A tutorial on building the perfect (quiet!) workstation for your office & lab by new contributor Bartek Plichta, a specialist in acoustics and linguistics.
About Bartek Plichta and how his article came to be featured here, in his own words:
"I work in the fields of hearing science, psychoacoustics, and acoustic phonetics at the University of Minnesota. I publish open-source acoustic analysis software and review research equipment (mostly microphones) for researchers in speech and hearing sciences. [See his web site, http://bartus.org/] I am also an avid PC builder and share your interest (and somewhat of an obsession, really) in building the quietest, but powerful, PC possible.
"I am writing to thank you for the wonderful resource that you and your colleagues have been providing for the enthusiast PC building community. Your website offers a wealth of useful information, and the strict testing methodologies you’ve been using make silentpcreview.com a truly unique resource.
"I am interested in becoming involved in your website, in some capacity. Perhaps I could offer my help in research, reviewing, or writing for the website."
This article chronicles my building of a silent research workstation. While I focus primarily on the needs of acoustics-related research, the workstation should prove to be a powerful tool for just about any research lab or office.
Over the last twenty years I have worked in a number of different research laboratories, ranging from make-shift research "mobiles" at Nene College, Northampton, to a state-of-the-art anechoic chamber at Michigan State University. Despite the obvious differences across these research environments, they all shared one major flaw: they were plagued by computer-generated noise. I am sure most of you have experienced it, too. The constant, buzzing, humming, the relentless drone of computer fans spinning frantically; all required to keep the workstations cool. The issue has become progressively worse over the years, as the increase in processing power inevitably involves an increase in thermal output, which, in turn, requires powerful and noisy air or liquid cooling.
Computer-generated noise is not merely an annoyance. It can seriously interfere with the quality of our research. In this article, I am going to document the process of building a powerful, versatile, and nearly silent computer workstation. My goal is to build a system with a total noise output that is lower than the typical ambient noise floor of a research laboratory. You can be sure that you would not be able to buy a similarly powerful and silent computer, at any price from any of the major PC makers. However, if you follow my tutorials, you should be able to put together a sweet system, at a fraction of the mainstream price.
The term "workstation" refers to a specific type of a computer. On the surface, it may look just like a typical office PC, but it is quite different. The first thing you’ll notice is that workstations cost thousands of dollars. The price is justified both by specially designed hardware components as well as some vendor-specific, value-added software and customer care. The ultimate goal of a research workstation is to provide power, stability, and longevity. In addition, certain components can be used for dedicated processing. For example, the workstations sold by Kay Pentax, include an audio processing board, and an external audio I/O interface box. The workstations running in our lab, for instance, use a Lynx L22 audio A/D and D/A processing PCI cards and NVIDIA Quadro FX graphics cards, for low-latency digital signal processing and stimuli presentation. The computer that we will be building in this series has most of the capabilities of a high-end workstation, such as power, stability, versatility, and extensibility, but, unlike most off-the-shelf workstations, it offers cool and virtually silent operation.
Why should one bother building a workstation if one can easily order one from Dell or Apple? Yes, these manufacturers do build powerful computers, but they do not pay nearly enough attention to noise. Modern designs are driven primarily by the need to provide ample computational power and sufficient cooling. Even though you may come across a workstation that is advertised as quiet, it will, no doubt, turn out not to be quiet enough for your needs. To be fair, there are small, boutique PC makers who might be able to manufacture a silent PC (e.g., Puget Systems), but if you want to have a truly quiet, yet powerful workstation, you have to build it yourself. Besides the obvious utility of a DIY system, the process of designing and building a silent workstation can be enjoyable and rewarding. Did I mention significant savings? Without a doubt, a DIY workstation will be considerably less expensive than most off-the-shelf systems.
Before we set out designing the system, we need to establish some standards for how quiet it needs to be. Measuring noise can be rather tricky. Noise varies not only in intensity (perceived as "loudness") but also in spectral distribution (colloquially referred to as "color"). Yes, we could come up with a very detailed description of acceptable noise levels, but it would be impractical for each builder to implement such guidelines without proper testing equipment and technique. Instead, let’s define "silent," quite simply, as producing noise that is at or below the existing noise floor in your work environment. In other words, we can define a silent workstation as one that does not significantly raise the ambient noise levels that are already present in the research laboratory or office.
