Recommended Fans

Fans|Controls | Reference|Recommended
Viewing page 1 of 2 pages. 1 2 Next
  • Sept 26, 2012: Complete rewrite & update of contents and recommendations
  • March 4, 2007: Updated and substantially rewritten to reflect the results of our fan test project
  • August 26, 2004: Added info on comprehensive fan project and added more fans to the list, including 92mm and 120mm models
  • January 5, 2003: Minor changes
  • December 24, 2002: First publication by Mike Chin

Silent PC Review has been interested in fans since its inception, way back in 2002. This is no surprise — fans are the primary noisemakers in most systems, and SPCR is a site about computer noise. However, for the first four years of its existence, the vast majority of SPCR's fan knowledge was buried in personal forum posts and the occasional mention of a particular "reference fan". No formal fan reviews were done until November 2006.

The first baby steps towards reviewing fans occurred in early 2004, when we announced a project called Calling All Good Fans — intended to be a comprehensive examination of the best fans available. A quick glance at the list of updates above shows how long this project took to bear fruit; this recommended list languished for almost three years as SPCR's knowledge of fans stagnated. Things got going again in 2007 and 2008 after another series experiments with fan testing systems. Still unsatisfied with our fan testing methodology, we went back to the drawing board, and came up with the basics for a new system in Fan Test System, SPCR 2010. Finally, couple months ago, another fan roundup was posted: Fan Roundup #6. This was a substantial empirical improvement over the previous tests as it was the first conducted in our custom-built 11 dBA hemi-anechoic chamber with acoustic instrumentation capable of reading accurately down to 9 dBA.


While there are numerous dimensions to fan performance, SPCR's recommendations are based on two:

  1. Noise
  2. Effective cooling


It's well known that the amount of noise a fan makes is based largely on rotation speed, so it might seem that we should just recommend the slowest fans available and leave it at that. However, that doesn't quite cover what we mean when we say a fan has low noise. Thanks to the widespread use of fan speed controllers, a fan's rated speed is almost irrelevant when it comes to choosing a fan — computer fans are very rarely used at their stock speed, especially in custom-built silenced systems. Besides, almost every fan on the market is too noisy at full speed. Reducing the speed of the fans is a mandatory step in silencing a PC.

Our recommendations assume that a fan controller will most often be used to reduce the fan speed. When we evaluate fan noise, we weigh our conclusions most on how a fan sounds when it has been slowed down to an acceptable level. What is an acceptable level? There are two rules of thumb here:

  1. A fan should be slowed down to the point where it provides enough airflow to prevent overheating.
  2. A fan should be slowed down to the point where there is no acoustic benefit to reducing the speed further — i.e. the fan has become inaudible.

Both of these rules are system- and user-specific; it is impossible for us to know what the ideal speed for your system will be.

There are other critical aspects of noise, beyond what we can easily measure. Anyone who has even scratched at the science of sound and noise knows about SPL or sound pressure level, which we routinely measure in decibels, weighted to the A scale, which most closely approximates the frequency and amplitude response of human hearing. This is a basic measure of "loudness" or amplitude. These SPL measurements are conducted with an extremely sensitive, low noise microphone capable of accurate readings down to 9 dBA in our own hemi-anechoic chamber, which has an ambient noise level of 10 dBA. But beyond SPL, there are other sonic factors:

Tonality: This is the subjective effect of a sharp audible peak in amplitude. It sounds like a tone, hence the word tonality, and its opposite is "broadband" or "random" noise. The latter is described as white noise when the frequency balance is flat from low to high frequencies, and called pink noise when the amplitude falls with increasing frequency. Pink noise is more common in nature, and it's the desired "character" for any noise if you want it to be unobtrusive. In general, random noise is far less objectionable than tonal noise.

Regular or irregular variances: This term is a catch-all to cover a range of fan noises that are not constant, but varying. "Ticking" is a pretty common phenomenon. It can be regular and vary proportionately with RPM, or irregular and somewhat random. It can also vary depending on the angle and position in which the fan is used. Depending on its amplitude, and the ambient noise floor of the computer or its environment, this noise can become quite annoying. "Chuffing" is a variation of the above, but the sound is stretched over a longer frame of time. There are other terms describing other noises, but what they all have in common is that they are not desired, they can be annoying, and they are not constant. Absence of all such noises is the ideal; very few fan actually exhibit this.


The second aspect of fan performance is what we describe as "effective cooling". This phrase is used in preference to "airflow", which does not have a linear or even linearly proportionate relationship to cooling. We tried for years to improve the accuracy of our airflow measurements, but in the end, we concluded that a $40,000 fan testing machine is beyond our reach, and that knowing the airflow really doesn't matter anyway in the context of quiet PC cooling.

Cooling in a PC depends on several interrelated factors. The relationship between measured airflow, pressure and temperature change is non-linear even for a single combination of heatsink + CPU and fan. It never scales up and down in a linear way.

So our current fan testing setup measures noise at various speeds, and the cooling effectiveness of the fan with a particular heatsink and heat source at those various speeds. The heatsink used has what we'd consider "medium" airflow impedance — it is neither very difficult nor very easy for a fan to "push" air through it. The word push touches on another technical parameter we're choosing to not measure: Static air pressure. This describes the ability of a fan to "push" air through varying impedance, which is important, for example, for good cooling performance with a heatsink that has big, tightly spaced fins.

