• Home
  • blog
  • Quiet PC for Torrid Thailand

Quiet PC for Torrid Thailand

blog image

An old friend from another life asked about a quiet computer… and the end result is a PC that’s been custom built to be quiet in the tropical heat of rural Thailand. Its components include an Athlon64 X2-3800+, EVGA nVidia 6800GS-256, Samsung 200GB HDD, and six fans in a modified Antec P150 case; still, measured SPL is just 23 dBA@1m. Quiet is in the details of this high heat optimized PC.

Jan 10, 2006 by Mike Chin

Along with the growth and increased visibility of SPCR, a degree of personal “fame” or “notoriety” has come my way. (I use both terms with some amusement. The latter is more likely to be used by some members of the PC industry; the former only by naive SPCR enthusiasts.) As a result of this exposure, I get asked occasionally to build someone a quiet computer. If I am not drowning in lab work or articles or whatever crisis happens to be on us, and if the project is interesting in some way, I usually comply. I charge for my time and expertise, and use review samples when I can, so there’s a little financial incentive. More importantly, it’s a good opportunity to keep in touch with the roots of how this site developed: Hands-on experience and experimentation building quiet/silent computers. It gives me a chance to work in a practical way with gear we’ve recently tested, to stay attuned to real-world applications of the products.

View during a trek in northern Thailand, Jan 2005. It may look cool, but it was actually sweltering.
It’s somewhere near the destination of this PC.

These custom computer requests usually come from people who know me personally, friends, friends of friends, relatives, etc., mostly in the Vancouver area. Sometimes they come from SPCR readers. Once in a while, though, I get requests from people much farther away, in other parts of Canada or the US. The project I just finished takes the cake for unusual destinations, though. It’s for an old friend who I haven’t seen in more than 20 years, who lives now in the north of Thailand somewhere in the vicinity of Chiang Mai, in a quiet rural area that’s apparently on the edge of the jungle. Frank, who hails originally from England, retired and built his dream oasis there after an unexpected inheritance made it possible for him to give up his university teaching job in Hong Kong.

Frank’s original e-mail asked my opinion on whether a particular Hewlett Packard PC would be quiet compared to his very quiet laptop, which is now a few years old. My assessment, after a close reading of the specs on HP’s website: No Way! We discussed various desktop and laptop options back and forth… and in the end, he asked me to build him a desktop with enough computing power and features to last several years, with some ability to play games, with some expansion capability, a computer that would be as quiet as the computers I build and use for myself.

There was some initial hesitation on my part, for several reasons:

1) The trials and tribulations of shipping a PC across the Pacific. I’d never tried to send anything this big (probably over 40 lbs) that far. Would the shipping companies do a good job? What if there was damage?

2) If transit damage did occur, how could the PC be repaired, especially if it involved anything I had modified or customized for low noise? Was there anyone near Frank who could tackle this? (The answer was no, which dictates extreme care in packing.)

3) Technically, it was a big challenge. The temperature range in my working environment here in Vancouver ranges from about 20°C to maybe 28~29°C in summer. Once in a great while, it might actually reach 30°C. In other words, thermally, it’s mild. Thailand, by contrast, is a hot and humid place. In the hot season, 40°C in the shade is almost routine.

However, as you know already since you’re reading this article, I could not resist the challenge of building a quiet computer that could stay quiet even in an environment as hot as Thailand can be.


Frank is not a hardware nut. His interest in computers is in using them productively and for fun. So I was free to make choices on his behalf based on his stated requirements and his budget. These are the components I chose:

Antec P150 mid-tower case: This is about the best “conventional” layout mid-tower case for quiet computing I know of. I would have gone for the P180, but its bulk and weight would have doubled the cost of air freight, estimated at ~$300 with the P150. With one major modification and the right choice of components, I would make the P150 the equal of the P180 for Frank’s system.

Antec NeoHE-430 power supply: It’s the PSU that comes with the P150. It’s quite efficient, has a great fan and a good fan controller. It’s also an 80mm fan unit, which is so rare these days and is perfect for the PSU fresh air intake duct system I had in mind for this system. It has Active PFC, can help in less stable AC conditions, and it’s a universal AC input design, which is great for Thailand’s 220VAC. The original PSU would not work with the motherboard of my choice, however, so I had to wait for an improved compatibility replacement, which arrived just before Christmas. The replacement PSU works just fine.

