2. Choosing and Buying Components

The components you choose for your system determine its features, performance level, and reliability. How and where you buy those components determines how much the system costs.

Sometimes it is a good idea to spend more for additional features or performance, but often it is not. The trick is to figure out where to draw the line—when to spend extra money for extra features and performance, and when to settle for a less expensive component. Our years of experience have taught us several lessons in that regard:

• Benchmarks lie. Buying PC components based solely on benchmark results is like buying a car based solely on its top speed. It’s worse, actually, because no standards exist for how benchmarks measure performance, or what aspect of performance they measure. Using one benchmark, Component A may be the clear winner, with Component B lagging far behind. With another benchmark, the positions may be reversed. When you select components for your new system, we suggest you regard benchmarks with suspicion and use them only as very general guidelines, if at all.
• Performance differences don’t matter if it takes a benchmark to show them. Enthusiast websites wax poetic about a processor that’s 10% faster than its competitor or a video card that renders frames 5% faster than its predecessor. Who cares? A difference you won’t notice isn’t worth paying for.

  Ron Morse Comments

  Conventional wisdom is that a change in performance has to be on the order of 25% to 30% before it becomes immediately obvious. This doesn’t mean that small improvements aren’t important, but it does mean you probably won’t notice them until you realize that a job that used to take many minutes doesn’t take quite as many minutes as it used to.

• It’s easy to overlook the really important things and focus on trivialities. The emphasis on size and speed means more important issues are often ignored or at best given short shrift. For example, if you compare two hard drives, you might think the faster drive is the better choice. But the faster drive may also run noticeably hotter and be much louder and less reliable. In that situation, the slower drive is probably the better choice.
• Integrated (or embedded) components are often preferable to standalone components. Many motherboards include integrated features such as video, audio, and LAN adapters. The integrated video on modern motherboards suffices for most purposes. Only hardcore gamers and others with special video requirements need to buy a separate video adapter. The best integrated audio—such as that on motherboards that use Intel and NVIDIA chipsets—is good enough for almost anyone. Integrated LAN adapters are more than good enough for nearly any desktop system.

The advantages of integrated components are threefold: cost, reliability, and compatibility. A motherboard with integrated components costs little or no more than a motherboard without such components, which can save you $100 or more by eliminating the cost of inexpensive standalone equivalents. Because they are built into the motherboard, integrated components are usually more reliable than standalone components. Finally, because the motherboard maker has complete control over the hardware and drivers, integrated components usually cause fewer compatibility issues and device conflicts.

• Buying at the “sweet spot” is almost always the best decision. The sweet spot is the level at which the price/performance ratio is minimized—that is, where you get the most bang for your buck. For example, Intel sells a broad range of desktop processors, from $50 dual-core Celerons to $1,000 six-core Core i7 processors. Dual-core Celeron processors are cheap, but slow. Six-core Core i7 processors are fast, but hideously expensive. There must be a happy medium. The sweet spot for processors is around $175 (give or take $25) for a retail-boxed CPU. If you spend much less, you get less performance per dollar spent. If you spend much more, you get only a slight performance increase. This sweet spot has stayed the same for years. The same $175 buys you a faster processor every time Intel cuts prices, but that $175 processor has always been the bang-for-the-buck leader.

  Bang for the Buck

  To find the sweet spot, just compare the price of a component to its performance or capacity. For example, if one processor costs $175 and the next model up is 10% faster, it should cost at most 10% more. If it costs more than that, you’ve reached the wrong part of the price/performance curve, and you’ll be paying a premium for little additional performance. Similarly, before you buy a hard drive, divide the price by the capacity. At the low end, you may find that a small hard drive costs more per gigabyte than a larger drive. At the high end, a very large drive probably costs significantly more per gigabyte than a medium-capacity model. The sweet spot is in the middle, where the cost per gigabyte is lowest. Make sure, though, that you compare apples to apples. Don’t compare a quad-core or six-core processor to a dual-core processor, for example, or a 7,200 RPM hard drive to an SSD model.

  Of course, all of that assumes that you care about bang for the buck. If you have the budget and performance is more important to you than cost, by all means buy the $1,000 processor. You’ll have the fastest system possible, which for a lot of people is no small thing.

• It’s almost always worth paying more for better quality and reliability. If the specs for two components look very similar but one sells for less than the other, it’s a safe bet that someone cut corners to reduce the price of the cheaper component. The cheaper component may use inferior materials, have shoddy build quality or poor quality control, or the manufacturer may provide terrible tech support or a very short warranty. If it’s cheaper, there’s a reason for it. Count on it. The best way to avoid the trap of poor-quality components is to be willing to pay a bit more for quality. The price difference between a mediocre product and a top-quality one can be surprisingly small. Throughout this book, we recommend only high-quality products. That’s not to say that products we don’t list are bad, but those we do recommend are good.
• Brand names really do mean something, but not all brands are good ones. Brand names imply certain performance and quality characteristics, and most manufacturers take pains to establish and maintain those links in consumers’ minds. Different brand names are often associated with different quality and/or performance levels in a good/better/best hierarchy, in the same way that General Motors sells its inexpensive models as Chevrolets and its expensive models as Cadillacs.

For example, ViewSonic makes several lines of LCD displays, including its high-end Pro Series, its midrange Graphics Series, and its entry-level Value Series. ViewSonic also maintains a separate brand name for its cheapest products: OptiQuest. If you buy a Pro Series monitor, you know it’s going to cost more than the lower-end models, but you also know it’s going to have excellent performance and will likely be quite reliable. Conversely, if you buy an OptiQuest monitor, you know it’s going to be cheap and not very good. A few manufacturers also have special high-end brand names, although that practice has declined as margins have eroded throughout the industry.

• If you’re on a tight budget, shop by brand name rather than by specifications. For the same price, it’s usually better to choose a component that has less impressive specifications but a better brand name rather than a component with better specifications but a poor brand name. For example, if you can’t afford a high-end 24” Samsung LCD display, but a 22” Samsung model with similar specifications or a Brand-X 24” display with similar specifications is within your budget, choose the 22” Samsung model. It may be a bit smaller than the Brand-X display, but the Samsung will almost certainly have better display quality and be more reliable. In other words, if you have to choose between better quality and higher performance, choose quality every time.
• Similarly, if you’re choosing between two lines offered by a particular manufacturer, choose the better-quality model. For example, if a display manufacturer offers a 24” model from its entry-level line at about the same price as a 22” model from its premium line, choose the latter. You give up a bit of screen size, but display quality will be higher, and possibly reliability as well. Conversely, if the manufacturer offers good/better/best lines, you’ll probably get optimum bang for the buck by choosing the midsize “better” model rather than the small “best” model or the large “good” model.

In this chapter, we tell you what we’ve learned based on more than 20 years of experience buying PC hardware components. By necessity, the project systems in this book are built with specific components. The obvious problem with any such list of specific components is that new products are constantly introduced and older products discontinued. A product that is leading-edge when we proof the final galleys may be midrange by the time the book arrives in bookstores and discontinued by the time you read it. Nor is this problem limited to printed books; even enthusiast websites can’t keep up with the flood of new products. So, rather than a detailed discussion of such ephemeral details, this chapter focuses heavily on the important characteristics of hardware components—the things you need to understand to make good decisions.

Fortunately, progress in PC components is generally evolutionary rather than revolutionary, and the quality level you can expect from any particular manufacturer is pretty predictable. If we use, for example, a Seagate Barracuda 7200.12 hard drive in one of the project systems, and that exact drive is no longer available when you start ordering parts for your own system, it’s probably safe to assume that a similar 7200.12 model will do the job just as well. And if Seagate has by then introduced a 7200.13 series, it’s very likely that those new models will be at least as fast and reliable as the earlier models.

With so many alternatives, it’s easy to buy the right part from the wrong source. Accordingly, the last part of this chapter distills what we’ve learned about how and where to buy PC hardware components. When you finish reading this chapter, you’ll have all the information you need to make the right buying decisions.

Choosing Components

The biggest advantage of building your own PC is that you can choose which components to use. If you buy a cookie-cutter system from Dell or HP, most of the decisions are made for you. You can specify a larger hard drive, more memory, or a different monitor, but the range of options is quite limited. Want a better power supply, a quieter CPU cooler, or a motherboard with built-in FireWire and enhanced RAID support? Tough luck. Those options aren’t on the table.

When you build from scratch, you get to choose every component that goes into your system. You can spend a bit more here and a bit less there to get exactly the features and functions you want at the best price. It’s therefore worth devoting some time and effort to component selection, but there are so many competing products available that it’s difficult to separate the marketing hype from reality.

The reality is that you get what you pay for—or perhaps we should say, you get no more than you pay for. If you’re not careful, you can get much less than you pay for, whether by choosing the wrong components, or the wrong vendor, or both. To avoid that problem, you need to educate yourself about what’s important and what’s not when it comes to choosing components.

Still, we understand that many people will want specifics, so we’ll do our best to provide them by naming names and offering advice. We don’t doubt that some people will take issue with some of our recommendations and advice. We don’t claim that the products we recommend are “the best” in any absolute sense, because we haven’t tested every product on the market and because what’s “best” is inherently subjective. What’s “best” for us may be just “very good” from your point of view; however, it almost certainly won’t be “terrible.”

So, keeping all of that in mind, the following sections describe the manufacturers we recommend for the various components you’ll need to build your own PC.

Case

The case (or chassis) is the foundation of any system. Its obvious purpose is to support the power supply, motherboard, drives, and other components. Its less-obvious purposes are to contain the radio-frequency (RF) interference produced by internal components, to ensure proper system cooling, and to subdue the noise produced by the power supply, drives, fans, and other components with moving parts.

A good case performs all of these tasks well and is a joy to work with. It is strongly built and rigid. Adding or removing components is quick and easy. All the holes line up. There are no sharp edges or burrs. A bad case is painful to work with, sometimes literally. It may have numerous exposed razor-sharp edges and burrs that cut you even if you’re careful. It is cheaply constructed of flimsy material that flexes excessively. Tolerances are very loose, sometimes so much so that you have to bend sheet metal to get a component to fit, if that is even possible. Using a cheap case is a sure way to make your system-building experience miserable.

Depending on the case size and type, expect to pay at least $50 for a good basic case without a power supply, and $100 or more for that case with a good power supply. Case and power supply quality is one of the major areas for cost-cutting in mass-market systems, many of which use a case and power supply combo that would retail for half the price of decent components, or less. If you must economize on these components, we strongly recommend that you buy an inexpensive case without a power supply and install a good power supply. Using a cheap power supply almost guarantees problems—if not initially, then certainly as the system ages.

Here are some important considerations when choosing a case:

Build quality and price

Build quality correlates closely to price. Better and more expensive cases are heavier, stiffer, and have no dangerous sharp edges or burrs. Panels are heavy-gauge and fit to tight tolerances, often with captive screws, and supplemental case fans are of good quality. One exception to the heavier-is-better rule is portable gaming cases, which are often made from aluminum or a light alloy but are engineered to provide stiffness and tight tolerances.

Size and type

Mainstream cases are generally categorized as server, full-tower, mid-tower, mini-tower, microtower, desktop, or slimline. There are also various special-purpose cases available that are optimized for specific purposes, such as media center/home theater cases and gaming cases. For maximum flexibility and expandability as well as better cooling, choose a large case. If size is a major consideration, choose the smallest case that will accept the motherboard that you want to use and has sufficient drive bays for all of the drives you want to install.

Motherboard form factor compatibility

Motherboards are available in numerous form factors, which differ in size and (sometimes) mounting hole positions. Fortunately, only three of those form factors are commonly used for desktop motherboards. From largest to smallest, these include: ATX (305 mm × 244 mm board size), microATX or μATX (244 mm × 244 mm) and Mini-ITX (170 mm × 170 mm).

With very few exceptions, cases built to accept these form factors are backward-compatible. An ATX case can accept an ATX, μATX, or Mini-ITX motherboard. A μATX case can accept a μATX or Mini-ITX motherboard, and a Mini-ITX case can accept only a Mini-ITX motherboard.

A few motherboards designed for gaming systems or servers use the Extended ATX (EATX) form factor, which has a board size of 305 mm × 330 mm. These motherboards fit only EATX cases, which can also accept motherboards in any of the three most common form factors.

Bundled power supply

Many mainstream cases include a power supply, which may or may not be suitable for your purposes. Case quality and power supply quality generally go hand-in-hand. We’ve never seen a cheap case that included anything other than a junk power supply, and we’ve seldom seen a high-quality case that included anything less than a midrange power supply.

Expensive and special-purpose cases—particularly server and gaming cases—often do not include a power supply, on the assumption that you’ll want to choose a power supply to fit your particular needs. Media center/home theater cases may or may not include a power supply. Unless you just happen to love a case that comes only with a bundled power supply, we recommend that you buy the case and power supply separately. You may pay a few bucks more than you would for a case with a bundled power supply, but you’ll also probably get a better power supply.

Be wary of very small cases and bare-bones systems like those sold by Shuttle, Foxconn, MSI, and others that come with a bundled power supply. Some of these cases use power supplies with proprietary physical form factors, which means the only source for a replacement power supply is the original vendor. These proprietary power supplies may be extremely expensive or no longer available when your original power supply needs to be replaced. If you buy such a case, either make sure that it accepts industry-standard replacement power supplies or resign yourself to paying through the nose to replace the power supply when it inevitably fails.

Front-panel connectors

Most cases provide front-panel connectors for USB and audio. Some also provide connectors for FireWire and/or eSATA, which are used to connect external hard drives, camcorders, and similar devices. A few cases include a built-in front-panel card reader, which is useful for transferring files from a digital camera or camcorder. If a case you otherwise like lacks one or more front-panel connectors that you want, you can purchase a port extender that fits in an open external drive bay and provides the connectors you need.

Case material

Nearly all cases use some plastic, particularly for the front panel, but the frame should be of stiff, heavy-gauge steel or aluminum, as should the side and top panels. Steel is rigid but heavy. Aluminum frames can be as rigid as steel, but are much lighter. In fact, the first time UPS showed up with an aluminum case without a power supply installed, we picked up the shipping box and thought they’d forgotten to put the case in the box. All other things being equal, an aluminum case costs more than a similar steel case but provides similar rigidity. Some sources claim that aluminum cases provide superior cooling, but we’ve never been able to confirm these claims. We suggest that you choose a steel case unless system weight is a major concern, such as for a LAN party gaming system or when building a stackable media center/home theater system.

Cooling

Inexpensive cases often make no special provisions for better cooling, and may even lack mounting positions for supplemental cooling fans. What fans and mounting positions they do include are often small—60 mm to 80 mm. (Of course, small cases are limited in both the number and size of fan positions.) Better cases have multiple fan-mounting positions, usually of larger size—typically 90 mm up to 250 mm (!), depending on the size of the case—and may have fans already installed in some or all positions. Midrange cases usually include inexpensive fans, which are not the quietest or most durable models available. Premium cases are often supplied without fans, on the theory that you’ll want to choose your own fans, but those that have fans installed usually include midrange or better fans. Some high-end gaming cases include provisions for installing liquid cooling, and a few of those have a radiator and tubing preinstalled.

Because larger fans move more air at slower (and quieter) rotation speeds, we recommend that you choose a case with as many and as large fan mounting positions as possible. If you’ll be running hot components, such as a gaming video card (or two) or an extreme processor, look for a case that includes ductwork designed to exhaust hot air from those components directly to the outside of the case.

