpipes. However, although heatpipes help to move heat away from the CPU core more quickly than traditional heat-conducting fins, a fan is still needed to transfer the heat to the air surrounding the HSF. Fortunately, fan technology has improved too, and HSFs are quieter than they were two years ago.

So, to find out which is the best HSF for your Athlon 64 or Pentium 4, which Pentium 4 HSF is cooler and quieter than most water-cooling kits, and which heatsink can passively cool an overclocked Athlon 64, read on.

The single most important characteristic of a HSF is how

well it cools your CPU and prevents the automatic thermal protection from activating and shutting down your PC.

Our method is to measure the CPU temperature using the Asus Probe utility, which in turns gets its reading from the built-in thermal diode in our test CPUs.

Although the motherboards we use have well-calibrated probes, you can never be entirely sure that the temperature reading is 100 per cent accurate, so instead we take the reading and compare the difference to the measurement taken when another HSF is fitted to the CPU. Once we have the relative temperature difference, it's then possible to make a meaningful comparison between different HSFs, and even water-cooling kits and phase-change cooling systems.

What you really need, though, is a reference point, and for this test, we used the HSFs that AMD and Intel supply with a retail boxed processor. These make great reference points, as most users will find these HSFs to be more than adequate, unless you're a serious overclocker.

So if you're thinking about upgrading, you'll be able to see which models are best, or if you bought an OEM processor without an

HSF then you can see which HSF, at prices ranging from £19 to £41, to buy.

For this Labs test, we used the two standard test rigs that we use to test all cooling hardware in Custom PC. The Athlon 64 test rig comprises an Athlon 64 FX-55 with the multiplier raised from 13 to 14 for a CPU frequency of 2.8GHz, and the vcore boosted to 1.55V.

The rest of the system comprises an Asus A8V WiFi-g Rev 2 motherboard and GeForce 6800 GT installed in a Cooler Master Stacker chassis. The Pentium 4 test rig comprises a 3.6GHz Pentium 4e overclocked to 3.96GHz by increasing the FSB from 200MHz to 220MHz (880MHz effective), and the vcore to 1.45V. This system is also built inside a Cooler Master Stacker chassis, along with an Asus P5AD2-E Premium motherboard and GeForce 6800 GT.

The temperature results you see below were taken once the CPU had reached its maximum stable temperature running one instance of Folding@home (http://folding.stanford.edu). Because folding is so CPU-intensive, this should be the highest temperature that the CPU ever reaches. The actual temperatures are shown relative to the reference AMD and Intel HSFs.

SCORING

Underneath each HSF review, you'll see four separate score boxes, one for Cooling, Design, Value and one for Overall. The Cooling score shows how well each HSF cooled the processor compared with the reference designs, and accounts for 60 per cent of the overall score. The Design score was calculated by giving a certain number of points for each of the key design characteristics.

For example, we penalised excessively noisy HSFs, and designs that exceed the AMD and Intel maximum weight recommendations, and score on how well the HSF mounts to the motherboard. The Value score is then derived by a weighted calculation of adding together the Cooling and Design scores, and then dividing by the price. The Overall score is also weighted and isn't an equal average of the three initial scores. For this ratio, we used a weighting of Cooling (60 per cent), Design (20 per cent) and Value (20 per cent), as cooling is obviously the most important characteristic of a HSF.