If you have the diagnostic equipment available, you can measure the noise levels with the workstation turned on (at idle and at full load) and off, and compare the results. I recommend that you use a spectrum analyzer, rather than a simple sound level meter. It it is important to identify noise levels across the audible spectrum, as some types of noise (e.g., 60 Hz hum) may be more objectionable than others.
If you don’t have a sound level meter, you have no reason to worry. I suggest that you assess the noise levels perceptually. Place the workstation in its designated location, turn it on and log into the operating system. Run your typical applications for at least 30 minutes so that the heat-producing components have a chance to reach their peak thermal output. Listen to the sound of the workstation from about 1 meter away. If you cannot hear it distinctly, i.e., you cannot easily identify its sound against ambient noise, then your workstation is probably quiet enough. Things get more complicated when we have more than one computer (or other noisy equipment) running in the lab, but as long as the workstation is perceptibly silent against ambient noise, it passes the test.
For those of you who are interested in rigorously tested, commercially available computer components, I strongly recommend visiting Silent PC Review, a fantastic web resource for all things related to silent computing.
What makes workstations noisy?
Powerful computers produce a lot of heat. There are three major components whose thermal output requires active cooling: (1) the power supply, (2) the CPU (also known as "the processor", and (3) the GPU (also known as "the video card" or "graphics card"). Other heat-generating components include the motherboard, optical drives, spinning hard drives, dedicated RAID adapters, as well as some of the audio (e.g., Kay Pentax CSL Model 4500) and video (e.g., Matrox Matrox RT.X2) processing add-on cards. Some of these components use dedicated heatsinks and fans, while others rely on proper air circulation inside the computer chassis. Computer cooling fans are used as the primary means of providing active cooling in most workstations. When they spin at speeds in excess of 1,000 RPM or so, they produce noise (mostly due to the resulting air turbulence, perceived as "white noise") and cause vibrations that often resonate through the computer chassis (sometimes perceived as all kinds of rattling and shaking noise). The more fans you have and the faster they spin, the more noise they are going to cause. In addition to fan noise, significant noise can also result from hardware components involving moving parts, such as the rotating data platters and a motor-driven spindle in hard disk drives. Finally, electronic components, such as capacitors, can sometimes become noticeably loud, especially if poorly designed or built.
Figure 1 shows a simple system built around the Lian Li A05NB case. I have labeled the typical components that you will need to build a basic workstation. Can we simply get rid of the noisy components? Having identified the most common sources of computer noise, we can now move on to discussing some of the most effective noise reduction strategies. The obvious question one should ask is whether we can get rid of any of the noisy components. Since our goal is to build a powerful and versatile workstation we cannot really skimp on too many components. However, I believe that we should try to get rid of at least the optical drive. These days, you can install all of the applications from the network and even the operating system can be easily installed from a USB memory stick. Yes, I realize that in some administrator-managed university computing environments software is distributed on optical disks only, but, one would hope, these policies will change and the optical disk will become obsolete soon.
Figure 1. An Intel Core i7 system built inside the Lian Li A05NB chassis. The labels indicate some of the typical components you are likely to use in a workstation.
What can we get rid of next? How about the hard drive? At this time (summer of 2011), we can install the operating system and all the applications on a solid-state drive (they are virtually silent), and use spinning drives only as data storage. For some workstations, data can be stored on a network-attached storage system (NAS), but for most of us, it is not a good idea. We need to run experiments, record audio and video, capture participant responses, etc., all of which require fast and stable data throughput. Therefore, let’s assume that every workstation is going to have to have an SSD and two at least two 2TB hard disk drives, possibly more.
Can we get rid of the GPU? I think that, in some cases, we can get rid of the discrete GPU altogether, but only if the workstation is controlled by a remote desktop connection from another machine and does not require any GPU-specific programming. All modern operating systems allow the control of a workstation desktop via the network. This requires another computer, with a GPU installed, a monitor, keyboard, and mouse connected to it. For example, the "headless" workstation may be located inside a sound-proof booth and used to collect participant responses, while the control PC is in the adjacent lab, behind a sound-proof window. I believe this is a rather rare scenario, but a very real one, nevertheless. Thus,if your workstation involves this particular use case, it can, in theory, be built without an add-on GPU card. In all other cases, we will need to make sure the power supply, the CPU, the GPU, and the hard disk drives are all kept cool and quiet.