Quiet PC technology

Many premium cases include various quiet PC technologies, such as sound-deadening panels, vibration isolators in the drive bays, thermally controlled case fans, and so on. Although some of these features can be added via third-party products, if noise level is a major consideration we suggest that you purchase a case that’s designed from the ground up to minimize noise.

Recommended case brands

There are many good case brands, but Antec (http://www.antec.com) stands out from the crowd. Antec offers a huge range of cases, from inexpensive basic cases to premium special-purpose cases, with or without bundled power supplies. The bundled power supplies, most of which are actually made by Seasonic, are generally excellent. Over the last decade, we’ve used Antec cases for probably 95% of the systems we’ve built, and we’ve never been disappointed.

Cooler Master (http://www.coolermaster.com) is Antec’s major competitor in the PC case segment. It’s a well-respected brand that’s particularly popular amongst gamers. Cooler Master offers a broad selection of case types and styles, although not quite as broad as Antec. The Cooler Master cases we’ve seen are similar to Antec cases in quality and features, although we generally prefer Antec’s bundled power supplies.

Lian Li (http://www.lian-li.com) makes one or two entry-level cases but is primarily known for its high-end cases, most of which use aluminum construction and sell for premium prices. Construction quality is excellent on every Lian Li case we’ve seen, and they’re particularly popular for portable gaming systems and media center/home theater systems. Most Lian Li cases are sold without a power supply.

Silverstone (http://www.silverstonetek.com) was originally best known for its media center/home theater cases but has since branched out to become a full-line case supplier, selling an extremely broad range of cases—everything from entry-level mini-tower cases to cube-style small form factor cases to super-premium media center/home theater cases. We have limited experience with Silverstone cases, but all of those we’ve seen have been of excellent quality. Most Silverstone cases are sold without a power supply.

Thermaltake (http://www.thermaltakeusa.com) cases are particularly popular among extreme gamers—including those who want a case designed from the ground up for liquid cooling—but Thermaltake also offers a complete range of general-purpose cases, including small form factor and media center/home theater cases. Construction quality is generally excellent. Most Thermaltake cases are sold without a power supply.

Power Supply

The power supply is one of the most important components in a PC, yet most people give it little consideration. In addition to providing reliable, stable, closely regulated power to all system components, the power supply draws air through the system to cool it. A marginal or failing power supply can cause many problems, some of which may be very subtle and difficult to track down. Most problems are not subtle, however. A poor or marginal power supply is likely to cause system crashes, memory errors, and data corruption and may fail catastrophically, taking other system components with it.

Here are some important considerations when choosing a power supply:

Form factor

Unless you buy a case with a bundled power supply, the first thing to determine is which power supply form factor(s) will fit the case you’ve chosen. Full-tower cases and most mid- and mini-tower cases accept ATX12V or ATX12V/EPS12V power supplies. Smaller cases, including most μATX cases, may accept only smaller power supplies that use the SFX12V, TFX12V, or a proprietary form factor. Mini-ITX cases accept Mini-ITX power supplies.

Wattage rating

Wattage ratings are not standardized, so you can’t simply compare numbers. High-quality power supplies have wattage ratings specified at 40° C or 50° C. Cheap power supplies are often rated at 25° C or even 20° C, which unrealistically inflates their capabilities. For example, a power supply rated for 500W output at 20 °C may actually be able to deliver only 300W at the higher temperatures actually present in a working power supply.

Size your power supply according to the system configuration, taking into account the wattage required by the processor and video adapter. The power draw of mainstream processors ranges from less than 30W to more than 130W. Video power draw ranges from just a few watts for most integrated video adapters to 150W or more for some high-end gaming video adapters, and 400W or more for extreme gaming systems with dual or quad video adapters.

Main power connector

The original ATX specification and the ATX12V 1.X specifications defined a 20-pin main power connector, which provides power to the motherboard. The ATX12V 2.0 specification (and later versions) expanded the main power connector to a 24-pin version that is a superset of the original 20-pin connector. That is, the 20 original pins remain the same, and 4 additional pins have been added to one end of the connector.

Most (but not all) current motherboards and power supplies have the 24-pin connector. Many 24-pin power supplies use a split connector, which allows them to function as a 20-pin or 24-pin power supply, depending on which type your motherboard requires. Most 20-pin motherboards can accept the 24-pin connector directly simply by allowing the extra 4 pins to overhang the motherboard connector, but some motherboards have capacitors or other components that prevent a 24-pin plug from being inserted completely into the 20-pin socket. Sometimes it’s possible to bend the interfering component slightly to provide clearance for the 24-pin power supply plug, but 24-to-20 pin adapter cables are readily available and a much safer solution.

Even if you use a 20-pin motherboard, we recommend buying a 24-pin power supply.

Supplemental power connectors

Early PCs required only the main power connector to power the motherboard and peripheral power connectors to power drives and other peripherals. As the power requirements of newer and faster processors and video adapters continued to climb, supplemental power connectors were added to power supplies to meet these larger power requirements. Modern power supplies provide the following supplemental connectors:

+12V power connector

All current ATX12V power supplies provide a four-pin (2×2) +12V power connector, which connects directly to the motherboard near the processor and provides power to the processor.

Power supplies that meet both the current ATX12V and EPS12V standards provide at least one eight-pin (4×2) secondary +12V power connector and, optionally, a tertiary +12V power connector that uses the same four-pin arrangement as the connector on a standard ATX12V 2.X power supply. To ensure backward compatibility with ATX12V motherboards, most EPS12V power supplies implement the secondary eight-pin connector or connectors as a splittable pair of four-pin 2×2 connectors, either of which may be used to provide processor power on an ATX12V motherboard.

Most standard motherboards require only a single four-pin +12V power connector, which is provided by any current ATX12V power supply. You need concern yourself with this issue only if the motherboard you use requires additional +12V power connectors, which very few desktop boards do.

PCIe power connector

Most entry-level and midrange video adapters can be powered by the PCIe slot itself, but the very fast video adapters used in some gaming systems may require much more power than the PCIe slot can provide. ATX12V 2.1 and later power supplies include one or more PCIe power connectors to provide additional power to the video adapter(s). These connectors plug directly into the video card, supplying supplemental power beyond the power that is provided by the PCIe slot itself.

The original PCIe power connector, defined in the ATX12 2.1 specification, is a six-pin (3×2) connector that must be capable of providing up to 75W of supplemental power to a video adapter. ATX12V 2.2 defined an eight-pin PCIe power connector, which is capable of supplying another 150W to a video adapter. The eight-pin adapter is backward-compatible with the six-pin adapter, so a video adapter that requires a six-pin PCIe power connector can be powered by an eight-pin PCIe power connector. The converse is not true, because a video adapter that accepts an eight-pin PCIe power connector may draw more current than the six-pin PCIe power connector can supply.

Current ATX12V power supplies provide at least one six-pin PCIe power connector. High-wattage power supplies may provide several PCIe power connectors (six-pin, eight-pin, or some combination thereof). For example, the PC Power & Cooling TurboCool 1200, a power supply popular among extreme gamers, provides three six-pin PCIe power connectors and three six/eight-pin PCIe power connectors.

Other than gaming systems, very few systems require the PCIe power connector at all, so most builders need not concern themselves with the number or type of PCIe power connectors present on the power supply. If you are building an extreme gaming system, pay close attention to the number and type of PCIe power connectors, particularly if you plan to install two or more video cards.

Peripheral power connectors

All power supplies provide some or all of the following peripheral power connectors:

Berg power connector

Some current power supplies no longer include the obsolete four-pin Berg power connector. This connector was used to power floppy disk drives, which are now nearly extinct. (Of the few floppy disk drives still available, most of them, including internal models, are USB-powered.) Absence of Berg connectors is no reason to reject a power supply. If for some reason you actually need a Berg connector, you can buy an inexpensive Molex-Berg adapter cable. Some power supplies that do not include Berg connectors include such an adapter cable in the box.

Molex power connector

All current power supplies still include the obsolescent Molex power connector. The Molex power connector was originally used to power ATA/IDE and earlier types of hard drives and is still used today to power case fans, front-panel lights, and some internal peripherals.

SATA power connector

The SATA power connector is the current standard for powering internal peripherals, primarily hard drives and optical drives. There are actually three types of SATA power connector. The 15-pin Standard SATA power connector is used for powering optical drives and 3.5” hard drives in desktop systems and is the type you’ll find on any current desktop power supply.

The six-pin Slimline SATA power connector is used to power optical drives and 2.5” hard drives in notebook systems and some Mini-ITX and other small form factor systems. The nine-pin Micro SATA power connector is used to power 1.8” hard drives in notebook systems. If you need to install a peripheral that requires a Slimline or Micro SATA power connector in a desktop system, you can purchase a Standard-to-Slimline or Standard-to-Micro SATA power adapter cable.

For mainstream systems, nearly any current power supply provides enough SATA power connectors to power all of the drives. If you are building a server that will have many hard drives installed, make sure you have sufficient SATA power connectors for all of the hard drives and optical drives you plan to install. (We once built a tower system that contained 12 internal drives; we ran out of drive bays and SATA power connectors at the same time.) The length of the SATA power cables is seldom an issue unless you are building a full-tower system populated with a lot of drives. In that situation, pay careful attention to the lengths of the SATA power cables and the positions of the connectors on the cables.

Manufacturer

Many power supply vendors don’t actually manufacture power supplies, but simply relabel power supplies made by other companies. That’s not necessarily a bad thing. For example, two of the most reputable power supply vendors, Antec and PC Power & Cooling, purchase most or all of their power supplies from other companies and rebrand them. Some first-rate companies, notably Seasonic, manufacture and sell power supplies under their own name and also distribute power supplies to other companies for relabeling. Still other companies do all three: manufacture and sell power supplies under their own brand names, manufacture power supplies for relabeling by other companies, and buy power supplies from other manufacturers and put their own brand names on them.

About the only way to determine which company actually made a particular power supply is to check the UL label and look up the manufacturer in a database. Here’s one website that maintains such a list:

http://www.jonnyguru.com/modules.php?name=NDArticles&file=print&ndar_id=24

Some power supply manufacturers produce only cheap power supplies. Some produce a wide range of models, from cheap to first-rate units. A few produce only high-quality power supplies. Among the manufacturers (as opposed to brand names) we’ve found to be consistently reliable are Channel Well (USA), FSP/Sparkle, Seasonic, and Win-Tact.

Here are some less important things to consider when choosing a power supply:

Modular power supply cables

Many power supplies have fixed power cables. If you’re building a system in a relatively large case, that’s not a problem; there are plenty of places to tuck unused power cables out of the way. But in smaller systems, space is at a premium. The last thing you want is unused power cables cluttering up the available space, blocking air flow and interfering with fans. That’s why some power supplies use modular cables. The cables that are always required—the main power cable, the +12V motherboard power cable, and perhaps one peripheral power cable—are permanently affixed to the power supply. You can connect optional cables, such as the PCIe power cable and additional peripheral power cables, only if you need them. There are several jacks on the power supply body that accept PCIe power cable(s) and additional peripheral power cables, allowing you to install only those cables you actually need. Most modular power supplies come with a good selection of optional cables, but if you need extra cables of a particular type you can usually buy them separately from the company that makes the power supply.

All other things being equal, a power supply that provides modular cables is usually a bit more expensive than one that does not. More importantly, it is also a bit less reliable, at least in theory, because it uses plugs and jacks rather than soldered connections. On balance, we tend to use a modular power supply when we build a small system and a standard power supply when we build a full-size system, but the difference is not enough either way to make it a major decision factor. Treat modular power cables as a tiebreaker if you’re choosing between two otherwise similar power supplies. Otherwise, don’t worry about them.

Efficiency

Typical power supplies of a few years ago had efficiencies around 70%, which meant that 100W of AC input was required to provide 70W of DC output, with the remaining 30W simply producing waste heat inside the power supply. Most current power supplies are rated at 80% or higher efficiency, and the most efficient exceed 90%. The issue is not simply wasted power and higher electricity bills; it’s that waste heat. Heat is the enemy of PCs, and it’s worthwhile to reduce waste heat whenever possible.

The efficiency of any power supply varies with load. Efficiency is maximum at somewhere near the middle of its range. For example, a power supply that is accurately rated for 400W maximum wattage will be most efficient somewhere near 200W output, with efficiency gradually decreasing as the load approaches either 0% or 100%. That’s why it’s a mistake to buy too much power supply for your system. If your power supply is routinely operating at, say, 25% of rated wattage, the good news is that it’ll probably last forever, but the bad news is that it’s operating inefficiently. Conversely, it’s a mistake to buy too little power supply. If your power supply routinely operates at 90% of rated capacity, it’s also operating very inefficiently and will probably die an early death.

Small differences in power supply efficiency may be significant if you’re a corporation with a fleet of thousands of desktop or server PCs. Otherwise, give preference to the more efficient unit, but don’t make it a deciding factor.

Power-factor correction

Power supplies of a few years ago typically had power factors of 67% or so. In essence, that means those power supplies drew 150% of their nominal maximum input power two-thirds of the time and zero power one-third of the time. For example, a power supply with a rated input of 300W and operating at full capacity would actually draw 450W for two out of every three cycles of the 60 Hz AC current, and 0W for the third cycle. In other words, it would draw 450W for 1/30th of a second and then no power for the following 1/60th of a second. That may seem a small difference from drawing 300W continuously, but it has major implications for businesses that must size their computer rooms’ electrical infrastructure and UPSs for the maximum current draw rather than an average draw.

Some years ago, power supplies with power-factor correction (PFC) started becoming more common. Most of these early PFC power supplies used passive PFC, which amounts to a simple coil that helps smooth out current draw. These units were able to improve power factors to 75% or more. A few current PFC power supplies use passive PFC, but most use active PFC, which uses live electronic components to smooth current draw. A typical active PFC power supply improves the power factor to 90% or more, and the best reach 99%. For home use, PFC efficiency is a minor issue, but it’s still worth choosing a power supply with a high PFC factor, if only because such power supplies are generally better-made than units with lower PFC factors.

Use the following guidelines to choose a power supply appropriate for your system:

• Above all, make sure the power supply you buy fits your case and has the proper connectors for your motherboard and, if applicable, your video card(s).
• Assuming honest wattage ratings (like those from the vendors we recommend), for a typical entry-level system, install a 300W or larger power supply. For a mainstream system, install a 400W or larger power supply. For a high-performance system, install a 500W or larger power supply (possibly much larger). If you’re installing multiple video adapters in an NVIDIA SLI (Scalable Link Interface) or AMD/ATI CrossFireX configuration, make sure to use a power supply that is certified for such use.

Recommended power supply brands

There are many good power supply brands, but Antec (http://www.antec.com), PC Power & Cooling (http://www.pcpower.com), and Seasonic (http://www.seasonicusa.com) stand out from the crowd. All three companies offer a wide range of power supplies to fit various needs.

Processor

A few years ago, if we checked an online PC component vendor for desktop processors, there would have been maybe eight or nine Intel processors on offer and about the same number of AMD processors. When we started working on this section, we checked the NewEgg desktop processor page. There were 44 AMD processors available, ranging in price from $36 to $296, and 42 Intel processors, ranging in price from $41 to $1,000. That’s a lot of options. How can AMD and Intel possibly make that many different processors? And how can anyone make the best choice among 86 processors ranging in price from $36 to $1,000?

