Antec P180 Review, Part 2: The Whole Nine Yards

IMPACT OF VIDEO CARD HEAT

The effectiveness of opening up the top hole suggested that the difficulty with CPU cooling was due primarily to the hot NVidia 6800GT sitting right under the CPU. To test this theory, a lower powered VGA card — an AOpen Aeolus 6600GT — was installed in place of the 6800GT. Configurations 1 (top vent sealed, back 120mm fan on low) and 2 (both 120mm fans on low) were then tested again with this slight system change.

Intel High End System: 6800GT vs. 6600GT (Stressed by 2 x CPUBurn)
Configuration	VGA	CPU Temp.
1	6600GT	63°C
	6800GT	74°C*
2	6600GT	60°C
	6800GT	72°C*
*CPU Throttling occurred in these configurations. AC Power Draw at full load: >> w/ 6800GT: 236W << >> w/ 6600GT: 210W <<

This time, the CPU did not throttle. In fact, with the top vent sealed it stabilized around 63°C — a 10°C difference compared with the same configuration when the 6800GT card was in place! This is a substantial difference.

The total difference in system heat can be estimated by examining the difference in AC power draw between the two systems: With the 6800GT, AC draw under load was 236W versus 210W with the 6600GT. At this power level, the Seasonic S12-430 is ~80% efficient, meaning that the actual difference in heat is roughly 21W. This is a huge difference for a GPU at idle.

Any nVidia 6800GT video card is a challenge to cool quietly.

VGA COOLING

Once the best means of cooling the CPU quietly was established, further tweaking could be done to improve the the temperature in the rest of the case, especially around the video card. The 6800GT card was returned to the system, and Configuration 4 was kept as the baseline for experimentation because it could cool the CPU adequately, and do it more quietly than Configuration 5, the only configuration that was cooler.

1. The CPU was overclocked by 5% under load. The FSB was run at 210 MHz, bringing the CPU frequency up to 3.8 GHz, on par with the top of the line Intel 670 processor. The system remained completely stable under this modest overclock.
2. The built-in fan on the VGA card was undervolted to 8.25V, the level at which the fan noise dropped to the level of the other fans.
3. The 120 x 38mm fan in the PSU chamberwas removed on the assumption that the airflow provided by the Seasonic S12 would be enough to cool the Raptor installed in the lower drive cage.
4. The upper drive cage and front fan holder were removed entirely to minimize the airflow impedance near the intake.
5. The VGA duct was also removed in hopes that the airflow around the VGA card would already be sufficient.
6. Airflow in the upper chamber was provided by the the rear Antec TriCool fan, set on Low.
7. The top fan was left installed but unplugged, and the top vent was uncovered to ensure optimal CPU cooling
8. No fan was used on the Ninja heatsink, thanks to our experience in the previous section.

Because we do not have a reliable, repeatable way to load the VGA card, thermal comparisons were done with the system at idle. VGA temperatures were read from the NVidia driver utility that is installed in Windows Display Properties. Keep in mind that the VGA fan was undervolted to 8.25V throughout testing, so the temperatures are higher than they would be if the card was run as intended.

VGA Duct Configurations

Because of compatibility issues (see the section above on installing the VGA duct), a low profile Antec 80x20 mm fan (tested in the above section on Fan Characteristics) was installed in the VGA duct of the P180. The fan was undervolted to 10.8V using a Zalman Fanmate, with the expectation that, if all went well, further undervolting could be done.

As mentioned before, this fan is quite impressive, especially for a thin, low profile fan — at least while it is spinning in free air. Unfortunately, installing it in the VGA duct ruins it completely. At full tilt, the duct greatly amplifies the volume of the fan and also accentuates a sharp, unpleasant buzz. This is entirely the fault of the duct, as it is prone to resonating and amplifying the vibrations produced by the fan. Both the duct material as well as the air in the duct contribute to the resonance. At 7V and below, the nasty effects of the duct are much subdued, but so is airflow and cooling.

In fact, the sonic effect of the duct was so noticeable that it was worth testing the duct outside of the P180 case. The results were quite dramatic.

80 x 20mm Antec Fan Characteristics
Position	Setting	SPL
Free Air	12V	22 dBA/1m
	7V	<17 dBA/1m
	5V	<17 dBA/1m
in VGA duct	12V	30 dBA/1m
	7V	21 dBA/1m
	5V	18 dBA/1m

The VGA duct was tested in the case in a number of different configurations, summarized in the table below:

P180 VGA Duct
Duct	GPU Temp. (idle)	SPL
Not installed (Configuration 4)	69°C	25 dBA/1m
Installed	67°C	27 dBA/1m
Installed at full extension (fan ~1.5" deeper into the case)	69°C	27 dBA/1m
Installed with fan reversed (exhaust instead of intake)	71°C	27 dBA/1m
Not installed (VGA card fan @ 12V)	57°C	29 dBA/1m

As expected, the GPU temperature went down with the duct installed. However, it's quite surprising how small the decrease was. The biggest (and only) improvement was a measly 2°C — perhaps not even statistically significant. Some configurations even made the temperature rise, indicating that the duct was disrupting airflow around the card.