The progress in semiconductor manufacturing has enabled Intel, AMD, NVIDIA and others to create physically smaller, more powerful, and more energy efficient chips. Some of the less powerful chips do not require dedicated active cooling. They do generate heat, but the amount of heat is small enough to be effectively dissipated by air circulation within the computer chassis. Is it perhaps possible to build a workstation entirely out of such passively cooled components? Unfortunately not. I believe that a modern research workstation must be built of components that will perform all of the required processing with ease and with a great deal of headroom. We must, therefore, focus on providing efficient and silent cooling.
This brings us to an important point in our discussion. There are two approaches to active cooling, namely liquid cooling and air cooling. Obviously, liquid cooling has more headroom, and is more effective in cooling extremely hot components. However, unless a truly elaborate, expensive, and physically large passive radiator system is used, liquid is going to have to involve several fans spinning at high speeds in order to facilitate efficient heat exchange between the radiators and the surrounding air. Perhaps ironically, liquid cooling simply requires too many noisy fans to be satisfactory for our needs. I argue that air cooling is the best choice for our silent workstation. It is efficient enough, yet reliable, inexpensive and relatively easy to implement.
I have seen several labs where the workstation is covered with sound dampening foam. Perhaps, intuitively, this might like a good idea, but it certainly is not. We should always keep the PC in a relatively well-ventilated area to facilitate efficient heat exchange. However, the chassis itself can play an important role in reducing the PC’s overall noise output. Computer cases are rarely optimized for noise reduction, but there are a handful of models that provide decent acoustic isolation. When choosing a chassis for your silent workstation, you should look for cases that provide some form of sound dampening and anti-vibration hardware mounting options. Of course, the case must also provide adequate air circulation and have the necessary expansion and good cable management capabilities.
For the purposes of this build, I have chosen the Antec P183 mid-tower case, but there are other equally good alternatives available (e.g., Fractal Design Define R3). The case comes with sound-treated side panels (a three-layer design) and anti-vibration hard drive mounts. Also, it uses a proven ventilation system with two intake in the front and two exhaust fans (one in the back and one in the top). The case is divided into two different thermal chambers to further facilitate airflow. It is important to point out that good air circulation is crucial to quiet operation because it allows fans to run at relatively slow rotational speeds, and, as a result, more quietly.
Obviously, you don’t have to use the Antec P183 if it doesn’t work for your needs. What should you look out for when buying a chassis for a silent build? Here are some points to consider.
The case must be large enough to accommodate all the components with ease, giving you plenty of room to work with.
The case must have good ventilation options. I recommend using the proven design of front intake and rear exhaust fans. The top fan is optional. You should use fans of 120 mm in diameter. Larger fans, especially 140 mm, can also be used, but there are far fewer silent options available in sizes other than 120 mm.
You must be able to do proper cable management to facilitate good air circulation. Look for cut-outs in the motherboard tray for routing cables and for enough room (say, at least 15 mm) between the back of the motherboard tray and the right side panel for tucking your cables away. Figures 2 and 3 show the cable management options of the small Lian Li PC-A05N case. The power supply and SATA cables are secured with adhesive cable anchors and zip ties. All cables are tucked away and do not obstruct airflow inside the chassis.
Figure 2. Cable management of Lian Li A05NB chassis (back)
To further facilitate cable management, be sure to use a modular power supply, such as the Antec BP550 Plus 550W. A modular power supply allows you to connect only the cables you need, thus substantially reducing cable clutter. More details about power supplies are included further on.
Figure 3. Cable management of Lian Li A05NB chassis (front)
It is crucial for the case to use rubber grommets for mounting vibrating components, such as fans and hard drives. Figure 4 shows rubber fan pins that are very effective at preventing fan vibrations. You can also use rubber dampeners for fans and power supplies (Figures 5). They are installed between the vibrating component and the chassis to further decrease noise levels.
Figure 4. Rubber fan pins. They are easy to use and effective in preventing vibrations.