The short answer is that AMD and Intel really don’t make all that many different processors. They each make only a few significantly different models, but offer those models in various speeds and with different combinations of minor features. It’s like buying a car. The base model may have a dozen options available, and by the time you total up all the available permutations, that base model can be ordered in thousands of different configurations.

A particular processor series may be available in numerous models that vary in speed, number of cores, number of threads per core, amount of L3 cache, power consumption, and so on. There is also frequently pricing overlap between models from different processor series. For example, the day we checked, no less than four different Intel processors were selling for exactly the same price, $199.99. Table 2-1 lists them.

Table 2-1. Four $200 Intel processor models
Model	Cores	Threads	Speed	L3 cache	Process	Socket	Power
Core 2 Quad Q9300	4	4 (1/core)	2.5 GHz	6 MB	45 nm	LGA 775	95W
Core i5-660	2	4 (2/core)	3.33 GHz	4 MB	32 nm	LGA 1156	73W
Core i5-661	2	4 (2/core)	3.33 GHz	4 MB	32 nm	LGA 1156	87W
Core i5-750	4	4 (1/core)	2.66 GHz	8 MB	45 nm	LGA 1156	95W

As Intel was fully aware when it set the prices for these four processors, their overall performance is very similar. Will a quad-core processor with a slower clock speed but more L3 cache outperform a dual-core processor with a faster clock speed and less L3 cache, or vice versa? Will a dual-core processor that runs two threads per core outperform a quad-core processor that runs only one thread per core, or vice versa? The answer to all of those questions is both yes and no, because the winner will be different for different applications. Some applications are clock speed–bound, while others are core-bound, thread-bound, or L3 cache–bound—or some combination of those factors. The upshot is that on average you get $200 worth of performance, whichever of these $200 processors you use.

But what about the Core i5-660 and -661 models? They appear to be identical other than in power consumption. Why would anyone pick the -661 for the same price as the -660, when the only difference appears to be that the -661 draws more power? Here’s where we get into really minor differences:

• The -660 and -661 both include integrated Intel HD Graphics, but the -660 uses a basic graphics frequency of 733 MHz, while the -661 boosts that to 900 MHz for faster video performance.
• The -660 supports Intel Virtualization Technology for Directed I/O (VT-d), a memory-mapping technique that has both advantages and disadvantages. The -661 does not support VT-d.
• The -660 supports Intel Trusted Execution Technology, which is essentially hardware support for Digital Restrictions Management (DRM). The -661 has no such support.

As it happens, we wanted to use a processor in this price range for our mainstream system. So which one did we choose, and how did we choose it? All four are excellent processors, but…

• We first eliminated the Core 2 Quad Q9300 from consideration. It’s based on older technology, but that’s not necessarily a problem. We ruled out the Core 2 Quad because it uses the obsolescent Socket LGA 775, which would have limited our choice of motherboards and future upgrade/repair options.
• We next eliminated the Core i5-750. It’s an excellent processor, and for some people would be the best choice of the four. However, it has the highest power consumption of the remaining candidates, and lacks integrated video.
• With the choice down to the Core i5-660 and -661, the decision becomes almost a toss-up. In return for its somewhat higher power consumption, the -661 provides faster integrated graphics. The -661 lacks VT-d (which we’d kind of like to have just on general principles), but it also lacks hardware DRM, which we’d definitely prefer not to have.

On that basis, we chose the Intel Core i5-661, but the truth is that any of the other three would have done nearly as well.

Most people spend too much time dithering about which processor to install. The only really important decision is how much to spend. After you make that decision, it really comes down to the minor details. Here are the considerations for each of the processor price ranges:

Embedded ($60 to $150, including motherboard)

Although tiny motherboards with embedded processors have been available for years, they didn’t really become mainstream items until Intel introduced its Atom series of low-power processors. Although the Atom is certainly no speed demon, it is fast enough for appliance systems, and even as the foundation for small, quiet, low-power general-purpose systems, such as a bedroom, kitchen, or kids’ room PC. Atom processors aren’t available separately at retail vendors. Intel sells them in bulk to OEMs and systems integrators, who build motherboards or complete systems around them.

Purchasing an Intel Atom processor means also buying the motherboard it’s soldered onto. Most such motherboards include “integrated everything,” including video, audio, LAN, and often WiFi. Such motherboards sell for $60 to $150 or more, depending on their features, their quality, and the Atom model they include. A very small minority of those motherboards are μATX, and there are even one or two full ATX models available, but the vast majority are Mini-ITX boards, designed to fit the smallest cases.

The Atom is available in single- and dual-core models at various clock speeds. The slowest single-core models are appropriate for an appliance system such as a network-attached storage (NAS) box or a Linux router. The faster dual-core Atom processors are fast enough that most people consider them suitable for light general use, such as checking email or browsing the Web. (Robert’s den system is a passively cooled and nearly silent dual-core Atom Mini-ITX box running Linux, which he uses for checking web mail and looking up details on IMDb while we’re watching videos.)

Entry level (under $100)

Our praise of the Intel Atom aside, these are the least expensive processors that are really suitable for a general-purpose PC. At the bottom of this range—sub-$60 processors—are the least expensive socketed AMD Sempron and Intel Celeron models. These processors are low-power in both senses, providing limited CPU performance while consuming 65W or less. They are single-core or very low-end dual-core, and are suitable if you need more performance than the fastest Atom processors can provide but don’t need the performance of a more expensive processor.

The $60 to $100 price range is the realm of low-end dual- and triple-core processors and a few of the least expensive quad-core units. AMD owns this market segment with its Athlon II and Phenom II series of two-, three-, and four-core processors. In this price range, AMD processors offer more cores, more threads, and more general bang for the buck than similarly priced Intel processors. Any of these processors is an excellent choice for an entry-level general-purpose system.

Mainstream ($100 to $200)

Although AMD offers many processors in this price range, Intel owns this market segment. Intel processors simply provide more bang for the buck than AMD processors in this price range—sometimes dramatically so.

In the lower half of this price range, dual-core Intel Core i3-series processors consistently edge out comparably priced dual- and triple-core AMD Phenom and Athlon II processors for most applications. In the upper half of this price range, dual- and quad-core Intel Core i5-series processors outperform comparably priced quad- and hex-core AMD processors.

Mainstream processors, particular those from the upper half of the price range, are the best choice for most systems. They are noticeably faster than entry-level processors and cost only a little more, while at the same time they are only a bit slower than performance processors.

Performance ($200 to $400)

With the exception of its fastest hex-core model, AMD no longer offers desktop processors in the $200 and higher price range. This is the realm of fast Core 2 Quad and Core i5 processors and midrange Core i7 processors. Performance processors are popular primarily among serious gamers, who are willing to spend extra money in return for the slight competitive edge that one of these faster processors provides.

Extreme ($400+)

This rarefied range is thinly populated by a few of the fastest Intel Core i7 processors, up to and including $1,000+ processors. Extreme processors are favored by well-heeled gamers and by business users like real-time commodity traders, to whom literally a fraction of a second may mean the difference between a huge profit and a huge loss. These processors are faster than mainstream and performance processors (sometimes noticeably so). If performance is the top priority and cost is not an issue, choose an extreme processor. Otherwise, spend that extra money on extra memory, a better display, a faster video card, or a larger hard drive. Or just keep the money in your bank account.

Also consider the following issues when you choose a processor:

• Even a $75 processor is fast enough for typical general-purpose computing, such as reading email, browsing the Web, running office productivity applications, burning DVDs, watching YouTube videos, and so on. If you never load the system heavily, you won’t notice much difference between that $75 processor and a more expensive model.
• Budget processors are hampered not just by their lower clock speeds and smaller number of cores and threads, but by small caches, which cripple performance, particularly if you work with large data sets (e.g., rendering video). In such situations, the performance difference between a budget processor and even a mainstream processor (let alone a performance or extreme processor) can be like the difference between night and day. For example, Robert uses a fast quad-core processor to render video much faster than real-time. (That is, rendering a 10-minute video takes only a couple of minutes.) Rendering that same 10-minute raw video segment on a budget processor might take 10 or 15 minutes, which is a huge difference when you’re actually sitting staring at the screen waiting for the render to finish.
• Processors in the “sweet spot” range—$150 to $200 for a retail-boxed processor—usually represent the best bang for the buck. They’re noticeably faster than budget processors for even moderately demanding applications, and are fast enough even for CPU-intensive tasks like casual video rendering.
• Buy the processor you need initially, rather than buying a slower processor now and planning to upgrade later. Processor upgrades, both AMD and Intel, are a minefield of compatibility issues.

Recommended processors

For appliance systems, the Intel Atom is hard to beat. Intel, ASUS, and GIGABYTE all offer very good Mini-ITX motherboards with embedded Atom processors. Zotec offers excellent Atom motherboards, but they’re very pricey (too much so, we think).

For budget systems, dual-, triple-, or quad-core AMD Athlon II or Phenom processors in the sub-$100 range offer more cores, threads, and overall performance than similarly priced Intel processors, although Intel sells a lot of Pentium E- and G-series processors in this price range due to its reputation for quality, compatibility, and reliability.

For mainstream and higher systems, it’s all Intel. We recommend choosing your processor according to your budget from among the broad range of Intel Core i3, Core i5, and Core i7 processors.

Heatsink/Fan Units (CPU Coolers)

Modern processors may consume from 25W to 130W or more. Nearly all systems deal with the resulting heat by placing a massive metal heatsink in close contact with the processor and using a fan to draw air through the heatsink fins. This device is called a heatsink/fan (HSF) or CPU cooler.

Most AMD and Intel processors are available with or without a bundled CPU cooler. The bulk, OEM, or white-box form includes just the processor. The retail-boxed form costs a few bucks more and includes the same processor with a bundled CPU cooler (and, usually, a longer warranty).

Bundled CPU coolers have gotten much better over the years. Although aftermarket units may provide better cooling and lower noise levels, the bundled CPU cooler is generally more than good enough to cool the CPU while keeping noise down to an acceptable level. Gamers, overclockers, and others who run their processors hot may find an aftermarket CPU cooler worth the extra money, as may those who want to build the quietest possible system. For most other people, a retail-boxed processor with bundled CPU cooler is the best choice.

Use the following guidelines when choosing an HSF:

• Make certain the HSF is rated for the exact processor you use. An HSF that physically fits a processor may not be sufficient to cool it properly. By definition, a bundled HSF is rated for the processor it comes with.
• Make sure the HSF is usable with your motherboard and case. Some HSFs are incompatible with some motherboards because clamping the HSF into position may crush capacitors or other components near the processor socket. Good third-party CPU coolers often list compatibility with specific motherboard models. Check the website. Premium CPU coolers are usually physically large. If you intend to purchase such a cooler, make sure your case has enough clearance between the motherboard and cover panel for the cooler to fit.
• Pay attention to noise ratings. Some high-efficiency HSFs designed for use by overclockers and other enthusiasts have very noisy fans. Other HSFs are nearly silent. The best (and most expensive) third-party CPU coolers are both highly efficient and very quiet. Note that these coolers are often supplied without a fan, which you must purchase separately from among the models recommended for that specific cooler.
• Use the proper thermal compound. When you install an HSF, and each time you remove and replace it, use fresh thermal compound to ensure proper heat transfer. Thermal compound is available in the form of viscous thermal “goop” and as phase-change thermal pads, which melt as the processor heats up and solidify as it cools down. Make sure that the thermal compound you use is approved by the processor maker and, if applicable, the cooler maker.

Recommended CPU cooler brands

Unless you have good reason to do otherwise, buy a retail-boxed CPU with a bundled cooler. If you need high-performance cooling or an extremely quiet cooler, purchase a CPU cooler made by Arctic Cooling (http://www.arctic-cooling.com), Scythe (http://www.scythe-usa.com), Thermaltake (http://www.thermaltakeusa.com), or Zalman (http://www.zalmanusa.com).

Motherboard

The motherboard is the main logic board around which a PC is built. It is the center of the PC, in the sense that every system component connects to the motherboard, directly or indirectly. The motherboard you choose determines which processors are supported, how much and what type of memory the system can use, what type of video adapters can be installed, the speed of communication ports, and many other key system characteristics.

Here are some important considerations when choosing a motherboard:

CPU socket

The fundamental characteristic of a motherboard is the CPU socket, because it determines which processors physically fit that motherboard. There are two types of sockets:

Intel sockets

The oldest of the current Intel sockets is the obsolescent Socket LGA 775, also known as Socket T. Socket LGA 775 processors and motherboards are still widely available, and are likely to remain so through at least 2011 and into 2012. It’s not necessarily a bad idea to build a new system around an obsolescent socket—prices on older gear are often excellent—as long as you’re aware that it may be difficult or impossible to find a replacement motherboard or processor two or three years down the road. At worst, you may need to replace both the motherboard and the processor if one or the other fails. Socket LGA 775 motherboards accept dual-core Pentium processors, single- and dual-core Celeron processors, and Core 2 Duo and Core 2 Quad processors.

Intel’s current entry-level/mainstream socket is LGA 1156, also known as Socket H. With this socket, Intel moved functions formerly performed by the separate northbridge section of the chipset onto the processor chip itself, which accounts for the need for almost 400 more pins than LGA 775. The high-speed Direct Media Interface (DMI) link formerly used to link the northbridge and southbridge sections of the chipset is now implemented within the processor package, as are the QuickPath Interconnect (QPI) formerly used to connect the processor to the I/O hub (southbridge), a PCI Express 2.0 x16 link for communication with a graphics adapter, and a dual-channel DDR3 memory controller. In effect, LGA 1156 moves all of the high-bandwidth functions formerly performed by a separate northbridge onto the processor itself and leaves only the former southbridge chipset functions on the separate southbridge chip, now called the Platform Controller Hub (PCH). Socket LGA 1156 motherboards accept Core i3, Core i5, and Core i7 processors. LGA 1156 processors and motherboards are likely to remain Intel’s mainstream desktop solution for the next few years.

Intel’s current performance socket is Socket LGA 1366, also known as Socket B. Like LGA 1156, LGA 1366 uses an external southbridge chip for USB and other communications functions, and incorporates a DDR3 memory controller on the processor (triple-channel DDR3, in this case). The major difference between LGA 1156 and LGA 1366 is that the latter moves PCI Express off the processor chip, using a QPI link to communicate with a separate partial northbridge chip that supports video and other PCI Express communications. Socket LGA 1366 motherboards currently accept only Core i7 and Xeon (workstation/server) processors. LGA 1366 processors and motherboards are likely to remain Intel’s performance desktop solution for the next few years.

AMD sockets

Like Intel, AMD has three current processor sockets, but in AMD’s case the three sockets are actually just very minor variations of one socket. In fact, the 940-pin Sockets AM2 and AM2+ and the 941-pin Socket AM3 are pin-compatible physically, although a processor designed for one of those sockets will not necessarily operate in one of the other sockets.

Socket AM2 dates from May 2006, when it replaced Sockets 754 and 939 for Sempron and Athlon processors. Socket AM2 is characterized by its support for DDR2 memory, HyperTransport 2.0 operating at 1.0 GHz, and one power plane to drive both cores and the Integrated Memory Controller (IMC).

Although Socket AM2 processors have been discontinued, Socket AM2 motherboards are physically and electrically compatible with current Socket AM2+ and AM3 processors. Note, however, that using a later processor in an AM2 motherboard requires a BIOS update that may or may not be available from the manufacturer, and the old motherboard will not be able to take advantage of additional features supported by the newer processor. Although Socket AM2 motherboards may be available new for some time to come, don’t buy one.