The minimal effect of the VGA duct may have something to do with the design of the cooling system on our test VGA card. Like many other high-end cards on the market, the Aeolus 6800GT uses a duct style heatsink that blows air across the card. This means that the bulk of the heatsink is actually covered up. The surface area that can benefit directly from the additional airflow of the VGA duct is minimal. The VGA duct is probably most effective when the VGA card has large, open heatsinks, such as AOpen's new passively cooled cards.

The smooth heatsink on the VGA card is designed to be be used as a duct with the stock fan, and is ineffective when used with the external duct fan.

Positioning the duct to blow across the back half of the card — where the exposed heatsink fins might benefit more from increased airflow — did not have the desired effect; instead the GPU temperature rose. Reversing the direction of the duct fan was even worse. Not only was there a danger of contact between the spinning fan blades and the VGA card, but temperatures were worse then with the duct removed entirely.

The minimal cooling benefit of the VGA duct in our configuration was not worth the 2-3 dBA/1m increase in system noise. If the GPU temperature becomes excessive and requires an increase in noise, it is much more noise-efficient to simply increase the voltage of the fan on the VGA card. Running the fan at full speed did sound louder than with the duct installed, but here the drop in temperature was 10 degrees, not two.

MP3: P180 "Hot Potato" Configuration 4 (no duct installed): 25 dBA/1m

MP3: P180 "Hot Potato" with VGA duct installed: 27 dBA/1m

MP3: P180 "Hot Potato" with VGA fan at full speed, no duct installed: 29 dBA/1m

Front Fan Configurations

An alternate method of cooling the VGA card needed to be found, preferably not increasing the voltage of the stock fan. With high hopes, the Nexus 120mm fan was reinstalled in the front fan bay, this time to gauge its effect on the VGA card. Testing was done in combination with the VGA duct (with the Antec 80x20mm fan at 10.8V running in it) in hopes that the two together might achieve better results than either could singly.

P180: Front Fan
Front Fan Voltage	VGA Duct	CPU	GPU Temp.	SPL
no front fan	Removed	65°C	67°C	25 dBA/1m
12V	Installed	66°C	65°C	29 dBA/1m
12V	Removed	70°C*	65°C	28 dBA/1m
5V	Installed	67°C	66°C	27 dBA/1m
5V	Removed	69°C	69°C	25 dBA/1m
The processor was under 100% load with 2xCPUBurn * CPU Throttling occurred in this configuration.

At 5V, the Nexus was insignificant, both thermally and acoustically. A slight increase in CPU temperature indicated that more heat from the VGA card was finding its way across the fins of the CPU heatsink.

At 12V, the effect on temperatures was similarly unimpressive, at least while the VGA duct was installed. Furthermore, the volume of the system noise increased by 2 dBA/1m — a needless increase.

However, with the VGA duct removed, the GPU temperature actually dropped slightly, perhaps in response to the increase in circulation over the top (trace) side of the card. The 65°C idle for the GPU without the VGA duct installed was the lowest idle temperature for the GPU during all testing. Although the SPL for the Nexus alone measured slightly higher than the VGA duct alone, subjectively the Nexus 120 fan sounded much better. At 12V, the Nexus' main contribution to the system noise is an increase in turbulence noise and a low, smooth hum. This is considerably more pleasant than the droning buzz of the resonating fan in the VGA duct.

The downside of running the 12V Nexus without the VGA duct is a 5°C increase in CPU temperature, enough to force the CPU to start throttling.

The increase in CPU temperature was enough to convince us that a better temperature for the GPU could not be easily achieved with the 6800GT without compromising system noise. The 69°C idle is high, but acceptable. According the NVidia driver, the GPU core can hit 127°C before it throttles itself back, which still provides a 50°C cushion. Some minor graphical glitches were detected while running 3DMark05, but Doom III, well known for its steep system requirements, was quite playable with all graphics options enabled. A hardcore gamer would no doubt be happier running the VGA fan at full speed and paying the acoustic price, but it was decided that the borderline thermal performance was worth the acoustic benefit for our test.

In the end, it seems likely that for the best cooling and lowest noise, a capable, massive aftermarket heatsink needs to be installed on this hot video card, perhaps combined with a low speed fan in the VGA duct. There was not enough time to fully explore this hypothesis.