Two things: (a) This is a temporary location for this page. Do not bookmark it. (b) I am working on a major revamp that, while incomplete, is much more complete and accurate than this site. The remainder of this page is pretty ancient, and does not contain much useful information. I'm going to take it down in not long.
This page is quite unfinished. Dunno how finished it'll ever get. Basically, I started putting down some notes as I ran into things. Then it started getting 100+ hits per day. At the same time classwork caught up with me, so I didn't update it for a while. Both versions contain portions the other doesn't. Both are often quite out-of-date (especially the original), but I think the revised is now the much more useful of the two sites.
Update:My computer is now basically silent. Here's what I did. I'll write up in more detail when I have time:
Note: This page is slightly out-of-date. The info on it is still good, but my machine is configured a bit differently at the moment. I'll update it when I get time again.
Right now, the most comprehensive site on the topic is The Silent Personal Computer. If I'm missing anything, or if you've managed to do something creative to make your computer silent, or if you just found this page useful, please drop me a line.
Neat tidbit I ran across: The 2ghz Pentium 4 can run at reduced speeds not only fanless, but heatsinkless. Tom's Hardware has the story. Amazing how much power usage drops for pumping the clock speed down. With a heatsink but no fan, this actually seems a reasonable solution to a fanless system. I don't fully understand why this works; the lowest power P4-2ghz burns 71.8W average, 95.7W peak. I was under the impression that the thermal diode slows it down to half speed. This would cut the dynamic power dissipation in half, and leave the static close to what it was. Even 35W should be able to fry a processor heatsinkless. But heck, Tom's Hardware is pretty reliable, and if his empirical evidence shows it works...
Computer noise comes from drives, cooling fans, as well as external peripherals like printers and scanners. They can be either amplified or attenuated by the computer's enclosure. Especially in schools, workplace environments, or other places where people are in the same room as computers many hours per day, this noise can have adverse psychological effects, and cause loss of productivity. A 1998 report published in the British Journal of Psychology found that excessive noise can diminish the performance levels of office workers by as much as 60 percent. The EPA considers noise levels of below 45dB indoors as adequate to prevent activity interference, although most sound consultants recommend 40dB. Above 55dB, voice intelligibilty begins to drop. According to the document, sound above 32db begins to interfere with sleep (although the Swedish government suggests 25dB in the bedroom). For offices, recommendations of sound consultants range from 30-45dB. Most PCs are on the higher end (or sometimes outside) of this range. With a little bit of effort, an experieneced user can pick out and put together quiet components to build a quiter PC. A neophyte must pay a premium, but may be able to find manufacturers that build quiet PCs.
For the less technically inclined, several companies manufacture quiet PCs. Apple has worked to eliminate noise from many of their desktop models since at least the original Macintosh, and still offers some of the best quietest designs. For those of us who cannot switch platforms, several companies build quiet PCs:
Fujitsu has long been one of the frontrunners in the drive for silent PCs.
I do not know which of the above are current, and this site has a much more comprehensive list of quiet computer makers. Bear in mind that not all machines from these companies are necessarily quiet; only the ones designed to be so.
The first step to building a quiet PC is to buy a quiet hard drive. Most drive manufacturers give detailed specs on their drives, including noise level, either in sound pressure bels or decibels. For the Americans in the crowd, 1 bel=10 decibels. Some manufacturers (e.g. Western Digital) either lie blatantly about their specs or use significantly different test conditions. For instance, my moderately quiet IBM 37GP claims 34dB, while the Fujitsu MPE3173AE claims 35dB. In reality, while the IBM is moderatly quiet, the Fujitsu is much, much quieter than the IBM. One can depend on sites like Storage Review and Tom's Hardware to get the low-down. Fujitsu has, and still does, make the quitest drives in the industry. IBM drives also tend to be pretty good. Quantum and, to a lesser extent, Seagate used to make high-performance, high-noise drives, but both have announced noise-reduction technologies recently, so their current drives are likely quiter.
Low-power (less than 5 watt) drives can be placed in a special enclosure called the SilentDrive that further reduces noise level. Why the power requirement? The enclosure traps heat in. Since every watt of power used changes into a watt of heat, high power drives would burn up in an enclosure like this. Companies give specs on power usage, but again, since this depends on testing methodologies, it's good to also check an independent reviewer's web site for how hot the drive runs.
I use the aforementioned enclosure. Although it's not very well engineered, it does significantly reduce noise, and it works fine in terms of heat with my IBM 37GP drive.
Again, both Fujitsu and IBM do well for power usage. Also, pretty universally, 5400 RPM drives tend to be much quieter and use a good deal less power than 7200 RPM ones. Several people have suggested looking to notebook hard drives; although significantly more expensive, these use very little power and some of them are extremely quiet. The seriously wealthy can buy a completely silent Solid State Drive (though look out for ones with fans). Even a small SSD can allow your normal hard drive to be spun down almost all of the time if you place your system and other commonly used files there.