Socket AM2+ is a transitional socket. The two major upgrades from Socket AM2 are the addition of HyperTransport 3.0 operating at 2.6 GHz and split power planes, one of which powers the CPU cores and the other the IMC. Socket AM2+ motherboards can use AM2, AM2+, or AM3 processors, assuming that a suitable BIOS is available, and are often promoted as “Socket AM3-Ready,” although they are limited to using DDR2 memory. Socket AM2+ motherboards are widely available and likely to remain so for some time to come.

Socket AM3 is AMD’s current mainstream socket. Socket AM3 differs from AM2+ in that it supports processors that use either DDR2 or DDR3 memory.

If you intend to build an AMD-based system, buy a Socket AM2+ or (preferably) a Socket AM3 motherboard.

Supported CPU types

Just because a CPU physically fits the socket on a motherboard doesn’t mean that CPU is compatible with that motherboard. Two issues affect compatibility:

VRM support

Voltage Regulator Modules (VRMs) on a motherboard supply power to the processor. VRMs are rated to deliver specific maximum current. If a particular processor draws more current than the VRMs can provide, that processor cannot be used on that motherboard.

BIOS compatibility

Even if a particular motherboard is compatible in every respect with a particular processor, that processor cannot be used in that motherboard if the BIOS does not support that specific processor model. Even an apparently trivial difference between a supported processor and a new model may be enough to make that new model unusable without a BIOS update.

Form factor

Like cases, motherboards are available in numerous form factors, the most popular of which are full ATX, μATX, and Mini-ITX. Although Mini-ITX motherboards are physically compatible with μATX and ATX cases and μATX motherboards with ATX cases, it seldom makes sense to install a smaller form factor motherboard in a larger case. A μATX motherboard generally costs about as much as a similar full ATX motherboard, and Mini-ITX motherboards are often actually more expensive than similar μATX and ATX boards. Using a smaller motherboard than the case can accept simply gives up memory and expansion slots, communications ports, and other potentially valuable features for no reason. It may also cause problems with cables that are too short to reach the smaller motherboard.

Memory slots

Every motherboard includes one or more memory slots, but the number and type of these slots differ, as follows:

Number

Typical ATX motherboards provide slots for four memory modules. μATX boards usually provide two or four slots. Mini-ITX boards provide one or two slots. All other things being equal, it’s better to have more memory slots than fewer, for three reasons. First, the number of memory slots (and the maximum capacity of each) puts a hard limit on the amount of memory that can be installed. Second, it may be less expensive—sometimes much less expensive—to install a given amount of memory using more smaller modules rather than fewer larger modules. For example, if you intend to install a total of 8 GB of memory, the price of four 2 GB memory modules may be much less than the price of two 4 GB modules. Third, if you decide to upgrade your system in mid-life, it’s easier and cheaper to add memory modules than to remove and discard the current modules and replace them with all new modules.

Type and capacity of modules supported

Current desktop motherboards use one of four types of memory modules: 240-pin DDR2-SDRAM DIMM, 240-pin DDR3-SDRAM DIMM, 200-pin DDR2-SDRAM SO-DIMM, or 240-pin DDR3-SDRAM SO-DIMM. The SO-DIMM modules are actually notebook memory, but they are used by a few Mini-ITX motherboards. Mainstream ATX, μATX, and Mini-ITX motherboards accept either 240-pin DDR2 or DDR3 DIMMs. A few hybrid Socket AM2+/AM3 motherboards accept either 240-pin DDR2 or DDR3 DIMMs, but only one type at a time.

Just because a memory slot physically accepts a memory module doesn’t mean that module is compatible with that motherboard. The chipset and/or BIOS may limit the module capacity. For example, although a 4 GB DDR2-SDRAM module may be physically identical to a 2 GB module, a particular motherboard may support 2 GB modules, but not 4 GB modules. If you install the 4 GB module, the results are unpredictable. The system may fail to boot. More likely, it will boot normally, but will recognize the 4 GB module as only 2 GB. If the BIOS is limiting module capacity, updating the BIOS may allow the system to recognize the larger memory modules. If the chipset is limiting module capacity, there’s no fix.

Note that the maximum memory module capacity (and the maximum overall memory) supported by a motherboard may depend on the number of memory slots occupied and/or the speed of that memory. For example, a particular motherboard may support 8 GB of total memory in the form of two 4 GB modules, but not in the form of four 2 GB modules. Similarly, a motherboard may allow all four of its memory slots to be populated with slower memory, but only two slots with faster memory.

Before you purchase a motherboard, make sure it supports the specific memory configuration you want to use. Check the detailed motherboard specifications web page and also use one or more of the memory configurator pages provided by Crucial, Kingston, and other memory vendors.

Expansion slots

Most motherboards include expansion slots, which you can populate with expansion cards to provide functions that the motherboard does not. For example, if you purchase a motherboard that provides only USB 2.0 communication ports, you can later upgrade your system by installing a USB 3.0 expansion card. Most current motherboards provide some mix of PCI and PCI Express (PCIe) slots.

PCI

The older PCI standard defines one specific set of physical and electronic requirements for a PCI expansion slot, so any current PCI expansion card can be used in any current PCI slot (assuming that the case puts no constraints on the physical height or length of the card). In June 2010, Intel announced that it was retiring the PCI specification in favor of PCI Express. PCI expansion cards and motherboards with PCI slots are likely to remain available for some time to come, but development of new PCI products has now largely ended, and future products will all be PCI Express.

PCI Express

The newer PCIe standard defines the physical and electronic characteristics of a connector that may be used individually (called PCIe 1x) or may be concatenated into a single combined 4x, 8x, or 16x PCIe connector to provide correspondingly higher bandwidth.

PCI Express is now used universally for video adapter cards, but as of mid-2010 PCI expansion cards still greatly outnumber PCI Express expansion cards, despite the fact that PCI Express is superior and has been available for years. Until now, most expansion card makers have adopted an if-it-ain’t-broke-don’t-fix-it attitude. With Intel’s deprecation of the PCI standard, we expect that to change, with more and more PCI Express expansion cards becoming available over the coming months and years and PCI expansion cards becoming increasingly hard to find.

Communications ports

All motherboards provide various communications ports. Most still provide at least one legacy serial port and parallel port, although these ports are now often present only as header-pin sets on the motherboard rather than as physical connectors on the rear I/O panel. If you need to use them, you’ll need to install a cliffhanger bracket (usually supplied with the retail-boxed motherboard) to extend the ports out to an expansion slot cover. Beyond these legacy ports, current motherboards may provide some or all of the following communication ports:

USB 2.0

All current motherboards provide multiple USB 2.0 ports, both as physical connectors on the rear I/O panel and as header-pin sets that can be extended to the front-panel connectors. Most current peripherals—from keyboards and mice to speakers, printers, scanners, and digital cameras—use the USB 2.0 interface, so it’s important to have enough for your needs. (Of course, you can always add a USB hub, but that increases costs and desktop clutter.) Look for a motherboard with at least six USB 2.0 ports; eight or more is better.

USB 3.0

The most recent USB standard, USB 3.0, will eventually replace USB 2.0, just as USB 2.0 eventually replaced USB 1.0 and USB 1.1. The raison d’être for USB 3.0 is speed. USB 2.0 has a nominal data rate of 480 Mb/s (60 MB/s) and typical real-world throughput (after protocol overhead) of 25 MB/s. For most peripherals, that’s more than fast enough, but it’s marginal for such high-speed peripherals as external hard drives.

USB 3.0 has a nominal raw data rate of 4 Gb/s (500 MB/s) and real-world throughput of about 400 MB/s, fast enough even for external hard drives and solid-state drives. USB 3.0 is also backward-compatible with devices that use earlier USB standards, although those devices still run at the lower data rates for which they were designed.

Eventually, USB 3.0 ports will be ubiquitous, but for now most motherboards do not provide any USB 3.0 ports at all. Those that do generally provide only one or two, in combination with several USB 2.0 ports. As devices that need the higher speed of USB 3.0 and chipsets that support USB 3.0 become more common, more motherboards will start including USB 3.0 ports. For now, we consider USB 3.0 support a nonissue. If you later need USB 3.0, it will be easy enough to buy an inexpensive USB 3.0 PCI Express adapter card and install it in your system.

eSATA

Many current motherboards provide eSATA ports, sometimes as internal connectors, sometimes on the rear I/O panel, and sometimes both. These ports are used primarily to connect external hard drives and hard drive docking stations, both of which are common backup devices. If you use or plan to use eSATA devices, you’ll want at least one or two eSATA ports available on the motherboard you choose.

IEEE-1394 (FireWire)

Some motherboards—particularly “media” series models—include one or two IEEE-1394 (FireWire) ports, usually as a physical connector on the rear I/O panel and perhaps a second port as header pins on the motherboard. These ports may be IEEE-1394a (FireWire 400), which has a maximum data rate of 400 MB/s, or IEEE-1394b (FireWire 800), which has a maximum data rate of 800 MB/s.

FireWire is a declining standard. It’s still widely used in some specialty markets (notably, camcorders), but it is being displaced by eSATA and USB 3.0. If you need FireWire to support your camcorder or other device, look for a motherboard that provides one or two FireWire ports. If a motherboard you otherwise like has no FireWire ports, you can add them easily by installing an inexpensive PCI or PCI Express FireWire expansion card.

Use the following guidelines when choosing a motherboard:

• For a general-purpose system, choose a μATX motherboard, unless you need the additional expansion slots available on a full ATX model. For a small system, use a μATX motherboard. For a really small system, use a Mini-ITX motherboard. (Note that Mini-ITX motherboards often cost significantly more than μATX models with similar features and quality, so use Mini-ITX only if you really need it.)
• For an Intel-based system, choose a Socket LGA 1156 or a Socket LGA 1366 motherboard that is compatible with your choice of processor. Avoid Socket 775 (Socket T) motherboards, which are obsolescent. For an AMD-based system, choose a Socket AM3 motherboard if it is compatible with your chosen processor; otherwise, choose a hybrid Socket AM2+/AM3 model. Avoid Socket AM2 and Socket AM2/AM2+ models.
• Make sure the motherboard supports the exact processor you plan to use. Just because a motherboard supports a particular processor family doesn’t mean it supports all members of that family. You can find this information on the motherboard maker’s website or in the release notes to the BIOS updates. It’s also important to know exactly what revision of the motherboard you have, because processor support may vary by motherboard revision level.
• Make sure the motherboard supports the type and amount of memory you need. Do not make assumptions about how much memory a motherboard supports. Check the documentation to find out what specific memory configurations are supported.
• Before you choose a motherboard, check the documentation and support that’s available for it, as well as the BIOS and driver updates available. Frequent updates indicate that the manufacturer takes support seriously.

Recommended motherboard brands

For systems with Intel processors, we recommend Intel (http://www.intel.com), GIGABYTE (http://www.gigabyte-usa.com), and ASUS (http://usa.asus.com) motherboards. For AMD processors, we prefer GIGABYTE and ASUS. All three of these companies offer top-notch motherboards, in everything from basic vanilla units to feature-laden, high-performance boards. Intel tends to be a bit “stodgier” than the other two, emphasizing stability over performance and features, but the differences are small in any case. (If it’s any indication, most of our own systems use Intel motherboards.)

If cost is a major consideration, we recommend ASRock (http://www.asrock.com) motherboards. Originally spun off from ASUS in 2002 to compete with low-end motherboard makers, ASRock has developed a reputation for inexpensive motherboards with very good quality, stability, and feature sets. Recently, ASRock has introduced higher-end motherboards as well.

Memory

The only real decisions are how much memory to install, what size and type of modules to use, and what brand to buy. Consider the following factors when choosing memory modules (DIMMs):

• For budget systems, install no less than 512 MB per core, and at least 1 GB total. (For a single-core budget system, install 1 GB or more; for a dual-core system, 1 GB or more; for a triple-core budget system, 1.5 GB or more; and for a quad-core budget system, 2 GB or more.) For mainstream systems, install 2 GB or 4 GB, with a minimum of 512 MB/thread and 1 GB per core. For performance systems, workstations, and multimedia/graphics systems, install 4 GB or more, with a minimum of 512 MB/thread and 1 GB per core. If you’re running a 64-bit operating system, increase these recommended amounts by 50% to 100%.
• Memory manufacturers such as Crucial (http://www.crucial.com), Kingston (http://www.kingston.com), Corsair (http://www.corsairmemory.com), and Mushkin (http://www.mushkin.com) provide online memory configurators that allow you to enter the brand and model of your motherboard and return a list of compatible memory modules. Before you buy memory, use these configurators to make sure the memory you order is compatible with your particular motherboard.
• For motherboards that use 240-pin DDR2 memory, buy DDR2 memory modules of at least the speed required by your motherboard/processor combination. DDR2 memory is available in PC2 3200, PC2 4200, PC2 5300, PC2 6400, and PC2 8000 variants. Choose the fastest modules that do not sell at a significant price premium over slower modules. Once again, choose modules that support fast CAS latency timings only if they cost little or no more than modules with standard timings.

  DDR2 Versus DDR3

  If your motherboard supports both DDR2 and DDR3 memory modules, install DDR3 for higher performance and ease of future memory upgrades.

• For higher performance, install DIMMs in matched pairs or triplets to enable dual- or triple-channel memory operation.
• It’s generally less expensive to buy a given amount of memory in fewer modules. For example, if you are installing 4 GB of memory, two 2 GB DIMMs will probably cost less than four 1 GB DIMMs. Using fewer but larger DIMMs also preserves memory slots for future expansion. However, the largest-capacity modules often sell at a substantial premium. For example, a 4 GB DIMM may cost five times as much as a 2 GB DIMM, rather than only twice as much.
• Verify the memory configurations supported by your motherboard. For example, a particular motherboard may support 2 GB DIMMs, but not 4 GB DIMMs. Similarly, one motherboard may support 2 GB DIMMs in all four of its memory slots, but another may support 2 GB DIMMs in only two of its four slots. Check the motherboard documentation to determine the memory configurations your chosen motherboard supports.

  Ron Morse Comments

  This needs to be emphasized. With the trend toward 64-bit operating systems and more than 4 GB of installed RAM, problems related to this are becoming more common.

• Non-parity memory modules provide no error detection or correction. ECC modules detect and correct most memory errors but are slower and more expensive than non-parity modules. Use ECC memory if you install more than 8 GB of memory and the motherboard supports ECC memory. For 8 GB or less, use non-parity modules.

Recommended memory brands

There are many good memory vendors. Some, such as Corsair, Mushkin, OCZ, and Patriot, are well-known names among extreme gamers and other enthusiasts. But Crucial (http://www.crucial.com) and Kingston (http://www.kingston.com) are the two top-tier desktop memory vendors whose products we’ve used almost exclusively for the last decade or more. Both offer a broad line of memory and memory-related products at competitive prices, from budget modules to high-performance premium modules. Other than from lightning strikes, catastrophically blown power supplies, and similar misfortunes, we don’t recall ever having a memory module from either company fail in use.

Kingston is a packager: it buys memory chips from other companies and builds modules with those chips. Crucial (actually, Micron Technology) is fully vertically integrated. Sand goes in one side of its plants, and finished memory modules come out the other side. Because we have found Crucial memory to be utterly reliable and competitively priced, and because we like the idea of a company building (and controlling) its products from start to finish, we order Crucial memory whenever possible.