You need some sort of acoustic treatment (sound dampening) to prevent the fan noise from leaking outside the chassis. Some cases come with pre-installed acoustic foam or bitumen but you can easily apply acoustic treatment yourself, provided the case have enough room for it.
The case can be made either of aluminum or steel, with some plastic parts usually used for the front bezel. The construction must be sturdy and made with tight tolerances in order to prevent unnecessary vibrations. There’s a debate on whether aluminum cases provide better cooling, but we don’t need to concern ourselves with that. Be prepared to be pay a little extra for a sturdy case. Try solid brands, such as Antec, Fractal Design, Silverstone, or Lian Li, and stay away from the entry-level cases. You should expect to spend around $100-150 (US) for a decent case.
It is important for the case to have sturdy rubber feet (avoid the flimsy adhesive ones) to keep the body off the floor and prevent vibrations.
Proper placement is crucial to proper air ventilation. I realize that labs and offices have limited space to spare, but try to place the system in a well ventilated area, away from the wall or furniture. Make sure the spot is clean and free from dust. If you are going to put the case in a closet or locker, be sure to install ventilation fans.
The 60-Hertz hum is a proper curse of acoustics laboratories. It is quite difficult to avoid it, especially in old buildings, with old power lines. I have written extensively on preventing hum in sound recording, and the same ideas apply here. Be sure to use a high-quality uninterruptible power supply (UPS) and a surge protector. Use properly grounded cables and outlets. Finally, try the Ebtech Hum-X for filtering out unwanted voltage on the ground line. You can read more about it in my review.
Proper cooling is paramount for silent operation. Fans, at least in our case, are going to be operating within a certain RPM range; their speed depending on system temperature. If your system runs hot, the fans are going to have to spin faster, thus causing significantly more noise. Let’s now discuss some of the low-noise cooling methods.
The three main hot components, i.e., the power supply, the CPU, and the GPU will need their own, dedicated cooling. However, we also need to provide proper air circulation inside the chassis itself. Minimally, you need one intake and one exhaust chassis fan (see Figure 1). The idea is that cool air is sucked in from outside the case and moved over the hot components to be then sucked out of the case by the rear (or top) exhaust fan (Figure 1).
Figure 1. Typical airflow inside a computer chassis. The blue arrow indicates cool air, and the red one warm air.
The case cooling system can get very complex and sophisticated, with multiple fans of different diameter, operating at different rotational speeds. I have seen Dell, Apple, and HP workstations with additional air ducts inside the chassis, made by means of removable plastic chambers. The utility of such separate chambers remains somewhat debatable, but the general idea of trying to keep heat-generating components cool and maintaining proper air circulation inside the case applies across the vast majority of modern workstations. We must remember, though, that, Inevitably, the faster the fans spin, the louder they get.
Computer fans are typically described in terms of their diameter, rotational speed, their air moving capacity (measured in cubic feet per meter, or CFM), the noise level (measured in dB), and the connector type. For example, the Thermalright TY-140 fan that we’re going to use for the CPU and GPU is 140 mm in diameter, offers speeds of 900-1300 RPM, has CFM of 28.3~74, and the noise level of 17~21 dBA. The fan uses pulse width modulation (PWM) technology to adjust its speed depending on the temperature in its immediate environment.
There are many different types of chassis fans available so, and, unfortunately, the specifications provided by manufacturers cannot always be trusted. There exist no standard noise testing methodologies, so it is best to rely on independent reviewers, such as Silent PC Review, for a more standardized evaluation. I have had very good experience with Nexus, Scythe, and Gelid 120mm, PWM fans.
Some of the modern computer motherboards (though not as many as one would like) support the PWM technology and provide a BIOS-level monitoring and control of fan speeds. When selecting a motherboard, I typically pay close attention to whether or not it supports at least three PWM fan headers. Other methods of fan speed control involve the change in voltage (e.g., “undervolting” to 5 or 7 V, instead of the full 12 V), but I think PWM is the most straightforward and effective approach.
The BIOS-level PWM technology enables motherboard manufacturers (e.g., Intel) to write temperature management software for Windows. Such software provides an additional level of temperature monitoring and fan speed control, though it is not required to keep your system running cool and quiet. Finally, a dedicated hardware-based fan controller can be used, especially if one requires fully manual fan speed control. Fan controllers can be typically installed inside the 3.5” or 5.25” drive bay.