One thing to look out for is that most hard drives get much louder after a few years (sometimes even a few months) use.
I haven't looked into quiet floppy drives or CD ROM drives much, since I almost never use either. However, at least several CD ROM drive makers have focused on reducing noise:
According to several sources, one of the best CD-ROM drives for noise is the Kenwood UCR421. It uses 7 lasers instead of one, and so can read data much more quickly at a much lower spin speed. It's at once one of the fastest and quietest CD-ROMs available, though it costs several times more than a normal drive. According to StorageReview.com, though, it has some fairly large problems with shaking the case when seeking, leading to some noise (though not when reading linearly, as would be the case when playing music, and even with that, it was the quietest drive in that particular roundup).
There is a bit of a tradeoff between cooling and noise. Computers are conventionally cooled by air. More fans and higher power fans result in better circulation and better cooling, but also in more noise. Likewise, noise can also be reduced by putting acousticly absorbant material around the case; however, this also acts as insulation, and so raises system temperature. Since heat and noise are both just vibrations of different frequencies, it is difficult to find materials that transmit heat but absorb noise.
Cooling, in turn, is directly related to speed. A silicon chip can run faster reliably when it is kept at lower temperatures. When running faster, it also generates more heat, and is therefore harder to cool. Some have approached this problem by using non-air cooling systems (water cooling, large Peltier thermoelectric heat pumps, heat pipes, etc.), but these are either impractical for various reasons, or out of the price range of most end-users.
According to Dr. Bose, fan noise is only necessary if there is air turbulance within the machine. At the air speeds from one of those fans, that turbulance should not be a problem. As such, it is theoretically possible to build an almost completely silent fan, and so several companies do make very quiet fans. Later note: I did some more research later on on heat transfer, and heat sinks are explicitly designed to generate turbulance. Turbulance cools much better than lateral flow. Seems like you can't win.
Although this isn't strictly theory, I got a note in the e-mail about how one guy managed to reduce CPU fan noise by wheel-balancing the fan with a couple of stickers.
CPU makers usually place specs for power usage on their web sites, and a global listing of specs can be found here. Often times, the same CPU will have several variants (called steppings), each with different power usage (a K6/3-400AHX uses 16.1W, while the K6/3-400AFR uses only 10.8A). Many resellers will let you order a specific stepping. Extremists can also try to order mobile versions of processors, which often use considerably less power, and can often tolerate higher temperatures, but cost considerably more, and sometimes require non-standard motherboards.
VIA's Cyrix division shipped the Cyrix III, a low-end CPU explicitly designed to be run fanless (so as to save the system manufacturer the cost of a fan). It's somewhat slow (especially at floating point), but certainly worth considering.
Two companies make quiet power supplies:
The holy grail, of course, is a completely silent, fanless power supply. TK Power makes one, although its not advertised on their web site. Unfortunately, it may not supply enough power for many systems (~150W). That's plenty for an underclocked K6/2 system (mine peaks at considerably less than 65W), but not for an Athlon (where the CPU itself can suck 65W). I ordered one of these, and found that the company sent the unit late, and one that was not ATX . The power supply had an ATX plug, but was not a true ATX power supply. It didn't implement a lot of ATX functionality (it did not output one of the required voltages, and it did not talk to the motherboard about turning on and off, but was permanently on). As such, it wouldn't work with my motherboard. I contacted them again, and after long negotiations, a bit more cash, an extra month and a lot more hassle, they sent me a true ATX fanless. It works great, and the construction is quite nice, but it was a pain to aquire. Contact: Ryan Solri at firstname.lastname@example.org or 510-770-1172 for more information. Note that it's fully ATX electrically, but not in form factor, so you will need to attach it in an alternative fashion to the case (this is fairly trivial to do). They also forgot to send me a power plug for it, but I soldered one up myself. The actual power supply appears solidly built, and cleverly designed from a thermal standpoint. All the components that run warm are thermally coupled to the case, which is made of 1/8 inch metal (probably aluminum). Since it was still a bit warmer than I liked, I stuck a giant heat sink on it, and now it's a pleasant warm (though I'm guessing it would work fine without it).
The German RSG makes a fanless 130 watt AT power supply, though its rather expensive (on the order of $150).
Several people have had success running normal ATX power supplies, with the fan ripped out. I'd strongly caution against this, as if the right parts of the PSU burn out, you may find 120VAC getting pumped into your hard disk, motherboard and CPU. If you really want to risk it, at least stick the fan on a heat-activated switch (All Electronics periodically stocks some for about a buck).
When going fanless, make sure the rest of the system has adequate cooling. Either run open-case, or find some alternative way to prevent hot air from getting trapped inside the machine. You may also want to run a burn-in test to see how hot the chips on your motherboard and cards get, and stick additional heatsinking on the ones that run warm (heatsinks can often be found cheap on surplus electronics web sites).
