Apex MI-008: A Cheap Quiet mini-ITX Case?

Viewing page 7 of 8 pages. Previous 1 2 3 4 5 6 7 8 Next


The stock chipset fan is a squirrelly one — very loud and whiney at full speed. It does undervolt well; it becomes bearable at about 7V though you can hear it tick from up close. The board has a fan control option in the BIOS with a minimum 50% setting for the SYS_FAN header. When we plugged the chipset fan into it, the fan only slowed down to the equivalent of 10V, which was unacceptable. For D945GCLF2 users with Windows, SpeedFan can be used instead, as the Speed01 option allows for full fan control right out of the box. For some reason, the controls are backwards — setting it to 100% turns the fan off, while 0% sets it to full blast. We set it 80% which resulted in a speed equivalent to about 6V. With the fan undervolted and the board on an open test platform both the chipset and CPU heatsink became very warm at full load — between 45°C and 50°C according to our IR thermometer.

System Noise Level
Chipset Fan Speed
SPL @ 1m
SPL @ 0.5m
12 dBA
15 dBA
13 dBA
16 dBA
6V +
Yate Loon @ 5V
13 dBA
16 dBA
21 dBA
25 dBA

The system, with the chipset fan turned off (not recommended — this was done for reference only), the measured noise level at one meter was only 12 dBA. Our noise floor is only 11 dBA, so this is an excellent result. The stock power supply unit is very quiet, even by our standards. The power supply fan did not increase in speed when the system was stressed, so the idle noise level was the same as during load.

13 [email protected] was recorded with the chipset fan at 6V. Adding the Yate Loon 120mm fan at 5V did not affect the measured SPL or our sound recordings. When the chipset fan was cranked up to 12V, its default setting, the noise level increased to 21 [email protected] This isn't bad at one meter distance, but at half a meter it was quite loud. For a small PC like this, we assume it is more likely to be placed on a desktop rather than underneath it, so the half-meter measurements are done in addition to the usual one-meter readings to better represent what it sounds like in the real world.

Impressively, the recorded noise level of 13 [email protected] and 16 [email protected] is only 1 dBA more than the recently reviewed mCubed HFX Micro, which features a completely passive cooling system. The Micro also happens to retail for 545 Euros (US$720). That's a lot of green for a barely noticeable difference in noise.


Thermal testing was done with the chipset fan undervolted to 6V — the lowest we were willing to go given the how hot the chipset heatsink got to the touch on an open testing platform. The temperature readings given by SpeedFan are somewhat difficult to decipher as they do not have proper labels and Intel does not provide a hardware monitoring utility for their Atom boards. Core 0-3 we can assume refer to the processor (dual core with hyperthreading = four logical cores) while "Ambient" most likely correlates to system temperature. "Remote 1," the remaining sensor with the largest differential is probably the chipset temperature.

Temperature Comparison
Test State
Load with Case Fan @ 5V
Remote 1
Remote 2
Core 0 / 2
Core 1 / 3
Chipset fan set to 6V (80% using SpeedFan), full CPU and GPU load applied (Prime95 + ATITool).

If our assumptions are correct, system temperature did not vary greatly, though it was coolest with the Yate Loon fan exhausting air out the right side panel. Without the added fan, chipset temperature increased 20°C when the system was stressed. The additional airflow from the 120mm fan effectively cut this increase in half. CPU temperatures were similarly affected. Given the dramatic difference, we highly recommend a side fan — you can reduce the noise level significantly and improve thermal performance at the same time.

Previous 1 2 3 4 5 6 7 8 Next

Cases|Damping - Article Index
Help support this site, buy the Apex MI-008 TOWER BLACK P4 CHASSIS TWR W/250W ITX PSU+SATA POWER SUPPLY from one of our affiliate retailers!