Asus A8N-SLI Premium motherboard: There was a handful of options among nVidia nForce4 motherboards, but this one won me over with its huge feature set and the passive heatpipe cooling of its chipset and voltage regulators.

AMD X2 3800+ dualcore processor: It was a natural choice for processing power, longevity and power efficiency. For all the tasks Frank considers important, this is more than adequate now, and should remain so for several years. It’s also very good value.

Scythe Heatlane Zen NCU-2000 heatsink: It is meant to be fanless, but I’d be using a fan on it. There are many ways to attach a Nexus 120 fan. The big size, the widely spaced cooling fins, the unidirectional flat Heatlane heatpipe — all these would be invaluable for quiet cooling of the CPU in Thailand’s hot climate.

Nexus 120 fan (for CPU heatsink) + 2 Nexus 92mm fans for the front intake

EVGA eGeForce 6800GS graphics card: It seemed a good card, a big step down in price from the top-of-line cards, but still quite powerful with 256MB of DDR3 memory. Bottom line: The Salesman talked me into it.

Arctic Cooling NV Silencer 5 (Rev 3): To replace the noisy stock heatsink/fan on the EVGA video card.

OCZ Dual Channel PC3700 Gold Edition (Rev.2) 2x512MB DDRAM: Reliable, fast pair of sticks for a gigabyte of RAM.

Samsung SpinPoint P120 SP2004C 200GB SATA HDD: The earlier P80s are quieter, generally, but this one is pretty close, and it does have more capacity and appears to be a true SATA.

LG GSA-4167B – 16x DVD +/- R/RW/RAM, Double-Layer: An all-purpose optical drive that’s pretty smooth and quiet, and seems to work well.

Memory Card Reader/ Writer + 1.44MB Floppy Disk Drive: A combo device to handle various memory cards and the old floppy.

Modder’s Mesh: For modifying the case.

Various other materials and parts, described later in the text.

Most of the system components were purchased through Canadian retailer NCIX, who supports SPCR with advertising. They offer very good pricing and have a good good online store.


The principles applied to the design of this PC are the same as that espoused for silent computing throughout SPCR:

  • Use cooler and quieter components whenever possible.
  • The case must be sturdy and provide for good, minimally restricted airflow
    through it, in and out.
  • Easy and direct access to cool outside air for hot components; fast and efficient evacuation of heat.
  • Vibrating components should be mechanically isolated to reduce noise by conduction to the chassis.
  • Fans should run as slowly and quietly as possible while providing good cooling
  • There should be substantial cooling headroom in anticipation of the hot operating environment of this system.

The project took shape in the waning weeks of 2005. It was finally finished, tweaked and fully tested over the last few days. Here’s a photo of the finished system, ready for packing after almost a week of intensive stress testing, which it passed with flying colors.

Frank’s system, ready to be packed.


Those who are familiar with the appearance of the P150 probably wondered about the last photo on the previous page. The grill at the top front is external evidence of the only serious modification I made to the P150 for this system: A fresh air intake duct for the power supply to prevent it from being affected by heat from the rest of the computer. This modification ensures that the power supply fan only has to deal with its own heat, and that it runs as quietly as possible without risking overheating even under extended periods of very high load. A PSU intake duct is not a new idea; I suggested it to Leo Quan years ago, and it’s been adopted by many SPCR readers, including Ralf Hutter, with his P150.

Here’s a photo of an unmodified P150 case, from Ralf Hutter’s review in October:

The unmodified P150 is an elegant looker.

Below is a closeup of the modified portion of the P150 I built for Frank. It’s a piece of Modder’s Mesh, 22 gauge perforated steel, courtesy of mnptech.com and coolcases.com. Wedged behind it is a piece of thin, light open-cell foam to act as a dust filter. This accounts for the dark color of the holes in the mesh.

Perforated steel grill lends a somewhat more high tech look.
It compliments the brushed aluminum inset of the P150 front panel .

The steel mesh was cut slightly larger than the opening using tin snips, then the four edges bent 90 degrees manually using a hammer, clamps and wood blocks. The final piece fitted tightly into the 2-optical bay space allotted for the duct. I used a hot glue gun to affix the steel grill securely to three sides of the P150 front bezel. The dust filter was cut from a piece of foam found beneath a motherboard in its retail box; it is very commonly used for this shock-protection purpose.