Theoretically, a medium-size, ATX chassis, such as the Antec P183 case I have chosen for the present build, should be adequately cooled with one or two intake fans and one exhaust fan, all controlled by the BIOS-level software. However, in order to facilitate good airflow, it is essential to keep the interior neat, tidy, and dust-free. Proper cable management will help, as well. Figure 2 illustrates airflow inside the Antec P183 V3 case. Note two thermal chambers; the top for the motherboard, CPU, and GPU, and the bottom for the power supply and hard drives.
Figure 2. Airflow inside the dual-chambered Antec P183 V3 workstation case
We have now reached a crucial stage in our workstation build, namely, deciding on how to cool the three hottest components: the CPU, GPU, and PSU. It is important to choose an option that performs extremely well at full load. Workstation noise is rarely constant; it can significantly increase along with the increase in computational load. The harder the chips work, the hotter they get, and the faster the fans are required to spin. The cooling solution we choose must be able to reduce this increase in noise at full load as much as possible. In other words, the cooler must have adequate headroom, both thermally and acoustically.
If you go to a computer showroom, such as the Apple Store, you may be tempted to believe that the Mac Pro you are listening to is quiet. It might as well appear to be, but it is likely operating at a very light load and its acoustic output is being drowned out by the high level of ambient noise in the store environment. In order to truly assess the acoustics of a workstation, you’d have to buy it, install it in your lab, and run a specially designed benchmark application that stresses the CPU, GPU, and PSU at levels comparable to those you are likely to operate on in your daily research computing. I predict that you would discover that the workstation you’d just purchased gets unacceptably loud at full load. It has happened to me every single time I worked with an off-the-shelf workstation. At some point I decided that it was going to be impossible to buy a quiet workstation and that I would have to design and build one myself. I encourage you to do the same.
The stock heatsink and fan that comes with a CPU is likely to be unacceptably noisy. We have to replace it with an aftermarket heatsink and fan. The basic thermodynamics of CPU cooling are that the larger the area of the heat exchanger, the more efficiently it is going to perform. CPU heatsinks are typically made of copper or aluminium, with numerous heat pipes and fins to maximize heat exchange. They can be rather large and are likely to require additional mounting hardware. In addition to efficient cooling, large heatsinks provide an extra layer of redundancy, a crucial feature of a research workstation. Should the cooling fan fail, a large heatsink is able to provide enough passive cooling capacity to prevent the CPU (or GPU) from overheating and failing.
Figure 1. Thermalright HR-02 heatsink. Note the asymmetrical design
For the purposes of this build, I am going to use the Thermalright HR-02 heatsink (Figure 1) and a Scythe 120mm PWM fan. There are several equally good heatsinks available, but the HR-02 has the advantage of an asymmetrical design, making it more likely to fit a motherboard, with all RAM slots filled, even with RAM heatsinks installed (Figure 2). I am going to attach a 120mm Scythe fan to the heatsink with the provided pins. We want the fan to push cool air through the fins of the heatsink towards the rear case exhaust. Thus, we’re going to create a sort of "push-pull" airflow configuration, which has been shown to work particularly well for heat exchange purposes.
Figure 2. Thermalright HR-02 heatsink and Scythe PWM fan. Note the "push-pull" airflow configuration and adequate clearance of the RAM modules.
In order to cool the video card, we are going to have to get rid of the stock heatsink and fan. There are a few after market solutions available. I have decided to use the Thermalright Shaman heatsink and fan for this particular build (Figure 3). The rationale is exactly the same as in the case of the CPU; we want to put a large heatsink on the chip, so it keeps the video card cool, even with a fan operating at low rotational speeds (say, up to 1,000 RPM). The Shaman is an excellent cooler, which, along with the TY-140 PWM fan, provides massive, whisper-quiet cooling to the GPU, even at full load. I must point out that both the CPU and GPU fans are pulse-width modulated and regulated by the BIOS. It is a very simple, elegant, and effective solution.
Figure 3. Thermalright Shaman VGA cooler installed onto the NVIDIA GTX 460 reference card. Note that the entire kit requires 4 PCI slots worth of space.
