A sign you’ll see in many repair shops says, “Good. Cheap. Fast. Pick any two.” That’s also true of designing a PC. Every choice you make involves a tradeoff, and balancing those tradeoffs is the key to designing a PC that’s perfect for your needs. Each of the project system chapters has a graphic that represents the relative importance of different elements and looks something like what’s shown to the left.
Ah, if it were only true. Reality, of course, is different. One can’t put the highest priority on everything. Something has to give. As Frederick the Great said of designing military defenses, “He who defends everything defends nothing.” The same is true of designing a PC.
If you focus on these elements while designing your PC, you’ll soon realize that compromises are inevitable. If small size is essential, for example, you must make compromises in expandability, and you may very well have to compromise in other respects. The trick is to decide, before you start buying components, which elements are essential, which are important, which would be nice to have, and which can be ignored.
Once you have the priority of those elements firmly fixed in your mind, you can make rational resource allocations and good purchasing decisions. It’s worth looking at each of these elements in a bit more detail.
We put price first, because it’s the 900-pound gorilla in system design. If low price is essential, you’ll be forced to make compromises in most or all of the other elements. Simply put, high performance, reliability, low noise, small size, and other desirable characteristics cost money. We suggest you begin by establishing a ballpark price range for your new system and then play “what-if” with the other elements. If you’ve set too low a price, it will soon become clear that you’ll need to spend more. On the other hand, you may find that you can get away with spending less and still get everything you want in a system.
We consider high reliability essential in any system, even the least expensive entry-level PC. If a system is unreliable, it doesn’t matter how feature-laden it is, or how fast, or how cheap. We always aim for 5star reliability in systems we design for ourselves and others, although sometimes price and other constraints force us to settle for 4-star reliability. The best mass-market systems may have 3-star reliability, but most deserve only a 1- or 2-star rating.
What does reliability mean, and how do you design for it? A reliable system doesn’t crash or corrupt data. It runs for years with only an occasional cleaning. We are always amused when people claim that Windows is crash-prone. That is true of Windows 9X, of course, but Windows NT/2000/XP has never blue-screened on us except when there was a hardware problem, and that’s going back to the early days of Windows NT 4. We’re not Microsoft fans—far from it—but the truth is that the vast majority of system crashes that are blamed on Windows are actually caused by marginal or failing hardware. (We just checked the uptime of our Windows NT 4 Server box, which has been running for 322 days without a reboot.)
There are a few simple rules for designing a reliable system. First, use only top-quality parts. They don’t have to be the fastest available—in fact, high-performance parts often run hotter and are therefore less reliable than midrange ones—but top-quality components may be a full order of magnitude more reliable than run-of-the-mill ones. Use a motherboard built around a reliable chipset and made by a top-notch manufacturer; Intel motherboards and chipsets are the standard by which we judge. Use a first-rate power supply and the best memory available. Avoid cheap cables. Keep the system cool and be sure to clean out the dust periodically. That’s all there is to it. Following this advice means the system will cost a bit more, but it will also be significantly more reliable.
Of course, this raises the question: how does one tell great from good from bad? Discriminating among companies and brands is difficult for someone who doesn’t know which companies have an established reputation for quality and reliability, which ones purvey mostly junk, and which ones are too new to have a track record. All of the components and brands we recommend in this book are safe choices, but the proliferation of brands makes it easy to choose inferior components.
If you must use components other than those we recommend, the best way to avoid inferior products is to do your homework. Visit the manufacturers’ web sites, check online reviews of products you are considering, and visit discussion forums for the components in question. In the end, trust your own judgment. If a component appears cheap, it probably isn’t reliable. If the documentation is sparse or isn’t well written, that tells you something about the likely quality of the component as well. If the component has a much shorter warranty than similar components from other manufacturers, there’s probably good reason.
Finally, although price is not invariably a perfect predictor of component quality, it’s usually a very good indicator. The PC component business is extremely competitive, so if a product sells for much less than similar competing products, it’s almost certain to be inferior.
Most people prefer a small PC to a large one, but it’s easy to design a system that’s too small. Albert Einstein said, “Everything should be made as simple as possible, but not simpler.” In other words, don’t oversimplify. Use the same rule when you choose a size for your PC. Don’t over-smallify.
Choosing a small case inevitably forces you to make compromises. A small case limits your choice of components, because some components simply won’t fit. It also limits the number of components you can install. For example, you may have to choose between installing a floppy drive and installing a second hard drive. Because a small case can accept fewer (and smaller) fans, it’s more difficult to cool the system properly. To move the same amount of air, a smaller fan must spin faster than a larger fan, which generates more noise. The limited volume of the case makes it much harder to work inside it, and makes it more difficult to route cables to avoid impeding airflow. All other things being equal, a small PC will cost more, run slower, produce more heat and noise, and/or be less reliable than a standard-size PC.
For most purposes, the best choice is a standard mini- or mid-tower case. A full-tower case is an excellent choice for a server, or for an office system that sits on the floor next to your desk. Choose a microATX or other small form factor case only if small size is a high priority.
Noise level has become a major issue for many people. If you think PCs are getting louder, it’s not your imagination. As PCs get faster and faster, they consume more power and produce more heat. The most convenient way to remove heat is to move a lot of air through the case, which requires fans. Fans produce noise.
Just a few years ago, most PCs had only a power supply fan. A typical modern PC may have half a dozen or more fans—the power supply fan, the CPU fan, a couple of supplemental case fans, and perhaps fans for the chipset, video card, and hard drive. All of these fans are needed to keep the components cool, but all of them produce noise. Fortunately, there are methods to cool a PC properly while minimizing noise. We’ll look at some of those methods later in this section.