Light open-cell foam acts as a dust filter; duct is visible above LG optical disc.

The following photos illustrate the making and installation of the duct. It’s pretty self-explanatory.

What it looked like before.

The duct was cut with a utility knife from an old poster mounted on 1/4″ foamboard.
Foamboard has a polystyrene foam core, and is laminated on both sides with a paper stock.
The dimensions had to be correct, and the cuts had to go through one of the paper skins but not the other.

This is how the duct was inserted.


Once in place, two holes were poked on each side…

…so that ordinary plastic wire straps (zap straps) could be used to secure the duct.

The duct had to be cut open a bit on this side for the PSU output cables.

Inside the duct: There was a small gap on the right, which I closed with a bit of sticky-backed foam.

Here’s how the finished duct looks from below.


The Air Intake

Behind the bezel, there is a dust filter over a nicely unrestrictive grills, and behind them a pair of Nexus 92 mounted with AcoustiProduct’s fan grommet kit. The fan panel opens to provide access to the hard drive, which is mounted using the rubber suspension.

AcoustiProduct’s soft fan mounting kit was used for the three case fans.

The upper fan of the two intake fans blows outside air across the hard drive and to the CPU area beyond. It runs at 7V constantly, using the 7 volt trick. The lower fan, which cools the HDD a bit by blowing beneath it, but is there mostly to insure that the VGA cooling fan gets adequate outside air. This fan is plugged into a Q-fan controlled motherboard fan header, and receives the same thermally-controlled voltage feed as the CPU fan. Nexus 92 fans are very quiet even at 12V; at these reduced voltages they are even quieter.

Airflow Patterns

The photo below shows the airflow patterns in the case. There are seven individual noise makers in the system: One hard drive, and six fans, including the VGA cooling fan, and the fan in the Antec NeoHE 430 power supply, which is very smooth and quiet. The dark line below the PSU and duct indicate that the PSU is in one thermal zone and the rest of the system is in another; the heat from one area does not pass into the other.

Typical airflow paths.

The sheer number of fans, each on different circuits, ensures minimal risk of catastrophic cooling failure. It also means they can all spin pretty slowly, and collectivity make very little noise while still moving a large volume of air. It’ easiest to think of the fan/cooling design as a relay system, with the intake air starting at the front and being passed from fan to fan, cooling components along the way, and rising in temperature, before finally being exhausted, through the PSU, the back case exhaust fan, or the VGA cooling fan.

The side where much of the wiring is hidden for improved airflow.


Before the duct: Innards don’t look so different from an ordinary PC… or do they?

There’s a lot to look at in the above picture. Let’s start at the back of the case.

Antec 120mm Tricool fan for Case Exhaust

The standard stock one is black; this one is a clear retail version that has a more convenient 3-pin plug that goes into standard motherboard fan headers. I would have left the stock one in place, but during testing, it developed a nasty buzz that I attribute to a manufacturing variance. I could have replaced it with a Nexus 120 fan, which is quieter, but it’s limited to ~40 CFM airflow at full speed, which may not be enough in hot weather. The retail Tricool has the same airflow characteristics as the stock black one, which we measured in the lab to have 75, 47 and 28 CFM airflow at the three switchable speeds. At the top speed, it’s loud, but should an emergency thermal situation arise, the extra cooling power could be very helpful.

The fan itself was mounted using soft silicone rubber grommets from a fan mounting kit by AcoustiProducts, which we reviewed last year. Using AcoustiProducts’ metal screws proved to be counterproductive here. The screws themselves make contact with the edge of the panel mounting holes; enough so that vibration noise is not much reduced. I came up with a way of securing the fan without using screws. A picture is sometimes worth many words:

Two short pieces of clothing elastic hold the fan in place.

Clothing elastic is one of my favorite quiet PC building materials. I get it from a nearby warehouse-style textiles store for around a dollar a meter. It’s the kind of elastic string used in jackets as a waist-cinch with a spring-loaded lock. It has multi-core rubber strands in the core with a stretchy woven material around it; it’s extremely durable. Great for hard drive suspensions and all kinds of other vibration reduction applications… like mounting fans. It makes for more effective vibration reduction than the screws provided in the AcoustiProducts fan grommet kit.