On my personal power supply, before getting the TK Power, I took of the original fan, and crimped on a quiter fan from Panasonic.
For the more ambitious: I spoke to a Bose engineer who had worked with Apple on designing one of their quiter machines. They found that sticking a muffler on the power supply worked rather well. It reduces noise considerably, but doesn't slow the air passing through much. If you wanna experiment with this, I wrote up a very brief summery of the physics.
I found out that some dude actually did build a muffler box for his computer. His is incredibly well-done. Far better than most anything else I've seen from the amateur computer hacking community.
Finding a good fan is hard. As always, some companies lie about specs. Listening to a fan in the store is also not always practical; I've found is that some fans work great in the store blowing in open air with no resistance. However, they're no longer quite as quiet ``loaded;'' when they try to push air through a heat sink, the noise level increases considerably. Also, several people have noted that many fans will increase in noise as time passes. This is still often your best bet...
For the same amount of air movement, larger, slower fans tend to generate less noise than smaller, faster fans. Intel, when moving from the AT to the ATX spec, recommended a larger fan on the bottom of the power supply instead of a smaller one on the back.
As with a hard drive or any other mechanical component, the fan's vibration can also be magnified if it causes the entire case to vibrate. This depends on how its connected to the case, and how the case is constructed. Putting the fan on a proper suspension can help a lot.
Practically though, one may never need to buy a fan. The CPU fan can be eliminated, and the power supply fan is bought with the power supply. However, if one does need to buy one, a good approach is to either go with a trusted company (such as PC Power and Cooling) or see one in action, either at a store or on an existing quiet computer.
By all accounts, Papst makes some of the world's quietest fans. Panasonic Panaflo fans are also known to be reasonably well engineered.
Several sites recommend covering the inside of the case with sound-absorbing material. While not a bad idea, most of these sites are incompetently made, and ignore issues of heat dissipation, and also often recommend materials like carpets that can introduce static electricity and fry your computer. Use common sense and good engineering.
If you do make a custom case, consider using some sort of metal lining as a Ferraday cage around the case. This can be as simple as aluminum netting (normally designed for window screens) or thick aluminum foil (basically what Dell uses in their plastic cases).
I added this section because there was some confusion as to what makes a good heat sink. Especially in the overclocker community, people often tend to buy heat sinks purely based on size and surface area, as opposed to good engineering and actual cooling ability.
First off, heatsinks designed for fanned operation don't make good fanless heat sinks, and vica versa. The goal of heat sink design is not to simply maximize surface area, but also to get a reasonable amount of air flow, heat transfer through the sink, etc. Still air is a near-perfect insulator. As such, although very closely packed fins will increase area, they will trap air, and the heat sink won't cool very well. Likewise, fins need to be wide enough to carry heat away from the CPU.
Heat sinks designed for forced air flow will have a fairly large level of air flow, and so space the fins very closely. For fanless heat sinks, to have reasonable convection flow, the fin spacing must be much larger. The fins are also generally somewhat wider, since there are fewer of them, and so increasing width hurts area less, but provides a decent improvement in heat flow through the sink. In a good fanless heat sink, expect to see about a quarter of an inch spacing between fins, and on the order of an eigth or a sixteenth of an inch fin width.
If you have a tower case, it is also important that you mount the fanless heat sinks with the fins vertically instead of horizontally. Convection causes hot air to flow up, and horizontal fins would restrict air flow.
There is also a maximum size beyond which enlarging a heat sink helps very little. While this limit is often hit with fanned heat sinks, in fanless operation, this size limit is large enough as to be almost purely theoretical.
One other approach is to simply keep computers in a seperate room from users. For Windows, this has to be done using extension cords for the keyboard, mouse and monitor. Linux users have the advantage of having X, and so only require a network connection between the main machine and a remote, fanless thin client.
Heat pipes are pipes filled with some liquid with a boiling point close to where one wants to keep the CPU temperature. Liquid condenses on a heat sink near the top of the pipe, and then flows onto the CPU hotplate. If the CPU rises above the desired temperature, the liquid near the CPU turns to gas, rises to the heat sink, while new, cool liquid flows onto the hot plate.
Heat pipes are commonly used in notebook computers, where size is critical, but the power usage of a fan is undesirable. as well as some high-end server or workstation applications, where cost is less important. However, they are gradually starting to make an appearance in the PC market as well.
Several people have tried immersing their entire computers in mineral oil instead of air. As a non-conductive liquid, it may cool better than air. On the high-end, NASA engineers immerse electronics in an engineered liquid from 3M called Therminert, though Therminert is prohibitively expensive for normal applications.
Copyright © 2000. Piotr F. Mitros. Questions? Feedback on the site? Feel free to contact me.