The power supply is typically cooled with a built-in fan. It is crucial to choose a relatively high-wattage and an 80-plus efficient power supply. The high wattage is going to provide ample, clean power to all of the components with a great deal of headroom. Such a power supply is extremely unlikely to overheat and will thus operate at a relatively low temperature. Consequently, the cooling fan is not going to have to spin very fast, and the power supply will remain quiet, even at full load. For the purposes of this build, I have chosen the Antec CP-850 power supply. It is larger than the ATX-standard PSUs, so it will fit only a handful of Antec cases. However, the benefits of high wattage, cool operation, and a low price are well worth the obvious compatibility compromise. Figure 4 shows the non-linear distribution of the fan’s rotational speed, while Figure 5 illustrates PSU efficiency.
Figure 4. Duty vs. RPM curve of the Antec CP-850 power supply (courtesy of Antec)
The Antec P183 V3 chassis has a unique design whereby the PSU is installed in a separate chamber. It, therefore, receives additional cooling from a 120mm intake fan (Figure 6). Other cases may have a dedicated PSU fan blow hole in the bottom. You would then position the PSU with the fan facing down to allow it to intake cool air from outside the chassis. Either design should work well.
Figure 6. Antec P183 V3 chassis airflow in the bottom chamber allowing the PSU to be properly ventilated
We have so far designed cooling solutions for the chassis and for the three hottest components. It is time to put it all together and run a few benchmarks to see if we’ve achieved our goal. Building a silent workstation is not very different from building a regular PC. The only thing to keep in mind is overall tidiness, attention to detail. and cable management. As with any PC build, you need to follow the manufactures’ guidelines and recommendations. Also, join a reputable PC-building forum, such as Silent PC Review, and ask questions. The biggest issue is component compatibility, both in terms of hardware and software. Make sure the operating system you choose supports all of the hardware you plan to use. Workstation building is a craft that requires careful planning, patience, and a little bit of creativity. You can be sure that all of the components chosen for this build are compatible with one another and work perfectly with Windows 7 64-bit as well as with Linux Ubuntu 11.04. Here’s the hardware inventory:
I benchmarked the system with Prime95 to stress-test the system and measure temperature increase from idle to full load. At idle, the CPU and GPU hover around 30C, while at full load (30-min of 100% load testing) the CPU reaches around 65C while the GPU peaks at around 60C. Of course, these temperatures will differ slightly depending on the ambient temperature in your work environment as well as your fan speeds. The most important result is, of course, the total noise output, both at idle and at full load. The workstation performs as expected. At one meter away, it is inaudible in my office and lab environment. This simple result has taken a lot of research and planning, but it has been a thoroughly satisfying experience. I hope you give it a try yourself. Know that it is currently the only way to make sure your workstation is powerful, stable, cool, and, most of all, virtually silent.
Figure 1 shows the finished system inside the Antec P183 V3 chassis. Figure 2 is a screen shot of CPUID Hardware Monitor software. The fans, controlled by the Intel BIOS, run at low speeds, thus keeping the system quiet. Finally, you can further improve cable management and the overall looks of your build by individually sleeving the PSU cables or making sleeved extensions (Figure 3). Yes, it is perhaps superfluous, but it can be utterly fun. Good luck with your build and be sure to email me with questions.
Figure 1. The finished system
— End —
Much thanks to Bartek Plichta for this highly valuable contribution to SPCR!
If you have any interest in acoustics including excellent reviews of microphones and related audio gear or linguistics, check out Bartek’s unique site, http://bartus.org/akustyk/
* * *
* * *
Quantifying noise performance
Quantifying noise performance is tricky. It is virtually impossible to come up with a fully satisfying approach for the purposes of this article. Remember, our ultimate goal was to build a workstation that will not be audible from the distance of 1 meter, in a typical, quiet research environment. Of course, ambient noise levels are going to differ dramatically from lab to lab, so if your laboratory happens to house, say, huge power generators, even a relatively loud PC is going to appear whisper quiet. One might, therefore, ask, why not just measure noise levels and report them "objectively?" Well, it is not that simple. Yes, in theory noise measurement can be more "objective" but there are many, potentially confounding variables at play. The fields of audiology, hearing science, psychoacoustics grapple with this problem on a daily basis, and researchers still do not fully agree on how to measure noise "objectively."