At the minimum speed the Tricool 120 is pretty quiet, measuring 20 dBA@1m in free air. But mounted on the back panel of the P150, even with the damped mounting, I didn’t think it was quiet enough. The speed is just a bit too high, and some still vibration gets transferred to the back panel, which resonates just enough to add a low frequency hum that further increases overall noise. The system measured 24~25 dBA@1m from the back with this Tricool fan set on low.

My solution was to add a couple of zener diodes in-line (series) to reduce the voltage going into the Tricool fan, so that it would run at even slower speed. (For more information on zener diodes, see Cpemma’s excellent web site on fan noise control.) The input voltage is dropped from 12V to around 10V. With the zener diodes, when the Tricool fan switch is set to low, it’s now slightly under 23 dBA@1m from the back. The dangling 3-position switch was pushed through a small lockhole to the outside of the case so that it is accessible without opening the side cover.

Scythe NCU-2000 heatsink with Nexus 120 fan

This innovative big heatpipe heatsink is marketed as fanless model. We tested it a while ago, and found it to be usable in a typical system with some airflow from peripheral fans, but there’s no way I’d ship a fanless PC to Thailand. I fastened a Nexus 120 fan on the NCU-2000 using enamel solid-core wire. (Clothing elastic is not wise here due to the stress of high temperature.) It is set up to blow the air through the HS fins towards the back of the case. It’s a basic straight through push-pull arrangement: The CPU heat is pushed by the Nexus 120 fan towards the Tricool 120 fan, which pulls it out of the case.

One thing to note in the picture below: The fan is actually mounted slightly off center, and it occupies the memory slot closest to the CPU. Yes, it is actually jammed right up against the memory slot. This was done to get the fan’s airflow as close to the surface of the motherboard as possible. Why? To make sure the the board-mounted components, particularly the voltage regulators, get adequate cooling. This is an issue with most of these tower-style heatsinks. The older style of heatsink where the fan blows straight own does have a major advantage: There is always some airflow across the motherboard components. Allowing these components to run too hot lower reliability, longevity and power efficiency.

Nexus 120 fan on the Scythe NCU-2000 heatsink mounted as close to the motherboard as possible.

The fan mounting also ensures decent airflow for the heatpiped heatsink that pulls the heat from the chipsets.
Some of the voltage regulation components are visible on either side of the chipset heatsink.

The Nexus fan is plugged into the CPU fan header of the Asus board. The output of this fan header is thermally controlled by the motherboard. Q-fan is what Asus calls their board-embedded thermal fan speed controller, which is accessible in the BIOS under Hardware Monitor. It’s a simple on/off switch, with a manual selection for the target CPU temperature at which the fan get the maximum 12V. The range is 51°C to 82°C — the latter number seems absurdly high. During the testing, this setting was at 66°C, and the CPU fan never ramped up, which is perfectly logical, given that the CPU never hit higher than 47°C. Just out of curiosity, I lowered the setting to 51°C and stressed the CPU with two instances of CPUBurn, bringing the CPU temperature up to 47°C. Again, there was no change in fan speed.

This suggests that Q-fan will bring the CPU fan speed up only when the trigger temperature is reached, but it may take it from minimum to maximum speed as soon as it’s triggered. If so, this ramp-up in fan speed will definitely be audible. It is not quite as bad as it sounds, as the Nexus 120 is still very quiet even at 12V, measuring just 22 dBA@1m in free air. I changed the trigger temp to 60°C in the final build. Frank may want to re adjust this after he has had some time to work with the machine in his environment.

EVGA eGeForce 6800GS graphics card with Arctic Cooling NV Silencer 5 (Rev 3)

The vidcard is visible in the above photos. The main heatsink and the fan are on the underside (of the video card) but there’s also a passive memory heatsink on the trace side of the board, which benefits from peripheral airflow of the case exhaust and CPU fans. Revision 3 of the NV Silencer is compatible with nVidia 7800 models, and it has a much better-behaved fan. The first revision had some nasty ticking/buzzing effects which made them considerably louder than the original VGA Silencer. The improvement to the fan was apparently implemented in all AC VGA cooler models in Rev.2. The AC VGA coolers remain the only ones on the market to evacuate the heat of the GPU rather than just recirculate it inside the case. The AC VGA coolers help to lower not only GPU temperatures, but system and CPU temperatures as well.

