Test Setups and Methodology

The setup has changed from my previous tests, for the past year I’ve been able to test each heatsink in the exact same system, set up on the exact same spot. I moved to a new home 2 weeks ago, and thus, my test setup has changed. The hardware is still the same, but the physical location of the case has been changed.

I re-tested a few heatsinks I did previously (Scythe Ninja / Tuniq Tower / …) and temperature results are still the same, so that’s good news :). However the bad news is that I’m not living in a quiet area anymore, I’m next to a big road with lots of traffic and there is a high way at about 1km also. The house has very bad isolation, so there is a lot of noise… everywhere, almost all of the time, even at night.

The end result is an increased ambient noise in my new test room, if I remain very quiet and ignore the occasional car which passes by and spikes up the noise reading, I’m getting 36-37dBA and that’s best case scenario, in the middle of the night, during the day it can easily go near 40dBA, without any appliance turned on in the house. The cars are clearly audible as they pass by and it’s quite a difference from what I’m used to before. However a quiet PC is still much appreciated.

It’s not because a fan which is tested at 15dBA (in a special room) generates 28dBA @ 1 meter, that it will not be audible in a room where the ambient noise is 40dBA. In fact, it will still be very easy to pick up the noise of that fan.

Practical example, the Scythe Ninja installed combined with 120mm Papst fan, I previously got a ~35dBA reading in a ~32dBA room at 50cm. When I start the machine in its new location, I can easily hear it above the ambient noise in the room.

The bottom line: this sucks. Noise readings from my previous tests are not comparable to those of the new test setup.

JMke's Test Setup
CPU	A64 3200+
Mainboard	DFI NF3 Lanparty
Memory	1 * 256Mb PC3700 OCZ
Other	ATI R9000 Passive Cooling Silverstone EFN-300 300W Passive Cooled PSU Maxtor 120GB IDE HDD

in-take temperature was measured at 22°C for all tests, but temp fluctuations, different mounting and user error can account up to 1-3°C of inaccuracy in the obtained results. Please keep this in mind when looking at the results. Each heatsink was tested repeatedly; if we got questionable results the test was restarted.

Noise level of each HSF combo was recorded with SmartSensor SL4001A, the sensor was placed ~20cm away from the side the closed case. The lowest dBA reading in the test room was 36-37dBA with everything turned off!

System was stressed by running K7 CPU Burn for 30min (after Thermal Compound's burn-in); this application pushes the temperature higher then any other application or game we've yet encountered. Speedfan was used to log maximum obtained temperatures.

Arctic Silver kindly send us their "Lumière" thermal testing compound which has the same colour as Ceramique, but only a break in time of 30min!

Arctic Silver's ArctiClean was used to clean off thermal paste of the CPU and heatsink between tests

Fans used for comparison

To eliminate as much variables in the tests I test each heatsink with a "reference" fan if I can mount them. If the HSF comes with its own fan, I will compare the performance of that fan to the reference one I use.

Delta NFB0912L 92mm: 42CFM

Papst 120mm 4412 F/2GLL: 40CFM


The Case

Since I'm only using an Athlon 3200+ for my tests, it would be interesting to overclock the CPU so its maximum heat output increases and it can simulate a higher clocked Athlon 64.

I recently purchased a power meter similar to this. Doing a few basic measurements with the test system gave these results for full system wattage usage.

Athlon S754 3200+ @ 2200Mhz - 1.5v: idle: 67Watt / Load: 125Watt

Athlon S754 3200+ @ 2420Mhz - 1.7v: idle: 78Watt / Load: 165Watt

In my days of Athlon XP HSF testing an increase of 0.1v vcore would result in 4-6°C higher CPU temps, so without much surprise the temperature results in this roundup with the 1.7v Athlon 64 will be much higher.

Noise was recorded approx. 20cm away from the side case, here's photo of how the dBA meter was position opposite the case and the test-room.



2 different test scenarios were configured as such:

Test Setup 1: Overclocked and Silent

- Athlon 64 3200+ @ 2400 @ 1.7v vcore
- Antec Sonata II + Silverstone Passive PSU
- AcoustiFan DustPROOF 120mm @ 5v in the rear as outtake (mounted with soft-mounts)
- nVidia TNT2 Passive cooled video card
- Noise produced with system running without HSF fan: 36-37dBA @ 50cm

Test Setup 2: Stock speeds and Silent

- Athlon 64 3200+ @ 2200 @ 1.5v vcore
- Antec Sonata II + Silverstone Passive PSU
- AcoustiFan DustPROOF 120mm @ 5v in the rear as outtake (mounted with soft-mounts)
- nVidia TNT2 Passive cooled video card
- Noise produced with system running without HSF fan: 36-37dBA @ 50cm


What was measured?

The CPU temperature was measured with SpeedFan and highest value recorded

Temperature of air coming into to the case at the front

PWM temperature through SpeedFan, this represent the area around the CPU socket, the power management caps which you see on a motherboard, they are there to make sure the power which is fed into the motherboard coming from the PSU is filtered and delivered the CPU and other components. Too high temperature will cause Vcore fluctuations which in turn causes system instability.

Onto the test results ->