Floppy Disk Drive (FDD)

You probably don’t need an FDD in your new system. Few current motherboards provide the traditional FDD interface, so old-style FDDs are no longer usable. If you need FDD support to transfer data from old floppies, buy an inexpensive external USB floppy drive. If for some reason you want an internal FDD, buy an internal USB model, which connects to a set of USB header pins on the motherboard. If you need to boot from the FDD—for example, to run a classic computer game—configure the boot sequence section of BIOS Setup to put the FDD first, where it will be listed as a USB Mass Storage Device.

Recommended FDD brands

If you need an FDD, any brand is fine.

Hard Drive

It’s easy to choose a good hard drive. Hitachi, Samsung, Seagate, Toshiba (notebook drives only), and Western Digital produce drives at similar price points for any given size and type of drive.

Compatibility is not an issue for hard drives. Hard drives are plug-and-play devices. Any recent hard drive will coexist peacefully with any other recent hard drive or optical drive, regardless of manufacturer.

Use the following guidelines when you choose a hard disk:

• Hard drives are available in standard ATA (Parallel ATA or PATA) and Serial ATA (SATA) interfaces. PATA drives are obsolescent and are suitable only for upgrading older systems that lack SATA interfaces. For a new system, choose a drive that supports the SATA 3.0 Gb/s interface. A few newer premium models support the more recent SATA 6.0 Gb/s interface, but even the fastest hard drives are incapable of saturating a SATA 3.0 Gb/s interface, so the SATA 6.0 Gb/s interface is really just marketing hype.
• Although it’s tempting to buy the highest-capacity drive available, high-capacity drives often cost more per gigabyte than midrange drives, and the highest-capacity drives are often slower than midrange models. Decide what performance level and capacity you need, and then buy a drive that meets those requirements. Typically, you can choose the model based on cost per gigabyte. However, you may need to buy the largest drive available despite its higher cost per gigabyte and slower performance, simply to conserve drive bays and SATA ports.
• Choose a 7,200 RPM SATA drive for a general-purpose system. 10,000 RPM drives cost more than 7,200 RPM models, are not all that much faster, and are much noisier and hotter-running than 7,200 RPM models.
• Get a model with larger buffer/cache if it doesn’t cost much more. Some drives are available in two versions that differ only in buffer size. One might have a 16 MB buffer and the other a 32 MB buffer. It’s worth paying a few extra dollars for the larger buffer.

Recommended hard drive brands

We’ve used many desktop hard drives from all four companies that currently make them, not to mention drives from another dozen or two companies that have fallen by the wayside over the years. We usually install Seagate (http://www.seagate.com) drives, because they are fast, quiet, cool-running, reliable, and competitively priced. Western Digital (http://www.wdc.com) went through a very bad patch a few years ago, with high out-of-the-box and in-use failure rates. For years, we wouldn’t even consider installing a Western Digital drive. Fortunately, Western Digital has since addressed its quality problems and now produces reliable models across the full range of capacities and performance levels. We still seldom use Western Digital drives, though, because they generally cost more than comparable Seagate models. Samsung (http://www.samsung.com) produces a much narrower range of drives than Seagate or Western Digital, but the drives Samsung does offer are competitive with comparable units from Seagate and Western Digital. Like Samsung, the last member of the Big Four, Hitachi (http://www.hitachi.com), produces a relatively narrow range of models. After being burned by the fiasco several years ago when Deskstar drives were dropping like proverbial flies, we never installed another Hitachi drive, so we can’t comment on their current quality.

Solid-State Drive (SSD)

Solid-state drives serve the same purpose as hard drives—mass storage—but are purely electronic devices. Rather than storing data on spinning disk platters, an SSD stores data in flash memory. All but the least expensive SSDs are faster than the fastest hard drives—sometimes much faster. (In fact, the fastest SSDs can saturate a SATA 3.0 Gb/s interface, which is the real reason that motherboard makers are beginning to introduce models with SATA 6.0 Gb/s ports.)

SSDs are simultaneously reliable and unreliable: reliable in the sense that they have no moving parts and are much less subject to shock damage than rotating drives, and unreliable in the sense that SSDs are inherently consumable items. We know the idea of “using up” an SSD sounds odd, but that is in fact what happens. The nature of flash memory is that the individual cells used to store data survive for only so many write cycles. Once a cell exceeds that number of writes, it is no longer usable.

Manufacturers implement various firmware schemes to load-level their drives, and recent operating systems (such as current Linux distributions and Windows 7) include support for TRIM, a technology that maximizes the usable life of an SSD.

But, inevitably, a heavily used SSD begins to lose both capacity and performance as time passes. Used improperly, an SSD may begin to show performance degradation after only a few weeks or even days of use. With TRIM enabled, that same drive will provide excellent performance for months or even years of use.

It’s difficult for us to write authoritatively about SSDs, because right now the entire market segment is in a state of flux. Better drives are released literally every month, and strides are being made in improving both performance and sustainability. If you buy an SSD now, just be aware that you’re buying into a rapidly developing technology. Anything you buy now will inevitably be replaced by a better, larger, faster, and more durable model for the same price soon after you buy it.

So, is buying an SSD now a sucker bet? Not at all. The best current models are blazingly fast and quite durable, assuming they’re not abused. Although an SSD is often the only drive in a notebook system, we think the best way to use SSDs in a desktop system is as a boot/system/working data drive in a system that also has one or more standard hard drives for bulk data storage.

If you browse the SSD section of an online component vendor’s site, the first thing you’ll notice is that SSDs have small capacities and high prices. At a time when 1 TB desktop hard drives are midrange in capacity and sell for well under $100, a typical SSD stores from 32 GB to 256 GB and sells for anywhere from 4 to 10 times as much.

Other than price, there are several considerations when choosing an SSD:

Capacity

All other things being equal, SSDs typically sell for about the same amount per GB across the line. For example, if a particular model of SSD is available in 32, 64, 128, and 256 GB capacities, each step up typically doubles the price. That’s because the vast bulk of the cost of an SSD is in the flash memory. Doubling capacity requires twice as much flash memory, and so doubles the price.

We think the best strategy is first to determine how much space you need for your operating system, applications, and working data, then to add a bit extra for growth and buy an SSD of that capacity. When we did that, we settled on a 128 GB model for Robert’s primary system, mainly because he does a lot of video editing, which involves manipulating many multi-gigabyte files. For most people, a 32 GB SSD will suffice.

Cell type

Two types of flash memory are used in SSDs. Single-level cell (SLC) memory is faster and can sustain many more write cycles than multi-level cell (MLC) memory, but it also costs much more. SLC memory is widely used in enterprise-class SSDs, some of which can store 1 TB or more and cost as much as a good used car. A few premium consumer-grade SSDs also use SLC memory. They are generally of very low capacity, and cost about four times as much per GB as drives of comparable capacity that use MLC memory. For most people, the additional speed of an SLC drive will be much less important than the higher price. A good MLC drive is so much faster than any hard drive that the faster-still SLC drive will seem a minor improvement.

Memory speed

There are large differences in speed between the least expensive and most expensive MLC SSDs, because fast flash memory chips cost more than slow ones. The least expensive SSDs use very slow flash memory chips, and may have read and write data transfer rates that are slower than a standard desktop hard drive. A similar SSD that uses the fastest available flash memory chips may have data transfer rates two or three times higher than the fastest hard drives.

Recommended SSD brands

SSDs have been available for years as niche products in the laptop/notebook market, so as they become increasingly mainstream there’s some history to judge them by. It would be useless for us to recommend specific models, because that information would be out of date long before you saw it. We can, however, recommend two vendors who have a history of producing SSDs that have been very competitive on price, quality, and performance. The first, unsurprisingly, is Intel (http://www.intel.com), which over the years has produced a series of excellent SSDs. We have no doubt that whichever models Intel offers when you’re ready to buy your own SSD will be equally competitive. Our second recommended manufacturer is Crucial (http://www.crucial.com), a company that knows as much about memory as any other on the planet. We suggest you start your search with what those two companies have to offer and use them as standards to which to compare competing offerings.

Optical Drive

With the exception of Blu-ray drives, optical drives are a mature product category and have become commoditized. Hobbyists continue to debate their finer points, but the truth is that one $20 DVD-ROM drive or $30 DVD writer is much like any other.

Every system needs at least one optical drive, if only for loading software, but for most desktop systems the only decisions you need to make are whether you need Blu-ray support and whether to buy a read-only drive or a writer (burner). For some small systems, the physical dimensions of the drive may be critical. Many small cases accept standard optical drives, but Mini-ITX and other very small cases may require a slimline optical drive and may limit the depth of the drive.

Unless there’s good reason to do otherwise, install a DVD writer in any system you build. Saving $10 by installing a read-only DVD-ROM drive instead of a DVD writer often turns out to be a poor economy, given the additional flexibility the writer provides.

If you need Blu-ray support, you can choose among three types of drives:

Blu-ray writers

Blu-ray writers can read and write Blu-ray discs, DVDs, and CDs. Write speeds for CD-R/RW and DVD±R/RW discs are typically a bit slower than those of standard DVD writers. Blu-ray writers can write to single-layer (25 GB) or dual-layer (50 GB) write-once (BD-R) or rewritable (BD-RE) discs. Write speeds for BD-R discs are typically 8X, 10X, or 12X, which means it takes about 8 to 12 minutes to write a full BD-R/SL disc and twice that for a BD-R/DL disc. BD-RE discs are written at 2X, which means it may take about 50 minutes to write a full BD-RE/SL disc, and twice that for a BD-RE/DL disc.

Blu-ray writers are still niche products, primarily because the drives and discs are expensive relative to DVD writers and media. On a per-byte basis, BD-R discs are nearing price parity with DVD±R discs, which is to say that a BD-R/SL disc has roughly six times the capacity of a DVD±R/SL disc and costs roughly six times as much. BD-RE/SL discs are considerably more expensive, but may make sense in some applications. BD-R/DL and BD-RE/DL discs are extraordinarily expensive and should be used only if there is no alternative.

If you install a Blu-ray writer, we recommend that you also install a standard DVD writer. Use the expensive Blu-ray writer only to read or write Blu-ray discs, and use the cheap DVD writer for reading and writing standard DVDs and CDs, to save wear and tear on the expensive drive. In fact, we might go further and install three drives in the system: a Blu-ray writer that we’d use only for burning BD-R/RE discs; a Blu-ray reader that we’d use for reading Blu-ray discs; and a standard DVD writer that we’d use for everything else.

Blu-ray combo drives

Blu-ray combo drives read Blu-ray discs, DVDs, and CDs, and can write DVDs and CDs (but not Blu-ray discs). Installing a Blu-ray combo drive makes no sense unless you happen to need this particular combination of features in a system that has room for only one optical drive, or you’re short of ports. Otherwise, install two drives: a standard DVD writer and a read-only BD-ROM drive. The total cost will be the same or less, the individual drives are likely to have better specifications than the combo drive, and if the reader outlives the writer (as is likely) you can replace just the $30 writer when it fails instead of replacing the $100 combo drive.

BD-ROM drives

BD-ROM drives read Blu-ray discs and standard DVDs, but cannot write discs. Because of higher component costs and licensing fees, BD-ROM drives typically cost three to four times as much as DVD-ROM drives. The least expensive models typically have 4X BD-ROM read speeds and 2 MB caches, which suffices for watching or ripping movies. More expensive models typically offer 8X, 10X, or 12X BD-ROM reads and 4 MB or larger caches, but unless you need additional read speed there’s little reason to spend more money.

Finally, a word about interfaces. Every type of optical drive is available in external form. All external optical drives provide a USB 2.0 interface, and some also provide an IEEE 1394 (FireWire) and/or eSATA interface. Internal optical drives are available in old-style ATAPI and SATA interfaces, with the exception of Blu-ray drives, which are nearly all SATA.

Ordinarily, this isn’t an issue. ATAPI is obsolescent going on obsolete, so it’s usually a no-brainer to choose a SATA optical drive. The one exception is if you’re building a system around a tiny motherboard that has a limited number of SATA interfaces and an ATA interface that would otherwise go unused. In that case, you may need all of the SATA ports for hard drives, so you may have no choice but to install an ATAPI optical drive. If you do that, make sure to use a UDMA ATA cable rather than a standard cable.

Recommended optical drive brands

Years ago, we recommended Plextor optical drives. They were built like tanks and (given high-quality blanks) produced perfect burns every time. As an experiment, we once burned through an entire spindle of 100 DVD+R discs, one after the other, and scanned the resulting discs. The first disc we burned was nearly perfect, as we expected. But the hundredth disc—written when the Plextor drive had been burning discs continuously for hours—was also nearly perfect, which we didn’t expect.

No other drive we torture-tested this way came even close to the Plextor. Most started producing bad discs after they’d burned only a few discs at this 100% duty cycle. More than a few simply died partway into the spindle.

Alas, those days are long gone. Plextor still sells competent optical drives, but the optical drive market has changed dramatically. When Plextor sold $200+ DVD burners in competition with lesser brands that sold for a third of that, there was room in the budget to build those drives to an extremely high standard. Nowadays, no one would buy a $200 DVD burner. Plextor now has to compete with $25 units and price its drives at $40 or $50, so there’s simply no longer anything but mediocre DVD burners available.

For standard optical drives, we recommend ASUS (http://usa.asus.com), LiteOn (http://us.liteonit.com), and Samsung (http://www.samsung.com). For Blu-ray burners, we recommend Pioneer (http://www.pioneerelectronics.com).

Backup Hardware

With typical desktop hard drives storing 1 TB or more, the thought of backing up a computer can be pretty intimidating. Maintaining proper backups is not quick, easy, or cheap. Nor is it optional, at least if you care at all about your data. Because it’s so much trouble to maintain proper backups, most people’s solution to protecting their data is simply to ignore the problem and hope nothing bad ever happens. Inevitably, they’re disappointed.

Like anything with moving parts, any hard drive will eventually fail. If you’re lucky, you’ll get at least some warning when your hard drive is about to fail. But don’t count on that. All too often, hard drives fail like light bulbs—fine one moment and dead the next. If all of your data is on that hard drive and you have no backups, you might as well kiss your data goodbye. Yes, it’s possible that some of your data may be recoverable, but—as many people have learned to their regret—that recovery may come at a cost of hundreds or even thousands of dollars.

The only solution is to back up your system regularly and maintain multiple copies of your data, including at least one off-site copy. Then, when the worst happens, you’ll have lost only the data you created or modified since your last backup.

Developing and implementing a backup strategy is beyond the scope of this book, but whatever backup procedure you use will obviously require hardware to make the copies. Disregarding tape drives, which are too expensive for a home or small office system, there are several practical alternatives.

Optical discs

A DVD+R disc stores about 4.4 GB, which for many people is sufficient for routine backups, and requires only a few minutes to write and verify. The disc costs only $0.25 or so and is small, light, and portable. Disc wallets that hold from 16 to 100 or more discs are readily available, and you can carry the wallet with you when you leave the house.

Obviously DVD+R isn’t a workable solution for a full backup of a large hard drive, which might require hundreds of discs, but it offers more than enough capacity to store most people’s primary working data, such as email, documents, digital photographs, and so on. We use DVD+R discs as a supplement, writing all of our working data to one every Sunday.