After cooling the GPU, the air from the Arctic Cooling heatsink gets blown out the back panel through the next PCI slot.
Photo shows the exhaust duct of the AC

A Zalman Fanmate1 is used to control the speed of the AC NV Silencer’s fan. At full speed, this fan is too loud for a quiet computer. The fan speed setting was chosen on the basis of noise (quiet enough not to be identifiable as a discrete noise source) and cooling (which you’ll see later). It turned out to be a very slow speed; I would guess the fan is not getting more than 7V. Still it is plenty cool enough, even at very high, extended loads. The Fanmate1 is mounted on the back panel in such a way so that its little knob protrudes through one of the grill holes, and can be adjusted from the outside. Because the knob is so small, it may require smaller fingers than Frank’s to turn… but that shouldn’t pose much of a challenge.

Externally accessible fan speed controls.
Arrow points to knob on Zalman Fanmate1 controller.

Finally, a shot of the back panel.
Note exhaust grill below VGA slot for the Arctic Cooling VGA cooler to exhaust hot air.


The tools used to stress test and burn-in this computer included:

Ambient room conditions were 18 dBA for noise and 21°C for temperature.

The system remained stable and generated no errors during 48 hours of continuous Prime95 testing, done simultaneously with one instance of CPUBurn. For some stretches, three of the stress programs — CPUBurn, Prime95 and 3DMark05— were run simultaneously in order to force the highest possible stress on the system and maximize demand on the power supply. The following table summarizes the results of the testing. Quiet Mode is the normal mode in which the system is meant to be used.



Quiet Mode

Cool Mode

Noise (SPL, 1m from front of case)

22 dBA

28 dBA

36 dBA
Noise (SPL, 1m from back of case)

23 dBA

29 dBA

36 dBA
CPU Temperature, maximum



CPU Temperature, idle



GPU Temperature, maximum



GPU Temperature, idle



HDD Temperature, maximum



HDD Temperature, idle



Board Temperature‡, max



Board Temperature‡, idle



Idle system power draw



Idle system power draw w/Cool ‘n’ Quiet



Maximum peak system power draw







Nexus 120 fan on CPU heatsink (Q-fan*)

600 RPM

1100 RPM†

1100 RPM†
Tricool 120 fan on back panel

700 RPM

1100 RPM

1700 RPM
Nexus 92 intake, upper

900 RPM

900 RPM

900 RPM
Nexus 92 intake, lower (Q-fan*)

1100 RPM

1550 RPM†

1550 RPM†


‡ Board temperature is one of three sensors on the motherboard picked up by Speed Fan.
*Q-fan is the thermal fan speed controller in the Asus motherboard.
†Q-fan turned off.


Quiet Mode: The thermal results are particularly nice. 47°C was the highest temperature seen for the CPU at any time during the testing. At idle, it’s typically 27°C with Cool ‘n’ Quiet engaged. The hard drive typically sits at 25~26°C; it took a long time to reach 34°C using the check noise feature of Hitachi’s Feature Tool.

Remember that the system will reside in northern Thailand where ambient temperature can reach 40°C. I know Frank has air conditioning, but still, this PC has to be able to withstand native conditions. So add 20°C to those numbers (the difference between my test room ambient and the 40°C anticipated maximum of Thailand): We get 67°C for the CPU and 69°C for the GPU. If the HDD was stressed as hard as I did, it would reach 54°C, which is high, but the chance of this happening is extremely remote. In normal use, I never saw the HDD temperature go past 28°C. If the ambient was 20°C hotter, it would still be under 50°C, which is a bit hot, but probably about the norm for most big-brand PCs. All of these temperatures are well within normal operating limits. Since the stress tests put more load for a longer time than just about anything anyone would run on this PC, the machine is probably over-engineered, thermally.