Part of the problem is ambient noise levels (also referred to as "the noise floor") in the testing environment. For the most accurate noise measurements, the testing environment must have an extremely low level of ambient noise and be free from electromagnetic interference. Anechoic chambers are designed to obtain precisely this type of environment, but even the most expensive and technologically advanced anechoic designs vary greatly in their "noise floor." For example, the anechoic chamber at Michigan State University has the noise floor of about 15 dBA. But, what exactly does that mean? Is this an absolute value, such as, say, somebody’s height? Not quite. Decibel scales differ from one another, partly, because they use different reference levels and weighting filters. And that gets complicated, as well.
Say we want to use a scale that best corresponds to human hearing. After all, we want the workstation to be virtually silent, as perceived by our own ears. Human hearing is greatly variable and non-linear, i.e., it performs differently for different types of sounds. I think you can already see the problem there. We would not be able to properly characterize noise levels in reference to human hearing by using just one number. We would need to characterize the entire audible spectrum, between 20 and 20,000 Hz. The next question is whether we should apply an octave-band, a third-octave-band, or a critical-band filter? On a linear or logarithmic scale? Then we get into periodic versus transient sounds, bandwidths, overtones, resonant frequencies… the list goes on.
It is equally difficult to compare our results to those published by workstation manufacturers. We would need to know all the details of their testing procedures, instrumentation, methodologies, etc. It is simply not possible, so we have no choice, but to take their numbers in good faith, or rely on independent reviewers such as Silent PC Review.
Given all the complications related to quantifying noise performance, I decided to make digital recordings of my quiet research environment (a sound-treated, audiology lab) with the workstation off and on (at idle and full load). To give you a better idea of what the levels are, figure 4 shows octave band maximum permissible ambient noise levels for each test frequency range and ANSI S3.1-1977 levels for comparison.
Figure 4. Octave band maximum permissible ambient noise levels for each test frequency range and ANSI S3.1-1977 levels for comparison
I decided to use the same recording technique as that I use in my microphone reviews. The idea is to give you a reasonable comparison of the PC noise levels with typical conversational speech. This should give you a decent indication of how much above a very quiet lab ambient noise level the workstation-generated noise appears to be. Loudness is in the ears of the beholder, so you be the judge! Here’s how I suggest you listen to the recording:
- Open the MP3 file in your playback application (e.g., Windows Media Player, iTunes, etc.).
- Listen to the speech sample and set the volume at a level that, subjectively, would similar to that of someone talking to you from about a meter away.
- Listen to the entire sound file, which starts with 0.5 s of pure tone (377 Hz), followed by speach, then pure tone, ambient noise, pure ton, PC noise at idle, pure tone, and PC noise at full load (see Figure 5).
- Try to hear any differences among the three noise samples.
Download MP3 files, encoded with the LAME MP3 encoder at 128 kbps (mono). The maximum CPU and GPU temperature peaks at around 60C, with ambient temperature of around 25C. The recordings were made with the DPA 4006-TL microphone (reviewed here) at the distance of 1 meter from the workstation, an ultra quiet Tucker-Davis microphone pre-amplifier, and the Sound Devices USBPre A/D converter (reviewed here) at 48,000 Hz, 24-bit.
Figure 5. Waveform of the test recording
I have plotted long-term average spectra (LTAS) of each type of noise (200 Hz bandwidth). Figure 6 shows LTAS of ambient noise, while Figures 7 and 8 show LTAS of the PC noise at idle and full load, respectively. Note the clear change both in noise level, as well as character, with the most pronounced increased below 500 Hz, possibly caused by the low-frequency "hum" of the fans. At full load, an additional peak shows up in higher frequencies, possibly the result of the PSU fan spinning faster due to a significant increase in both power and temperature during the test. Still, the Antec CP-850 PSU remains relatively (and surprisingly) quiet throughout.
You will recall that the fans operate at rather low speeds (Figure 2 above), thus the increase in noise is rather small and barely perceptible (at least to my ears) at the distance of 1 meter from the workstation.
Figure 6. LTAS of ambient noise