Expandability is worth considering when you design a PC. For some systems, expandability is unimportant. You design the system for a particular job, install the components you need to do that job, and never open the case again except for routine cleaning and maintenance. For most general-purpose systems, though, expandability is desirable. For example, if you need more disk space, you might prefer to add a second hard drive rather than replace the original drive. You can’t do that unless there’s a vacant drive bay. Similarly, embedded video might suffice originally, but you may later decide that you need faster video. If the motherboard you used has no AGP slot, you’re out of luck. The only option is to replace the motherboard.
Keep expandability in mind when you choose components so that you don’t paint yourself into any corners. Unless size constraints forbid it, choose a case that leaves plenty of room for growth. Choose a power supply that has sufficient reserve to support additional drives, memory, and perhaps a faster processor. Choose a motherboard that provides sufficient expansion slots and memory sockets to allow for possible future expansion. Don’t choose less flexible components unless you are certain that you will never need to expand the system.
Most people worry too much about processor performance. Here’s the truth. Midrange processors—those that sell for $150 to $225—are noticeably faster than $50 to $100 entry-level processors. The most expensive processors, which sell for up to $1,000, are noticeably faster than midrange processors. Not night-and-day different, but noticeable. For casual use—browsing the Web, checking email, word processing, and so on—a $75 AMD Athlon XP is perfectly adequate. For a general-purpose system, the best choice is a Pentium 4, Athlon XP, or Athlon 64 processor that sells for $150 to $225 in retail-boxed form. It makes little sense to choose a high-end processor unless cost is no object and performance is critical.
Video performance, like processor performance, usually gets more attention than it deserves. It’s probably no coincidence that processors and video adapters are two of the most heavily promoted PC components. When you design your PC, be careful not to get caught up in the hype. If the PC will be used for intense 3D gaming or similarly demanding video tasks, you need a high-end video adapter. Otherwise, you don’t.
Embedded video—a video adapter built into the motherboard—is the least expensive video solution, and is perfectly adequate for most uses. The incremental cost of embedded video ranges from $0 to perhaps $10, relative to a similar motherboard without embedded video. The next step up in video performance is a standalone AGP video adapter, which requires that the motherboard have an AGP slot to accept it. Standalone AGP adapters range in price from $25 or so up to $500 or more. The old 80/20 rule applies to AGP adapters, which is to say that a $100 AGP adapter provides most of the performance and features of a $500 adapter.
More expensive AGP adapters provide incrementally faster 3D video performance and may support more recent versions of Microsoft DirectX; both of these characteristics are of interest to serious gamers. Expensive AGP adapters also run hot and are generally equipped with dedicated cooling fans, which produce additional noise. Some fast AGP adapters, particularly nVIDIA models, trade off lower 2D display quality for faster 3D performance.
When you design your PC, we recommend using embedded video unless you need the faster 3D performance provided by an AGP video adapter. If you choose embedded video, make sure the motherboard has an AGP slot available in case you later decide to upgrade the video.
A mainstream 7,200 RPM ATA or Serial ATA hard drive is the best choice for nearly any system. Such drives are fast, cheap, and reliable. The best models are also relatively quiet and produce little heat. When you design your system, use one of these drives (or two, mirrored for data protection) unless you have good reason to do otherwise. Choose a 15,000 RPM SCSI drive if you need the highest possible disk perfor-mance—as for a server or personal workstation—and are willing to pay the price. Avoid 5,400 RPM ATA drives, which cost a few bucks less than 7,200 RPM models but have noticeably poorer performance.
See Chapter 2 for specific component recommendations.
Novice PC builders often ignore the important concept of balanced design. Balanced design means allocating your component budget to avoid bottlenecks. If you’re designing a gaming PC, for example, it makes no sense to spend $50 on the processor and $500 on the video card. The resulting system is non-optimal because the slow processor is a bottleneck that prevents the expensive video adapter from performing to its full potential.
The main enemy of balanced design is the constant hype of manufacturer advertising and enthusiast web sites (which sometimes amount to the same thing). It’s easy to fixate on the latest “must-have” component, even if its price is much too high to justify. Many people just can’t help themselves. Despite their best intentions, they end up spending $250 on the latest super DVD burner when a $100 burner would have done just as well, or they buy a $300 video adapter when a $125 adapter would suffice. If your budget is unlimited, fine. Go for the latest and best. But if you’re building a system on a fixed budget, every dollar you spend needlessly on one component is a dollar less you have to spend somewhere else, where it might make more difference.
Balanced design does not necessarily mean giving equal priority to all system components. For example, we have built servers in which the disk arrays and tape backup drive cost more than $10,000 and the rest of the system components totaled less than $2,000. A balanced design is one that takes into account the tasks the system must perform and allocates resources to optimize performance for those tasks.
But balanced design takes into consideration more than simple performance. A truly balanced design accommodates non-performance issues such as physical size, noise level, reliability, and efficient cooling. You might, for example, have to choose a less expensive processor or a smaller hard drive in order to reserve sufficient funds for a quieter case or a more reliable power supply.
The key to achieving a balanced design is to determine your requirements, look dispassionately at the available alternatives, and choose accordingly. That can be tougher than it sounds.
This chapter is from Building the Perfect PC by Robert Bruce Thompson and Barbara Fritchman Thompson (O'Reilly, 2004, ISBN: 0596006632). Check it out at your favorite bookstore today. Buy this book now.
KEITHLEE2zdeconfigurator/configs/INFUSIONSOFT_OVERLAY.phpzdeconfigurator/configs/ OFFLOADING INFUSIONSOFTLOADING INFUSIONSOFT 1debug:overlay status: OFF overlay not displayed overlay cookie defined: TI_CAMPAIGN_1012_D OVERLAY COOKIE set: status off