Cool Mode: Just as a check for thermal cooling, Q-fan was turned off so that both the CPU Nexus 120 fan and the bottom intake Nexus 92 fan were running at full speed. The back Tricool 120 fan was sped up to a point where is began to add to the overall noise; this turned out to be with the rheostat knob just below mid point. The overall measured noise rose to 30 dBA@1m. Overall airflow through the case was substantially increased. The stress tests were run again, and the results are compiled in the Cool Mode column in the above table. The CPU temperature maxed out at 45°C. The hard drive temperature dropped to 23°C at idle and 29°C maximum. The maximum GPU temperature also dropped 10°C, down to 57°C. Those are pretty amazing numbers.

Coolest: This was an extreme test out of sheer curiosity, with all the fans that can be controlled set to run as fast as they can. The only difference between this and the Cool Mode was the speed of the exhaust fan, which went up from 1100 RPM to 1700 RPM. Noise jumped to 36 dBA@1m, the sound being almost entirely broadband “wooshing”. Cooling of the CPU improved a bit, but not much else changed. It appear that the under the test conditions, the airflow of Cool Mode had already brought temperatures down as far as they could go. With higher ambient temperatures, I would expect bigger differences between all three of these settings.


In late 2005, Antec NeoHE power supplies came under much criticism for incompatibility with many motherboards, including the Asus A8N-SLI Premium motherboard used here. The original PSU in the P150 used here did not work with this motherboard, although it worked with many other motherboards. It was replaced with a recently released revised NeoHE-430 that’s supposed to be free of previous problems, and sailed through 48 hours of continuous stress testing (at 155W~190W AC load). Still, I decided to give it yet another round of tests: Over a period of 2 days, it was booted into Windows, then turned off and back on almost immediately as often as I could remember to do it (while I went about my other work). In all, I would guess it remained on nearly 48 hours and was cycled on/off/on at least 100 times. No error was seen in this ad hoc “test”.

I will not cover the performance of this computer. The performance of an AMD Athlon64 X2-3800+ processor is well documented, as is the motherboard. The graphics card is fast, probably faster than Frank really needs right now. Suffice it to say it is much faster than my own main machine.


This PC is very quiet, just above the threshold of my hearing from a couple meters away in a room ambient of 18 dBA (at 1:00 AM). From three meters distance, it’s hard to decide whether I can hear it. At four meters, I definitely cannot hear it. The main sound character is a low level “wooshing” broadband noise. The computer is as quiet as the main Intel P4-2.8 / dual-HDD rig I run under my desk. The character of the sound is the same: Smooth, constant and unchanging, except for a touch of HDD chatter when it is seeking.

The 80mm fan in the PSU power supply, the 120mm Nexus fan on the CPU heatsink, and one of the two 92mm Nexus intake fan are all thermally controlled. They will speed up when things get hot. However, none of them ramped up even once during all of my testing. This is a very good sign, but I would expect that in >30°C weather, these fans will ramp up somewhat under load. They should!

The two manual fan controls on the back panel for the case exhaust fan and the VGA cooling fan assume some degree of hands-on participation in system cooling by the user. These fans could have been thermally controlled, but Frank would then lose direct control over the noise emitted by the computer. The price, of being a bit watchful during hot weather and turning up the fans if it’s called for, is a small one.


The final challenge is to pack this PC up so that it will survive the air journey to Chiang Mai from Vancouver without damage. After much thought, I’ve decided to fill all the empty spaces in the PC with packing foam so that if something should get loose due to external shock, it simply can’t go anywhere. I can’t think of a safer way to protect the components. Frank will have to open it up and remove all the foam when it arrives there. He needs to check all the connectos anyway, especially the SATA drive cables which are so easy to dislodge.

Starting to stuff the PC with bubble packing material…

The P150 comes in a typical Antec retail box with top and bottom end caps made of Styrofoam. I will use those but go a step further and stuff all the empty spaces with Styrofoam pieces or chips. Then I’ll double-box it in another carton so that if the box is dropped, it will be extra difficult to perforate it and do damage to the case.

It’s been clear to me for some time that to make a quiet or silent computer, you need to take a holistic systems approach. But the execution has to be good as well; silence is in the details. All it takes a one small flaw for noise to jump up. Only Frank can tell us whether this quiet computer is a success in the long run. Maybe I can persuade him to send me an e-mail about it in a couple of months.

* * *

Thanks again to Canadian retailer NCIX for helping to source the components for this system.

Comment on this article in the SPCR Forums.

Leave a Comment

Your email address will not be published. Required fields are marked *