While our recommendations are based mainly on how the fan sounds at low speeds, this is not the only thing we examine. A near-complete run-down of our test procedures and tools can be found in Fan Test System, SPCR 2010. Slight changes were noted in Fan Roundup #6.


Aside from noise and effective cooling, reliability is another consideration. Unfortunately, build quality and reliability is impossible to test without long term study with a large sample size. Such testing is well beyond the resources any free hardware review site, so we are limited to making educated guesses about fan quality. There are a number of factors that affect our judgment of quality, some of which are listed below.

  • Bearing type — ball bearings appear to last longer than sleeve and other bearings, but this is not universally agreed.
  • Rated MTBF — how long the manufacturer thinks the fan will last. Not all manufacturers are trustworthy in this respect, and interpreting what MTBF means exactly can be challenging at the best of times...
  • Manufacturer — some manufacturers are known to be generally better — or worse — than others.
  • Sample Variance — high sample variance suggests poor quality control, and thus a higher chance of some kind of failure.
  • General Build Quality — this covers everything from the type of plastic, to the fit and precision of the molding, to the presence of "wobble" while the fan is spinning. Basically, anything that appears to be out of the ordinary is enough to raise our suspicions.

One thing to keep in mind that if you use fans the way we recommend — at reduced speed — they should last much longer than usual. With a decrease from rated RPM, there should be a proportionate increase in run time. Our 10+ years experience in the lab suggests longevity is much less of a concern for quiet PC enthusiasts because we run fans much slower than they are rated for. We've had very few fans over the last 10 years that become too worn out or too noisy to remain in service.

Notes on Sample Variance: Where possible, we try to examine more than one sample of a given fan model, as sample variance is often quite high among fans, especially for "minor" attributes such as noise. Not only is good consistency between samples helpful for ensuring that you don't get a bad sample, it is also an indication of good quality control at the factory.

One uncontrollable variance is that rough handling (like dropping a box of fans several feet) at any point during a fan's journey to SPCR (or the cumulative effects of many instances of rough handling) can cause subtle bearing damage that affects its noise. We always ask for the fans to be packed like thin-shelled eggs, but damage could have occurred before they're shipped to us. This type of damage may be subtle enough that very few users would actually notice and return or report such samples, and yet we might mark those same samples down for poor acoustics. Without huge sample sizes, we really have to shrug and say we do the best we can.


There are many ways of control fan speed.

Modern motherboards all have some form of fan speed control for at least one or two fan headers, particularly the 4-pin CPU header for the PWM fans that are now standard in all stock Intel (and AMD) CPU coolers. These can usually be controlled in the BIOS and/or via a desktop utility from the motherboard manufacturer, and often with SpeedFan. Most often, the fan speed is tethered to CPU temperature sensors, and varied automatically in accordance to preset profiles, and sometimes with user-defined profiles. The number of controllable fan headers (both 4 and 3 pin) and the flexibility of the control system varies tremendously, with the best, most flexible controls usually appearing on the most high end motherboards. To date, the Fan Xpert 2 from ASUS featured on many of their Intel Z77 boards is about the best we've seen. Often, in quiet PCs where relatively few fans are used, the motherboard fan headers are all you need to speed-control the fans.

• For those who want a simple control for a single 3-pin fan, a variable voltage controller like the Zalman Fanmate 2 is ideal. It is a small, inexpensive voltage controller with a tiny knob that sits between the motherboard fan header and the fan, providing a range of 5 to 11 volts. Cost is usually ~$5. These devices can also be built for as little as a dollar or two if you are handy with a soldering iron.

• A rheostat is a simple high power variable resistor that allows fan speed to be controlled much like with a voltage controller. It is often more expensive and less energy (heat) efficient than most variable voltage controllers.

• For more elaborate, multiple fan control, products with names like fan bay, bay bus, and so on, are available. These generally occupy a 5.25" drive bay on the front panel of the PC case and allow control of 2 to a dozen fans, with a variety of features and options, including thermistor control. They range in price from US$10 to $100.

• The 5V and 7V tricks: These date back to hardware hackers before SPCR, but are still used by those who prefer simplicity. Do-it-yourselfers often tap into the voltage lines available from any PSU. Three conveniently available voltages can be obtained from the standard 4-pin Molex connector: 12V (yellow), 5V (red), and 7V (the difference voltage between 12V and 5V). The 7V line is not really that, and it is not recommended for more than one or two fans, especially if they draw much power. For technical reasons we won't cover here, it can potentially damage the PSU. For typical fans, however, it is usually perfectly safe with a good quality PSU.

• Switches can be configured for multiple voltage feed to fans using the basic wiring information shown here. The DIY 12/5V switch is one example. There are many more variations that have been described in the SPCR Forum and all over the web. Your imagination is your main limitation.

• A simple way to get 6V is to wire two identical fans in series to 12V.

• Resistors in series (at least 1W rating) with the fan can also be used. 50~60 ohms usually provides around 5V from a 12V source. Adding ~25 ohms to 5V will give you around 4 volts. The numbers are approximate because the inductance/capacitance of the fan will affect the voltage drop.

• Zener diodes can also be used to reduce voltage — but still allowing the full 12V to pass on startup.



On the Web

1 2 Next

Reference|Recommended - Article Index
Help support this site, buy from one of our affiliate retailers!