USB flash drives

USB flash drives are another popular choice for backing up data. Their primary advantages are small size and—because they have no moving parts—high durability. Robert once overlooked a USB flash drive in the watch pocket of his jeans. After being run through the washing machine and dryer, it worked as well as ever. In an even more extreme case, one of our readers reported accidentally running over a USB flash drive with his car. The case was crushed, but he was able to plug in the drive and read the data from it. True grit.

You have to make a trade-off when you choose a USB flash drive: capacity versus speed versus cost. Pick any two. Current models range from 1 GB to 256 GB in capacity. The fastest USB flash drives use memory fast enough to sustain USB 2.0 transfer rates (about 25 to 30 MB/s) for both reads and writes. The slowest have real-world write speeds of only 2 or 3 MB/s. And the price of a USB flash drive correlates directly to both capacity and speed.

Ideally, you’d probably like a USB flash drive with huge capacity and stunning speed. Alas, the price of such a drive would be more than you’re likely to want to pay. For example, Kingston offers a 256 GB flash drive (rated at only 12 MB/s for writes) that sells for $850. Economic realities mandate that the fastest flash drives are available only in small capacities—typically 1 to 4 GB—while the largest flash drives are available only with pedestrian write speeds.

We find the lower write speeds of high-capacity flash drives acceptable, because we use flash drives in the same way we use DVD+R discs: for quick daily or weekly backups of our working data sets. For example, it takes 6 or 7 minutes to write and verify a 4 GB backup set to a DVD, which corresponds closely to a USB flash drive with a write speed of about 10 MB/s. In effect, the USB flash drive is equivalent to a miniature stack of DVDs.

Several manufacturers offer USB flash drives that are specifically designed to back up Windows (and, usually, Mac) systems. These drives contain special backup software that backs up changed files automatically. You just plug the drive into a USB port, and it automatically runs the backup.

Among those, our favorite is the new Lexar Echo SE Backup Drive (http://www.lexar.com/echo/echo_se.html), which is available in 16, 32, 64, and 128 GB models. Lexar rates these drives as supporting “up to” 10 MB/s writes, which raised a red flag for us. Ordinarily, the rated write speed of a USB flash drive is unrealistically high because it’s achievable only when writing a single large file. When writing many small files, the actual write speed plummets, often to half or less the rated speed.

We tested a 32 GB Lexar Echo SE by writing a data set of 1,363 files and directories totaling 4,100 MB. The write completed in 6:42 (402 seconds), for a sustained data transfer rate of about 10.2 MB/s, better than advertised. We next wrote a single 4,100 MB compressed archive file, expecting even faster results. For the first few hundred MB, the Echo SE indeed ran faster, at about 13 MB/s. However, it completed the 4,100 MB write in 6:39 (399 seconds), just three seconds faster than the same-sized collection of small files, and with an identical write speed. Lexar has obviously done something effective to optimize this drive for real-world backups of many smaller files, and we recommend the Lexar Echo SE Backup drive without qualification.

External hard drives

As useful as DVD+R discs and USB flash drives are for routine partial backups, they have neither the capacity nor the speed for full backups of a system with a large hard drive or drives. For that, the only practical solution is to back up to another hard drive.

The mainstream solution is to use external hard drives, which are available from numerous manufacturers—including drive manufacturers Hitachi, Samsung, Seagate, Toshiba, and Western Digital, as well as many other vendors—in capacities from 20 GB to 3 TB. The smallest units use 2.5” 5,400 RPM notebook hard drives, are powered by the USB cable, and fit in your shirt pocket. The highest-capacity units use 5,400 or 7,200 RPM desktop hard drives, and are still small enough to be easily portable.

If you decide to use an external hard drive (or, better, multiple drives) for backing up, pay close attention to the interface(s) it offers. Units are available for USB 2.0, USB 3.0, FireWire, eSATA, and various combinations thereof. The vast majority of external hard drives currently in use are connected to USB 2.0 ports, which is fine if you have only a few GB to back up but is much too slow for doing routine backups of large data sets. If you routinely do backups of 100 GB or more, an eSATA or USB 3.0 model is a much better choice because the data-transfer rate is limited only by the speed of the hard drives.

Most external hard drives include backup software for Windows. The first time you use the drive, you simply plug it in, power it up, and allow the software to install on Windows. The software does a full backup of your Windows system. (If you’re using USB 2.0, this initial backup may well take overnight to finish.) When the backup completes, you can shut down the drive, disconnect it, and take it with you when you leave the house. When you reconnect the drive to your system, it automatically scans for new and changed files and does a quick differential backup of only those files. Most external drives can be used with more than one computer to maintain separate backups for each.

Among the many USB 2.0 external drives available, we recommend the Seagate Free Agent series (http://www.seagate.com). They’re available in capacities from 240 GB to 3 TB, and the bundled software is excellent. If you have a USB 3.0 port available and need a fast, high-capacity external drive, choose one of the Western Digital My Book models (http://www.wdc.com).

But, before you buy an external drive, read on.

Removable hard drive bays

External hard drives are convenient, but the problem is that you really need at least two or three of them to maintain an adequate set of backups. Each time you buy an external hard drive, you’re paying for the external enclosure and the backup software. If you can get along with only two or three external drives, that cost premium is not outrageous. But if you need more, the cost starts to mount up. (Robert, for example, routinely uses more than a dozen external drives, both for backups and for archiving data off our disk arrays and into deep storage.)

One good solution to this problem is to install a removable hard drive bay in your system. Such bays fit into an external 5.25” drive bay and are connected to a standard SATA data cable and power cable. The front of the removable drive bay has an opening that accepts a standard 3.5” desktop hard drive, which you can just slide in and out as needed. (You will have to enable AHCI in your BIOS and operating system if you want to hot-swap drives.)

Removable hard drive bays have several advantages. They accept standard desktop hard drives, which are inexpensive (relative to external drives) and appear to the computer as just another hard drive. Because the drives use the SATA interface, they are as fast as an internal drive. Because the removable drive bay accepts any standard SATA hard drive, you can recycle older hard drives that you’ve retired when upgrading a system. And, when you leave home, it’s much easier to carry a bare hard drive in a plastic shell or wallet than it is to disconnect and haul an external hard drive around with you. (Be careful when removing a hard drive; depending on how hot-running the drive is and how long it’s been running, it may be too hot to touch.)

We’ve used several models of removable drive bays, but our favorite is the Antec Easy SATA (http://www.antec.com). It just works, and at $20 or so, there’s no reason not to install one.

Hard drive docking stations

Hard drive docking stations are similar conceptually to removable hard drive bays, but they are external devices rather than internal. Like removable drive bays, docking stations use the SATA/eSATA interface, and so support hot-swapping.

A typical hard drive docking station holds one or two 3.5” desktop hard drives or 2.5” notebook hard drives vertically in a plastic or metal base, is powered by a wall wart, and connects to the computer with an eSATA cable(s). The hard drives are exposed and can be inserted or removed simply by sliding them into or out of the docking station. (Once again, for hot-swapping you’ll need to enable AHCI in your BIOS and operating system.)

The disadvantage of a hard drive docking station versus a removable drive bay is that the docking station adds desktop clutter. The advantage is that the docking station is the most cost-efficient way of using removable hard drives. For example, the hard drive docking station we recommend, the SYBA SD-ENC50020 (http://www.syba.com), sells for about $25 and holds two hard drives simultaneously.

Video Adapters

The video adapter, also called a graphics adapter, renders video data provided by the processor into a form that the monitor can display. Many motherboards include embedded (integrated) video adapters. You can also install a standalone video adapter, also called a video card or graphics card, in a PCI Express expansion slot. Keep the following in mind when choosing a video adapter:

• Unless you run graphics-intensive games, 3D graphics performance is unimportant. Any recent video adapter is more than fast enough for business applications and casual gaming. Even integrated video, let alone any current standalone video adapter, is fast enough to support the 3D requirements of the Windows Aero Glass interface.
• Choose integrated video unless there is good reason not to. Integrated video adds little or nothing to the price of a motherboard, and generally suffices for anyone except hardcore gamers or those with other special video requirements. Make sure any motherboard you buy allows integrated video to be disabled and provides a PCI Express slot, though. That way, you can upgrade the video later if you need to by installing a separate video card.
• Make sure that the video adapter you choose (or the integrated video on your motherboard) provides the type of video output connector you need. Nearly all displays provide a 15-pin analog (DSUB) VGA connector. All but the least expensive displays also provide a DVI connector, which may be analog, digital, or hybrid. Some displays provide an HDMI connector, which may or may not be HDCP-enabled.
• If you plan to use dual displays, make sure that your integrated video or video adapter supports dual displays, and that it provides the type of video connectors you need for both displays. Note that some video adapters provide one analog and one digital video connector, but allow only one of those to operate at a time. The most flexible choice is a card with dual DVI-I hybrid video connectors, which support both analog and digital displays, or dual HDMI connectors.
• If you need a 3D graphics adapter, don’t overbuy. A $400 video adapter is faster than a $100 adapter, but nowhere near four times faster. As with other PC components, the bang-for-the-buck ratio drops quickly as the price climbs. If you need better 3D graphics performance than integrated video provides but you don’t have much in the budget for a video adapter, look at “obsolescent” 3D video adapters—those a generation or two out of date. If you buy an older adapter, make sure the level of DirectX it supports is high enough to support the games you play.
• Make sure that the adapter you choose has drivers available for the operating system you intend to use. This is particularly important if you run Linux, BSD, or another OS with limited driver support.
• In any case, but particularly if you will use a large, high-resolution display, make certain the video adapter and its drivers support the native resolution of the display. For example, if the highest resolution supported by your video adapter is 1920×1080, you will not be pleased with the image quality from that adapter on your expensive new 30” 2560×1600 display.
• If you want to use dual displays, it simplifies matters to install a video card that has two identical video outputs (preferably digital) and connect it to two identical displays. It’s often possible to configure a mixed environment—for example, one integrated video output and one video output from a separate video adapter driving two dissimilar displays—but doing so makes configuring the system more complex.

Recommended video adapter brands

Our first recommendation is to use integrated video unless you have good reason to do otherwise. ATI, Intel, and NVIDIA integrated video are all excellent, and more than sufficient for most purposes.

If you need a standalone video adapter, consider models from ASUS (http://usa.asus.com), EVGA (http://www.evga.com), GIGABYTE (http://www.gigabyte-usa.com), Sapphire (http://www.sapphiretech.com), VisionTek (http://www.visiontek.com), or XFX (http://www.xfxforce.com). All of those companies build video adapters around chipsets produced by ATI or NVIDIA, or both. The video adapter market is extremely competitive, so price is usually an excellent guideline to performance, features, and quality.

If you simply need a bit more 3D horsepower than integrated video provides, look for an adapter in the $25 to $50 range from any of these manufacturers, using either an ATI or NVIDIA chipset. The $50 to $100 range covers entry-level 3D gaming adapters, the fastest of which are sufficient for all but the most intense 3D games. If you’re seriously interested in 3D gaming, you probably know more than we do about gaming graphics adapters. If not, spend some time on AnandTech (http://www.anandtech.com) and Tom’s Hardware (http://www.tomshardware.com) before you plunk down money for a video adapter.

Display

The good news is that nowadays buying a bad display takes some serious effort. If you stick to good brand names, the chances that you’ll get a bad display are very close to zero. The display business is extremely competitive, and bad vendors simply don’t stay in business for long. The competitive nature of the business also means you don’t have to worry much about over-paying for a display. Doing a bit of price shopping for the models you’re interested in just about guarantees you’ll get a good display at a good price.

Here’s what to look for when you shop for a display:

Resolution

Resolution is the number of pixels (dots) that make up the image, specified as the number of columns by the number of rows. For example, a display whose resolution is listed as 1920×1080 uses a rectangular matrix of 1,920 columns in 1,080 rows. Such a display has a total of 1,920×1,080 pixels, or 2,073,600 pixels. Resolution of current desktop displays ranges from 1024×768 to 2560×1600. All other things being equal, higher resolution is better because at any given screen size it produces a sharper image and finer detail.

Aspect ratio

The aspect ratio of a display is calculated simply by dividing the number of columns by the number of rows. For example, a 1024×768 display has an aspect ratio of 1024/768 = 1.33:1 and a 2560×1600 display has an aspect ratio of 2560/1600 = 1.6:1. Rather than using such fractional numbers, aspect ratios are usually converted to whole numbers by multiplying both sides by an appropriate factor. For example, the 1.33:1 aspect ratio is normally multiplied by 3 to give a stated aspect ratio of 4:3, and the 1.6:1 aspect ratio is normally multiplied by 10 to give a stated aspect ratio of 16:10. (Of course, it could also be stated as 8:5 with equal accuracy, but apparently marketers like larger numbers.)

Until a few years ago, most computer displays used aspect ratios similar to those of older television sets, typically 4:3 or 5:4. Such displays are often referred to as “square,” even though they’re really not. However, they are nearly as tall as they are wide, in contrast to modern widescreen displays, most of which are, like modern televisions, much wider than they are tall. Although some people prefer “square” displays, the market is shifting quickly to widescreen models. As of mid-2009, the number of widescreen displays available outnumbered square displays by about two to one, and as of mid-2010 that ratio had increased to about three to one. “Square” displays are a dying breed.

Screen size

Screen size is simply the diagonal measurement of the screen. Current desktop displays range from 17” to 30” diagonally. There are two schools of thought about screen size. One argues that you should buy the largest, highest-quality display that you can afford and have space for. The other argues that two or more smaller displays are more useful than a single larger display. If you choose the latter route, make sure that your video adapter and drives support multiple displays.

All three of these characteristics are interrelated in real-world displays. Various standards—official, unofficial, and semi-official—exist that tie these characteristics together. Table 2-2 lists important desktop display standards in order of increasing total display resolution.

Table 2-2. Desktop display standards
Name	Display size and type	Resolution	AR	Megapixels
WXGA+	17” and 19” widescreen	1440×900	16:10	1.296
SXGA	17” and 19” square	1280×1024	5:4	1.310
“900P”	20” widescreen	1600×900	16:9	1.440
WSXGA+	22” widescreen	1680×1050	16:10	1.764
UXGA	20” and 21” square	1600×1200	4:3	1.920
1080P	21” through 27” widescreen	1920×1080	16:9	2.073
WUXGA	23” through 28” widescreen	1920×1200	16:10	2.304
WQXGA	30” widescreen	2560×1600	16:10	4.096

Beyond the basics, here are some other factors to consider:

Interface(s)

Displays may provide one or several interfaces. A few inexpensive models provide only a DB-15 analog (VGA) connector. Most displays provide both a DB-15 and a Digital Visual Interface (DVI) connector, which may be analog-only (DVI-A), digital-only (DVI-D), or hybrid analog/digital (DVI-I).

Some displays provide a High-Definition Multimedia Interface (HDMI), which is compatible with DVI-D but also includes the audio signal on the same connector. If your video adapter has a DVI-D or DVI-I connector and your display has an HDMI connector, or vice versa, you can purchase an adapter cable that will allow the two to function together.

If you intend to watch copy-protected content (such as Blu-ray movies) on your display, make sure at least one of the DVI or HDMI connectors includes High-bandwidth Digital Content Protection (HDCP) support.

Brightness and contrast

A bright display is important, particularly if your viewing environment is well lit, and also to give you some reserve as the display backlighting gradually dims with usage. Brightness is specified in nits (candellas/square meter, or cd/m2). The least-expensive displays may be rated at only 200 nits. Better displays are generally rated at 250 to 300 nits, and a few high-end or specialty displays may be rated at 400 nits or more. We consider a display rated at 250 nits or more to be sufficient.

