Viewing page 4 of 5 pages.
Previous 1 2 3 4 5 Next
TESTING
Test Platform
- Intel
Pentium M 770 with a measured power draw of 23.3W
under CPUBurn. This number includes power efficiency losses in the VRMs on
the motherboard.
- AOpen
i915GA-HFS motherboard
- 2 x 1024 MB Corsair DDR2 RAM
- Hitachi E7K100
100 GB notebook drive
- FSP Green PS
400W power supply
- Arctic Silver
Ceramique Thermal Compound
- Two-level plywood platform with foam damping feet. Motherboard on
top; most other components below. Eases heatsink changes and setup.
On the test bench...
Measurement & Analysis Tools
- CPUBurn
processor stress software
- AOpen PowerMaster
software to show CPU temperature, clock speed, and core voltage, as well as
fan speed.
- B&K model 1613 sound level meter
Official SPCR mascot, hard at work.
The two samples of NorthPole XE were tested in a direct comparison with the
stock AOpen heatsinks. The tests measure the combined noise of CPU
and northbridge fans.
Noise measurements were made with the fan in the power supply temporarily stopped
to ensure that it did not affect the results. A notebook drive was used to minimize
the effect of extraneous system noise; the ambient noise level was measured
while the drive was running.
Load testing was accomplished using CPUBurn to stress the processor. Our usual
temperature measurement software, SpeedFan, did not work on the AOpen board,
so AOpen's PowerMaster utility was used to read the thermal conditions of the
CPU. Unfortunately, no equivalent software could be found to reliably measure
the temperature of the northbridge.
The ambient conditions during testing were 20 dBA and 21°C.
TEST RESULTS
Our original intent was to use the Smart Fan settings in the BIOS to adjust
fan speed as needed to maintain adequate cooling. Unfortunately, these settings
did not seem to have any effect while Windows was running; All fan headers supplied
a full 12V in Windows, no matter what settings were used. A software version
of the fan controller, called SilenTek, was also downloaded and installed, but
it proved to be little better, either running the fan at full speed or not at
all, with almost no adjustable range in between. Even SpeedFan let us down ?
it would crash the whole system if we tried to force it to
run.
Eventually, we decided to run the board in its stock form, with all headers
supplying a full 12V. This gives the NorthPole XE a bit of an advantage, since
it is thermally controlled internally, independent of its input voltage. Realistically
speaking, both heatsinks could be used with SilenTek to cool the processor fanlessly
in idle, but the repeated starting and stopping of the fan would be
more irritating than constant fan noise.
| Microcool NorthPole XE Whisper vs. AOpen stock
heatsinks |
|
CPU State
|
Temperature
|
Fan RPM
|
°C Rise
|
°C/W
|
Noise
|
| Microcool NorthPole XE Whisper (w/o pads) |
|
Idle (SpeedStep enabled)
|
28°C
|
1,400
|
7
|
0.30
|
24 dBA@1m
|
|
CPUBurn
|
61°C
|
2,700
|
40
|
1.72
|
27 dBA@1m
|
| AOpen stock heatsinks |
|
Idle (SpeedStep enabled)
|
26°C
|
3,500
|
5
|
0.21
|
32 dBA@1m
|
|
CPUBurn
|
61°C
|
3,500
|
40
|
1.72
|
32 dBA@1m
|
Temperature: In degrees Celsius.
Fan RPM: Fan rotation speed, in rotations per minute
°C Rise: Temperature rise above ambient at load.
°C/W: Temperature rise per Watt, based on CPU's known power dissipation
Noise: SPL measured in dBA@1m distance with high accuracy B &
K SLM
*Test stopped for fear of damaging the CPU |
Initial testing showed the NorthPole to be a very poor performer. CPU temperature reached 80°C on CPUBurn. Then we
started tinkering and noticed that pressing down on the heatsink
instantly dropped by >10°C. This suggested that there
wasn't enough pressure on the CPU die. The cause was fairly obvious: The pads designed to take protect the edges
of the CPU die from pressure during installation were too thick. As a result, the pressure between HS base and die was too low.
The pads were removed, and a second set of tests run. This time, the NorthPole
XE turned in acceptable results, almost identical to the stock
heatsink. However, the twin NorthPoles were quieter than AOpen's fans, especially
in idle. Even with the CPU heavily loaded, however, they remained substantially quieter than
AOpen's stock heatsinks.
The fan on AOpen's tiny northbridge heatsink is the worse of the two stock HSF. It generates a shrill whine. If the fan controller on the motherboard
worked properly, the
northbridge fan would not run at full speed, which would make the noise of the stock cooling system much more palatable. With the northbridge
fan temporarily stopped and only the CPU fan running, the noise was roughly the same volume as the NorthPole
HSFs at full load. However, it is possible that running the NB fan at a lower speed is not an option because the stock HSF does not provide enough cooling otherwise.
Changes in NorthPole's fan speed were gradual, and only occurred
after the processor had been under heavy load for some time. For most usage
patterns, the fan should rarely ramp up. Rapid changes in fan speed could be
produced by increasing heat transfer by pressing down on the heatsink, but this
is hardly typical behavior. However, it does suggest that contact pressure was
still not as high as it could have been. We could not monitor the fan
speed of the northbridge fan, but we never heard it increase substantially
in speed.
There was some intermodulation noise between the fans as they increased
and decreased in speed ever so slightly. Another issue was a ringing tone somewhere
in the midband; this problem was especially
noticeable when the fan sped up under load. Further investigation revealed that
the aluminum fan shroud was the culprit. The noise could be eliminated by putting
pressure on the shroud to prevent it from resonating. The impact of damping
the shroud was significant ? as much as 2~3 dBA@1m under load.
In terms of noise character, the NorthPoles had surprisingly deep voices for
such small fans, although the dominant pitch tended to change with distance. From one meter away the primary noise was
the resonance mentioned above, but the noise turned into a lower-pitched growl
closer in. With the exception of the resonant tone, the noise
character did not change a lot under load.
Because the northbridge fan was by far the noisier of the two AOpen coolers,
one final test was run using a NorthPole XE on the northbridge and the stock
cooler on the CPU. This proved to be the most successful combination. The noise
level dropped to 27 dBA@1m in both idle and under load. This is the same
noise level as the twin NorthPoles achieved under load ? but the character
of the noise was nicer, since the interaction between the
two NorthPole fans was absent, as was the resonating shroud. The noise was more broadband
and easier to ignore than the two NorthPole fans together. As an added bonus,
this solution only costs half as much, since only one NorthPole cooler is needed.
| Help support this site, buy from one of our affiliate retailers! |
|