Contrast specifies the ratio between the luminances of the brightest white and darkest black the monitor can display, and is determined by panel type and quality as well as backlight intensity. A monitor with high contrast can display very dark shadows and very bright highlights simultaneously and shows gradation in both the shadows and the highlights. Inexpensive displays have contrast ratios as small as 300:1, and some high-end or specialty displays are rated as high as 10,000,000:1.

Unfortunately, contrast specifications are essentially useless. Different display makers use different methods to calculate them. Two comparable displays from different makers may have identical actual contrasts, and yet one manufacturer may rate its display at 500:1 and the other at 1,000:1. As far as we know, each manufacturer is consistent. For example, if one ASUS display is rated at 500:1 and another at 1,000:1, it’s safe to assume that the latter display has superior contrast.

Response time

Fast response time is important to avoid smearing and other artifacts during fast-motion video segments, particularly with 3D computer games. Typical displays have 4 millisecond (ms) or 5 ms response times, which are sufficient for anything other than extreme gaming. Premium models and those targeted at gamers typically have 2 ms to 3 ms response times, with a few models at less than 2 ms. Inexpensive displays and very large displays may have slower response times, typically 6 ms to 10 ms, but sometimes 20 ms or more. For general use, we consider a response time of 10 ms or less adequate. If you watch a lot of HD video, look for a model rated at 5 ms or less. If you’re heavily into extreme gaming, look for a model rated at 3 ms or less.

Backlighting

LCD displays backlight the image one of two ways:

CCRT

Most displays use cold cathode ray tube (CCRT) illumination. CCRTs are similar to fluorescent tubes and are subject to failure and to gradual dimming over time.

LED

Some premium displays substitute white LEDs for the CCRTs. Entry-level LED models—which are still more expensive than CCRT models—are edge-lit, with LEDs along the edges of the panel. More expensive full-field LED displays use a large numbers of LEDs behind the entire surface of the panel.

LED backlighting has several advantages. First, LEDs do not burn out or dim noticeably over time (a typical LED might dim to 50% of its original brightness after 50 years of constant use). Second, LEDs can be significantly brighter than CCRTs, which allows higher contrast and a brighter image. Third, full-field LED displays can greatly increase the dynamic contrast range of the image by selectively illuminating only some of the LEDs in the array. In brighter image areas, all LEDs are lit; in dimmer areas, only a fraction of the LEDs are lit.

We consider CCRT displays acceptable for almost any purpose. Rather than spending money on LED backlighting, we’d spend it on a larger or higher-resolution CCRT display—for now, at least. Currently, about 10% of all LCD displays use LED backlighting. As the technology becomes less expensive, we expect it to replace CCRTs entirely.

Minor features

We suggest you choose your display based on the factors we’ve already discussed. If those aren’t sufficient to narrow down your choice to one display, it’s time to start looking at less important factors. Taken individually, all of these “extra” features may be minor, but taken as a group they may be enough to swing your decision to one particular model.

Some displays include built-in speakers, which are at best mediocre but may be good enough for your purposes (if only to reduce desktop clutter). Some displays also include a built-in microphone that’s useful for Skype and similar VoIP services. Although it is less common than it was a few years ago, a few displays include a built-in webcam, which is useful for Skype or recording YouTube videos. Finally, to reduce desktop clutter, a few displays include a built-in USB 2.0 hub. Note that some of these are passive hubs, and may not have sufficient power for many USB devices. Others are powered USB hubs and work just the same as any other USB hub.

Recommended display brands

Our opinion, confirmed by our readers and colleagues, is that ASUS (http://usa.asus.com), NEC (http://www.necdisplay.com), Samsung (http://www.samsung.com), and ViewSonic (http://www.viewsonic.com) make the best displays available. Their displays, particularly midrange and better models, provide excellent image quality and are quite reliable. You’re likely to be happy with a display from any of these first-tier manufacturers. One up-and-coming display manufacturer that we’ve heard lots of good things about is Hanns.G (http://www.hannsg.com/us). Its displays are generally a bit less expensive than comparable models from the top-tier makers, but have competitive specifications. Most Hanns-G owners we’ve spoken with are extremely pleased with their displays.

Audio Adapter

Audio adapters, also called sound cards, are almost a dead breed. Integrated audio is more than good enough for nearly any purpose. Consider buying a separate audio adapter only if you are building an extreme gaming rig, a media center/home theater system that will connect to a professional-level sound system, or a professional audio/video-production system. Even for those purposes, chances are you’ll find integrated audio sufficient, so try it before you buy a separate audio adapter.

Internal audio adapters are available to fit PCI slots or PCI Express slots. Before you buy an internal audio adapter, make sure you have an available slot of the proper type. External audio adapters, intended primarily for notebook systems, connect to the computer via USB.

Recommended audio adapter brands

Unless you have good reason to do otherwise, use the integrated motherboard audio. For extreme gaming, choose a Creative Labs (http://us.creative.com) Xi-Fi card or an Auzentech (http://www.auzentech.com) Xi-Fi card with the features and specifications you need. Some gamers and media center/home theater system–builders prefer ASUS (http://usa.asus.com) Xonar series sound cards. If you’re an audiophile or do serious audio/video production, choose an M-AUDIO (http://www.m-audio.com) adapter with the features and specifications you need.

Speakers

Computer speakers span the range from $10 pairs of small satellites to $500+ sets of five speakers that are suitable for home theater systems. Personal preference is the most important factor in choosing speakers.

Speakers that render a Bach concerto superbly are often not the best choice for playing a first-person shooter game. For that matter, speakers that one person considers perfect for the Bach concerto (or the game), another person may consider mediocre at best. For that reason, we strongly suggest that you attempt to listen to speakers before you buy them, particularly if you’re buying an expensive set.

Speaker sets are designated by the total number of satellite speakers, followed by a period and a “1” if the set includes a subwoofer (also called a low-frequency emitter, or LFE). Speaker sets are available in 2.0 (front left/right satellites), 2.1 (front left/right satellites and a subwoofer), and 5.1 (front/back and left/right satellites, a front-center satellite, and a subwoofer) configurations.

Other than configuration, the main specification for a speaker set is the wattage of the various speakers. That value should be provided as RMS wattage but is sometimes specified as the much higher peak wattage value. Don’t fall prey to wattage overkill. Just a few watts per satellite is sufficient to produce uncomfortably loud sound, and a 100W subwoofer can literally rattle the walls.

The advantage of high wattage ratings is that such speakers have a wide dynamic range with full sound quality. A high-wattage speaker may typically actually use only a few watts, but when a very loud transient (such as a gunshot in a game or a crescendo in a concert) occurs, the high-power amplifier has sufficient power to accurately reproduce that sound at the proper high volume. A less expensive speaker with a lower-power amplifier has to strain to reproduce such transients, and will do so at lower sound quality and volume than the more capable speaker.

The price of a speaker set doesn’t necessarily correspond to the number of speakers in the set or their wattage. Both speaker (driver) quality and amplifier quality are important. There are inexpensive 5.1 speaker sets with relatively high wattage—which use cheap drivers and cheap amplifiers—and some low-wattage 2.0 or 2.1 sets that cost a bundle but provide top-notch sound.

We recommend that you decide on the number of speakers according to your needs and your budget. Buy a 5.1 set only for serious gaming or media center/home theater use; otherwise, buy the best 2.0 or 2.1 set you can afford. If you have $75 to spend, for example, you’re better off buying a good 2.1 speaker set than a cheesy 5.1 set. (Barbara uses a 2.0 M-AUDIO studio monitor set on her primary computer; Robert uses a 2.1 Logitech Z-series set on his.)

Recommended speaker brands

Altec-Lansing (http://www.alteclansing.com), Creative (http://us.creative.com), and Logitech (http://www.logitech.com) all offer speaker sets in all three configurations in a range of prices, from $15 2.0 sets to 5.1 sets that cost several hundred dollars. All of them offer excellent value in their price ranges, although we admit to a preference for Logitech. At the high end, M-AUDIO (http://www.m-audio.com) offers premium-quality 2.0 and 2.1 sets and has an excellent reputation with audiophiles.

Keyboards

The best keyboard is a matter of personal preference. A keyboard we really like, you may dislike intensely, and vice versa. Ultimately, your own preferences are the only guide.

Keyboards vary in obvious ways—layout, size, and style—and in subtle ways like key spacing, angle, dishing, travel, pressure required, and tactile feedback. People’s sensitivity to these differences varies. Some are keyboard agnostics who can sit down in front of a new keyboard and, regardless of layout or tactile response, be up to speed in a few minutes. Others have strong preferences about layout and feel. If you’ve never met a keyboard you didn’t like, you can disregard these issues and choose a keyboard based on other factors. If love and hate are words you apply to keyboards, use an identical keyboard for at least an hour before you buy one for yourself.

That said, here are several important characteristics to consider when you choose a keyboard:

• Keyboards are available in two styles, the older straight keyboard and the modern ergonomic style. Some people strongly prefer one or the other. Others don’t care. If you’ve never used an ergonomic keyboard, give one a try before you buy your next keyboard. You may hate it—everyone does at first—but then again, after you’ve used it for an hour or so you may decide you love it.
• The position of the alphanumeric keys is standard on all keyboards other than those that use the oddball Dvorak layout. What varies, sometimes dramatically, is the placement, size, and shape of other keys, such as shift keys (Shift, Ctrl, and Alt), function keys (which may be across the top, down the left side, or both), and cursor control and numeric keypad keys. If you are used to a particular layout, purchasing a keyboard with a similar layout makes it easier to adapt to the new keyboard.
• Most current keyboards use the USB interface natively and are supplied with an adapter for those who need to connect them to a PS/2 keyboard port. We use mostly USB keyboards, but it’s a good idea to have at least one PS/2 keyboard available (or a PS/2 adapter) for those times when Windows shoots craps and won’t recognize USB devices.
• Some keyboards provide dedicated and/or programmable function keys to automate such things as firing up your browser or email client, or to allow you to define custom macros that can be invoked with a single keystroke. These functions are typically not built into the keyboard itself, but require loading a driver. To take advantage of those functions, make sure a driver is available for the OS you use.
• The weight of a keyboard can be a significant issue for some people. The lightest keyboard we’ve seen weighed just over a pound, and the heaviest nearly eight pounds. If your keyboard stays on your desktop, a heavy keyboard is less likely to slide around. Conversely, a heavy keyboard may be uncomfortable if you work with the keyboard in your lap.
• Some manufacturers produce keyboards with speakers, scanners, and other entirely unrelated functions built in. These functions are often clumsy to use, fragile, and have limited features. If you want speakers or a scanner, buy speakers or a scanner. Don’t get a keyboard with them built in.
• Wireless keyboards are ideal for presentations, TV-based web browsing, or just for working with the keyboard in your lap. Wireless keyboards use a receiver module that connects to a USB port or the PS/2 keyboard port on the PC. The keyboard and receiver communicate using either radio frequency (RF) or infrared (IR). IR keyboards require a direct line of sight between the keyboard and receiver, while RF keyboards do not. Most IR keyboards and many RF keyboards provide limited range—as little as five feet or so—which limits their utility to working around a desk without cables tangling. Any wireless keyboard you buy should use standard AA, AAA, or 9V alkaline or NiMH batteries rather than a proprietary battery pack.

Recommended keyboard brands

Logitech (http://www.logitech.com) and Microsoft (http://www.microsoft.com) both produce a wide range of excellent keyboards, one of which is almost certainly right for you. Even their basic models are well built and reliable. The more expensive models add features such as RF or Bluetooth wireless connectivity, programmable function keys, and so on.

Mice

Choosing a mouse is much like choosing a keyboard. Personal preference is by far the most important consideration. If possible, try a mouse before you buy it.

Use the following guidelines when choosing a mouse:

• Mice are available in various sizes and shapes, including small mice intended for children, notebook-sized mice, the formerly standard “Dove bar” size, the mainstream ergonomic mouse, and some oversized mice that have many buttons and extra features. Most people find standard-sized mice comfortable to use for short periods, but if you use a mouse for longer periods small differences in size and shape often make a big difference in comfort and usability. Although oversized mice provide attractive features and functions, people with small hands may find such mice too large to use comfortably. Pay particular attention to mouse shape if you are left-handed. Although asymmetric ergonomic mice are often claimed to be equally usable by left- and right-handers, many lefties find them uncomfortable and resort to right-handed mousing. Some manufacturers, including Logitech, produce symmetric ergonomic mice.

• Get a wheel mouse. Although some applications do not support the wheel, those that do are the ones most people are likely to use a great deal—Microsoft Office, Internet Explorer, Firefox, and so on. Using the wheel greatly improves mouse functionality by reducing the amount of mouse movement needed to navigate web pages and documents. Mice with a tilt-wheel allow you to scroll vertically and horizontally.
• Standard two-button mice (three, counting the wheel) suffice for most purposes. However, five-button mice are ideally suited to some applications, such as games and web browsing. For example, the two extra buttons can be mapped to the Back and Forward browser buttons, eliminating a great deal of extraneous mouse movement.
• Mice have cords ranging in length from less than 4 feet to about 9 feet. A short mouse cord may be too short to reach the system, particularly if it is on the floor. If you need a longer mouse cord, purchase a PS/2 keyboard or USB extension cable, available in nearly any computer store.
• Consider buying a cordless mouse. The absence of a cord can make a surprising difference.

Recommended mouse brands

Logitech (http://www.logitech.com) and Microsoft (http://www.microsoft.com) both produce a wide range of excellent optical mice, in corded and cordless models. One of them is almost certainly right for you. Even their basic models are well built and reliable. The more expensive models have more features, are more precise, and are probably more durable. We use both Logitech and Microsoft mice on our personal systems.

Buying Components

We’ve bought a boatload of PC components over the last 25 years, for ourselves and on behalf of employers and clients. In the following sections, we’ll tell you what we learned along the way.

Buying Guidelines

Until the early 1990s, most computer products were bought in computer specialty stores. Retail sales still make up a significant chunk of computer product sales—although the emphasis has shifted from computer specialty stores to local “big-box” retailers like Best Buy, Fry’s, Walmart, and Costco—but online resellers now account for a large percentage of PC component sales.

Should you buy from a local brick-and-mortar retailer or an online reseller? We do both, because each has advantages and disadvantages.

Local retailers offer the inestimable advantage of instant gratification. Unless you’re more patient than we are, when you want something, you want it Right Now. Buying from a local retailer puts the product in your hands instantly, instead of making you wait for FedEx to show up. You can also hold the product in your hands, something that’s not possible if you buy from an online reseller. Additionally, local retailers have a big advantage if you need to return or exchange a product: if something doesn’t work right, or if you simply change your mind, you can just drive back to the store rather than dealing with the hassles and cost of returning a product to an online reseller.

Online resellers have the advantage in breadth and depth of product selection. If you want the less expensive OEM version of a product, for example, chances are you won’t find it at local retailers, most of which stock only retail-boxed products. If an online reseller stocks a particular manufacturer’s products, it will tend to stock the entire product line, whereas local retailers often pick and choose only the most popular items in a product line. Of course, the popular products are usually popular for good reasons. Online resellers are also more likely to stock niche products and products from smaller manufacturers. Sometimes, if you must have a particular product, the only option is to buy it online.

Online resellers usually advertise lower prices than local retailers, but it’s a mistake to compare only nominal prices. When you buy from a local retailer, you pay only the advertised price plus any applicable sales tax. When you buy from an online retailer, you pay the advertised price plus shipping, which may end up costing you more than buying locally.

Although online resellers may have a lower overall price on a given component, it’s a mistake to assume that this is always the case. Local retailers frequently run sales and rebate promotions that cut the price of a component below the lowest online price. For example, we bought a spindle of 100 discs on sale from a local retailer for $19.95 with a $10 instant rebate and a $20 mail-in rebate. After the cost of the stamp to mail in the rebate form, they paid us $9.68 to carry away those 100 discs, which is pretty tough for an online reseller to match.

In particular, local retailers are usually the best place to buy heavy, bulky, or fragile items, such as monitors, cases, UPSs, and so on. Local retailers receive these items in pallet loads, which makes the cost of shipping an individual item almost nothing. Conversely, online resellers have to charge you, directly or indirectly, for the cost of getting that heavy item to your door.

Whether your purchase your PC components from a local brick-and-mortar store or a web-based retailer, here are some guidelines to keep in mind:

• Make sure you know exactly what you’re buying. For example, a hard drive may be available in two versions, each with the same or a similar model number but with an added letter or number to designate different amounts of cache. Or a hard drive maker may produce two models of the same size that differ in price and performance. Always compare using the exact manufacturer model numbers. Before you buy a product, research it on the manufacturer’s website and on the numerous independent websites devoted to reviews. We usually search Google with the product name and “review” in the search string.
• Vendors vary greatly. Some we trust implicitly, and others we wouldn’t order from on a bet. Some are always reliable, others always unreliable, and still others seem to vary with the phases of the moon. We check http://www.resellerratings.com, which maintains a database of customer-reported experiences with hundreds of vendors.
• The list price or suggested retail price (SRP) is meaningless. Most computer products sell for a fraction of the SRP, others sell for near the SRP, and for still others the manufacturer has no SRP but instead publishes an estimated selling price (ESP). To do meaningful comparisons, you need to know what different vendors charge for the product. Fortunately, there are many services that list what various vendors charge. We use http://www.pricescan.com, http://www.pricewatch.com, http://www.pricegrabber.com, and http://www.froogle.com. These services may list 20 or more different vendors, and the prices for a particular item may vary dramatically. We discard the top 25% and the bottom 25% and average the middle 50% to decide a reasonable price for the item.
• Many components are sold in retail-boxed and OEM forms. The core component is likely to be similar or identical in either case, but important details may vary. For example, processors are available in retail-boxed versions that include a CPU cooler and a three-year warranty. They are also available as OEM components (also called tray packaging or white box) that do not include the CPU cooler and have only a 90-day warranty. OEM items are not intended for retail distribution, so some manufacturers provide no warranty to individual purchasers. OEM components are fine, as long as you understand the differences and do not attempt to compare prices between retail-boxed and OEM.
• The market for PCs and components is incredibly competitive and margins are razor-thin. If a vendor advertises a component for much less than other vendors, it may be a “loss leader.” More likely, though, particularly if its prices on other items are similarly low, that vendor is cutting corners, whether by using your money to float inventory, by shipping returned products as new, by charging excessive shipping fees, or, in the ultimate case, by taking your money and not shipping the product. If you always buy from the vendor with the rock-bottom price, chances are you’ll waste a lot of time hassling with returns of defective, used, or discontinued items and dealing with your credit card company when the vendor fails to deliver at all. Ultimately, you’re likely to spend more money than you would have by buying from a reputable vendor in the first place.
• The actual price you pay may vary significantly from the advertised price. When you compare prices, include all charges, particularly shipping charges. Reputable vendors tell you exactly how much the total charges will be. Less reputable vendors may forget to mention shipping charges, which could be very high. Some vendors break out the full manufacturer pack into individual items. For example, if a retail-boxed hard drive includes mounting hardware, some vendors will quote a price for the bare drive without making it clear that they have removed the mounting hardware and charge separately for it. Also be careful when buying products that include a rebate from the maker. Some vendors quote the net price after the rebate without making it clear that they are doing so.
• Some vendors charge more for an item ordered via their 800 number than they do for the same item ordered directly from their website. Some others add a fixed processing fee to phone orders. These charges reflect the fact that taking orders on the Web is much cheaper than doing it by phone, so this practice has become common. In contrast, some of our favorite vendors do not even provide telephone order lines.
• It can be very expensive to ship heavy items such as UPSs and printers individually. This is one situation in which local big-box stores like Best Buy have an advantage over online vendors. The online vendor has to charge you for the cost of shipping, directly or indirectly, and that cost can amount to $50 or more for a heavy item that you need quickly. Conversely, the big-box stores receive inventory items by the truckload or even via railcar shipments, so the cost to them to have a single item delivered is quite small. They can pass that reduced cost on to buyers. If you’re buying a heavy item, don’t assume that it will be cheaper online. Check your local Best Buy or other big-box store: you may find that it actually costs less there, even after you pay sales tax. And you can carry it away with you instead of waiting for FedEx to show up with it.
• Most direct resellers are willing to sell for less than the price they advertise. All you need to do is tell your chosen vendor that you’d really rather buy from them, but not at the price they’re quoting. Use lower prices you find with the price comparison services as a wedge to get a better price. But remember that reputable vendors must charge more than the fly-by-night operations if they are to make a profit and stay in business. If we’re ordering by phone, we generally try to beat down our chosen vendor a bit on price, but we don’t expect them to match the rock-bottom prices that turn up on web searches. Of course, if you’re ordering from a web-only vendor, dickering is not an option, which is one reason why web-only vendors generally have better prices.
• Using a credit card puts the credit card company on your side if there is a problem with your order. If the vendor ships the wrong product, a defective product, or no product at all, you can invoke chargeback procedures to have the credit card company refund your money. Vendors who live and die on credit card orders cannot afford to annoy credit card companies, so they tend to resolve such problems quickly. Even your threat to request a chargeback may cause a recalcitrant vendor to see reason.
• Some vendors add a surcharge, typically 3%, to their advertised prices if you pay by credit card. Surcharges violate credit card company contracts, so some vendors instead offer a similar discount for paying cash, which amounts to the same thing. Processing credit card transactions costs money, and we’re sure that some such vendors are quite reputable, but our own experience with vendors that surcharge has not been good. We always suspect that they discourage using credit cards because their business practices result in a relatively high percentage of chargeback requests.
• Good vendors allow you to return a defective product for replacement or a full refund (often less shipping charges) within a stated period, typically 30 days. Buy only from such vendors. Nearly all vendors exclude some product categories, such as notebook computers, monitors, printers, and opened software, either because their contracts with the manufacturer require them to do so or because some buyers commonly abuse return periods for these items, treating them as “30-day free rentals.” Beware of the phrase “All sales are final.” That means exactly what it says.

  A Cunning Plan

  Nearly all retailers refuse to refund your money on opened software, DVDs, etc. and will only exchange the open product for a new, sealed copy of the same title. One of our readers told us how he gets around that common policy: he returns the open software in exchange for a new, sealed copy of the same product, keeping his original receipt. He then returns the new, sealed copy for a refund. That’s probably unethical and may be illegal for all we know, but it does work. Recently, though, some stores, including Best Buy, have begun annotating the original receipt when you make an exchange. Oh, well. It was too good to last.

• Check carefully for any mention of restocking fees. Many vendors who trumpet a “No questions asked money-back guarantee” mention only in the fine print that they won’t refund all your money—they charge a restocking fee on returns. We’ve seen fees as high as 30% of the purchase price. These vendors love returns, because they make a lot more money if you return the product than if you keep it. Do not buy from a vendor that charges restocking fees on exchanges (as opposed to refunds). For refunds, accept no restocking fee higher than 10% to 15%, depending on the price of the item.
• If you order by phone, don’t accept verbal promises. Insist that the reseller confirm your order in writing, including any special terms or conditions, before charging your credit card or shipping the product. If a reseller balks at providing written confirmation of its policies, terms, and conditions, find another vendor. Most are happy to do so. If you’re ordering from a vendor that uses web-based ordering exclusively, use a screen-capture program or your browser’s save function to grab copies of each screen as you complete the order. Most vendors send a confirming email, which we file in our “Never Delete” folder.
• File everything related to an order, including a copy of the original advertisement, the email (or faxed or written) confirmation provided by the reseller, copies of your credit card receipt, a copy of the packing list and invoice, and so on. We also jot down notes in our PIM regarding telephone conversations, including the date, time, telephone number and extension, person spoken to, purpose of the call, and so on. We print a copy of those to add to the folder for that order.
• Make it clear to the reseller that you expect them to ship the exact item you have ordered, not what they consider to be an “equivalent substitute.” Require that they confirm the exact items they will ship, including manufacturer part numbers. For example, if you order an EVGA 512-P3-N871-AR GeForce 9800 GTX+ graphics card with 512 MB of RAM, make sure the order confirmation specifies that item by name, full description, and EVGA product number. Don’t accept a less detailed description such as “graphics card,” “EVGA graphics card,” or even “EVGA 9800 GTX graphics card.” Otherwise, you may get less than you paid for—a lesser GeForce card, a card with a slower processor or less memory, or even a card with a GeForce processor made by another manufacturer.
• Verify warranty terms. Some manufacturers warrant only items purchased from authorized dealers in full retail packaging. For some items, the warranty begins when the manufacturer ships the product to the distributor, which may be long before you receive it. OEM products typically have much shorter warranties than retail-boxed products—sometimes as short as 90 days—and may be warranted only to the original distributor rather than to the final buyer. Better resellers may “endorse the manufacturer warranty” for some period on some products, often 30 to 90 days. That means that if the product fails, you can return the item to the reseller, who will ship you a replacement and take care of dealing with the manufacturer. Some resellers disclaim the manufacturer warranty, claiming that once they ship the item dealing with warranty claims is your problem, even if the product arrives DOA. We’ve encountered that problem a couple of times. Usually, mentioning phrases like “merchantability and fitness for a particular purpose” and “revocation of acceptance” leads them to see reason quickly. We usually demand that the reseller ship us a new replacement product immediately and include a prepaid return shipping label if they want the dead item back. We don’t accept or pay for dead merchandise under any circumstances, and neither should you.
• Direct resellers are required by law to ship products within the time period they promise. But that time period may be precise (e.g., “ships within 24 hours”) or vague (e.g., “ships within three to six weeks”). If the vendor cannot ship by the originally promised date, it must notify you in writing and specify another date by which the item will ship. If that occurs, you have the right to cancel your order without penalty. Make sure to make clear to the reseller that you expect the item to be delivered in a timely manner. Reputable vendors ship what they say they’re going to ship when they say they’re going to ship it. Unfortunately, some vendors have a nasty habit of taking your money and shipping whenever they get around to it. In a practice that borders on fraud, some vendors routinely report items as “in stock” when in fact they are not. Make it clear to the vendor that you do not authorize them to charge your credit card until the item actually ships, and that if you do not receive the item when promised you will cancel the order.

Even if you follow all of these guidelines, things may go wrong. Even the best resellers sometimes drop the ball. If that happens, don’t expect the problem to go away by itself. If you encounter a problem, remain calm and notify the reseller first. Good resellers are anxious to resolve problems. Find out how the reseller wants to proceed, and follow their procedures, particularly for labeling returned merchandise with an RMA number.

If you seem to have reached a dead end with the vendor, explain one last time why you are dissatisfied and ask them to resolve the problem. Tell them that unless they resolve the matter you will request a chargeback from your credit card company. Mail-order and Internet vendors live and die on credit card revenue, so maintaining a good relationship with the credit card companies is critically important to them. Finally, but only as a last resort, contact your bank or credit card issuer and request a chargeback. Be prepared to provide a full explanation of the problem with documentation.

Recommended Sources

The question we hear more often than any other is, “What company should I buy from?” When someone asks us that question, we run away, screaming in terror. Well, not really, but we’d like to. Answering that question is a no-win proposition for us, you see. If we recommend a vendor and that vendor treats the buyer properly, well that’s no more than was expected. But Thor forbid that we recommend a vendor who turns around and screws the buyer.

So, which online resellers do we buy from? Over the years, we’ve bought from scores of online vendors, and our favorites have changed. For the last decade or so, our favorite has been NewEgg (http://www.newegg.com). NewEgg offers an extraordinarily good combination of competitive prices, wide product selection, good customer support, timely shipping, and fair return or replacement policies. We know of no other direct vendor that even comes close.

NewEgg’s prices aren’t always rock-bottom, but they generally match those of any other vendor we’re willing to deal with. NewEgg runs daily specials that are often real bargains, so if you’re willing to consider alternatives and to accumulate components over the course of a few weeks, you can save a fair amount of money. NewEgg ships what it says it’s going to ship, when it says it’s going to ship it, and at the price it agreed to ship it for. If there’s a problem, it’s rectified. It’s hard to do better than that.

In the last year or two, we’ve started ordering some computer components from Amazon (http://www.amazon.com). Amazon’s selection is not as wide as NewEgg’s for most items, but its prices are excellent. Amazon generally ships quickly, and we consider it on a par with NewEgg for hassle-free ordering. Packaging is one area where Amazon has an advantage over NewEgg. Although NewEgg has recently taken steps to improve its packaging, in the past we have received components from NewEgg that were damaged in shipping. Conversely, Amazon’s attitude is apparently that nothing succeeds like excess. In the past, Amazon over-packaged, to put it mildly. So much so that it’s received complaints about all the packaging material that needs to be recycled. Fortunately, Amazon has responded to these complaints and now uses sane packaging.

All of that said, if you buy from NewEgg or Amazon and subsequently your goldfish dies and all your teeth fall out, don’t blame us. All we can say is that NewEgg and Amazon have always treated us right. Things can change overnight in this industry and, while we don’t expect NewEgg or Amazon to take a sudden turn for the worse, it could happen.

As for local retailers, we buy from—in no particular order—Best Buy, Target, Office Depot, OfficeMax, and our local computer specialty stores, depending on what we need and who happens to have advertised the best prices and rebates in the Sunday ad supplements. Walmart, which used to sell only assembled PCs, has recently started stocking PC components such as ATI video adapters, so we’ll add Walmart to our list as well.

Final Words

We’ve done our best in this chapter to tell you how to choose components for your new system and where and how to buy them. The specific components you need differ according to the type of system you plan to build. We describe how to make component-specific decisions in the “project system” chapters later in the book. So, before you actually start ordering components, you might want to read some (or all) of those chapters.

When the components arrive, restrain yourself. Don’t start building your system before the FedEx truck even pulls out of your driveway, particularly if this is your first system build. Read or re-read the relevant project chapter first.

One thing you should do immediately, though, is check the contents of the boxes that were just delivered. Verify what you ordered against the packing list and invoice, and verify what’s actually in the box against those documents. Usually everything will be right, but if you have components coming from different sources, you don’t want to wait a week or two before finding out that an early shipment was wrong or incomplete.

Take it a step further. Once you’ve verified that everything is correct with the order, start opening the individual component boxes. Look for a packing list in the front of the manual, and make sure that you actually received everything that was supposed to be in the box. It’s not uncommon for small parts—mounting hardware, cables, driver CDs, and so on—to be missing. If that happens, call the vendor immediately and tell them what’s missing from your order.

At this point, you should have everything you need to start building your new PC. It’s kind of like being a kid again, on Christmas morning.