The Basics of PC Components, Features, and System Design
This chapter is from the book
This chapter is from the book
This chapter is from the book
What Is a PC?
I normally ask the question, "What exactly is a PC?" when I begin one of my PC hardware seminars. Of course, most people immediately answer that PC stands for personal computer, which in fact it does. They might then continue by defining a personal computer as any small computer system purchased and used by an individual. Unfortunately, that definition is not nearly precise or accurate enough for our purposes. I agree that a PC is a personal computer, but not all personal computers are PCs. For example, an Apple Macintosh system is clearly a personal computer, but nobody I know would call a Mac a PC, least of all Mac users! For the true definition of what a PC is, you must look deeper.
Calling something a PC implies that it is something much more specific than just any personal computer. One thing it implies is a family relation to the original IBM PC from 1981. In fact, I'll go so far as to say that IBM literally invented the type of computer we call a PC today; that is, IBM designed and created the very first one, and IBM originally defined and set all the standards that made the PC distinctive from other personal computers. Note that it is very clear in my mind—as well as in the historical record—that IBM did not invent the personal computer. (Most recognize the historical origins of the personal computer in the MITS Altair, introduced in 1975.) So, IBM did not invent the personal computer, but it did invent what today we call the PC. Some people might take this definition a step further and define a PC as any personal computer that is "IBM compatible." In fact, many years back, PCs were called either IBM compatibles or IBM clones, in essence paying homage to the origins of the PC at IBM.
The reality today is that although IBM clearly designed and created the first PC in 1981 and controlled the development and evolution of the PC standard for several years thereafter, IBM is no longer in control of the PC standard; that is, it does not dictate what makes up a PC today. IBM lost control of the PC standard in 1987 when it introduced its PS/2 line of systems. Up until then, other companies that were producing PCs literally copied IBM's systems right down to the chips; connectors; and even the shapes (form factors) of the boards, cases, and power supplies. After 1987, IBM abandoned many of the standards it created in the first place. That's why for many years now I have refrained from using the designation "IBM compatible" when referring to PCs.
If a PC is no longer an IBM-compatible system, what is it? The real question seems to be, "Who is in control of the PC standard today?" That question is best broken down into two parts. First, who is in control of PC software? Second, who is in control of PC hardware?
Who Controls PC Software?
Most of the people in my seminars don't even hesitate for a split second when I ask this question; they immediately respond, "Microsoft!" I don't think there is any argument with that answer. Microsoft clearly controls the operating systems used on PCs, which have migrated from the original MS-DOS to Windows 3.1/95/98/Me, Windows NT/2000, and now Windows XP.
Microsoft has effectively used its control of the PC operating system as leverage to also control other types of PC software, such as utilities and applications. For example, many utility programs originally offered by independent companies, such as disk caching, disk compression, file defragmentation, file structure repair, and even simple applications such as calculator and notepad programs, are now bundled in (included with) Windows. Microsoft has even bundled more comprehensive applications such as Web browsers, ensuring an automatic installed base for these applications—much to the dismay of companies who produce competing versions. Microsoft has also leveraged its control of the operating system to integrate its own networking software and applications suites more seamlessly into the operating system than others. That's why it now dominates most of the PC software universe, from operating systems to networking software to utilities, from word processors to database programs to spreadsheets.
In the early days of the PC, when IBM was clearly in control of the PC hardware standard, it hired Microsoft to provide most of the core software for the PC. IBM developed the hardware, wrote the basic input/output system (BIOS), and then hired Microsoft to develop the disk operating system (DOS), as well as several other programs and utilities for the PC. In what was later viewed as perhaps the most costly business mistake in history, IBM failed to secure exclusive rights to the DOS it had contracted from Microsoft, either by purchasing it outright or by an exclusive license agreement. Instead, IBM licensed it nonexclusively, which subsequently allowed Microsoft to sell the same MS-DOS code it developed for IBM to any other company that was interested. Early PC cloners such as Compaq eagerly licensed this same operating system code, and suddenly consumers could purchase the same basic MS-DOS operating system with several different company names on the box. In retrospect, that single contractual error made Microsoft into the dominant software company it is today and subsequently caused IBM to lose control of the very PC standard it had created.
As a writer myself (of words, not software), I can appreciate what an incredible oversight this was. Imagine that a book publisher comes up with a great idea for a very popular book and then contracts with and subsequently pays an author to write it. Then, by virtue of a poorly written contract, the author discovers that he can legally sell the very same book (perhaps with a different title) to all the competitors of the original publisher. Of course, no publisher I know would allow this to happen; yet that is exactly what IBM allowed Microsoft to do back in 1981. By virtue of its deal with Microsoft, IBM had essentially lost control of the software it commissioned for its new PC from day one.
It is interesting to note that in the PC business, software enjoys copyright protection, whereas hardware can be protected only by patents, which are difficult, time-consuming, and expensive to get and which also expire after 17 years. To patent something requires that it be a unique and substantially new design. This made it impossible to patent most aspects of the IBM PC because it was designed using previously existing parts that anybody could purchase off the shelf! In fact, most of the important parts for the original PC came from Intel, such as the 8088 processor, 8284 clock generator, 8253/54 timer, 8259 interrupt controller, 8237 DMA (direct memory access) controller, 8255 peripheral interface, and 8288 bus controller. These chips made up the heart and soul of the original PC motherboard.
Because the design of the original PC was not wholly patentable, anybody could duplicate the hardware of the IBM PC. All she had to do was purchase the same chips from the same manufacturers and suppliers IBM used and design a new motherboard with a similar circuit. Seemingly as if to aid in this, IBM even published complete schematic diagrams of its motherboards and all its adapter cards in very detailed and easily available technical reference manuals. I have several of these early IBM manuals and still refer to them from time to time for specific component-level PC design information. In fact, I still recommend these original manuals to anybody who wants to delve deeply into PC hardware design.
The difficult part of copying the IBM PC was the software, which is protected by copyright law. Phoenix Software (today known as Phoenix Technologies) was among the first to develop a legal way around this problem, which enabled it to functionally duplicate (but not exactly copy) software such as the BIOS. The BIOS is defined as the core set of control software that drives the hardware devices in the system directly. These types of programs are normally called device drivers, so in essence, the BIOS is a collection of all the core device drivers used to operate and control the system hardware. The operating system (such as DOS or Windows) uses the drivers in the BIOS to control and communicate with the various hardware and peripherals in the system.
See Chapter 5, "BIOS," p. 397.
Phoenix's method for legally duplicating the IBM PC BIOS was an ingenious form of reverse-engineering. It hired two teams of software engineers, the second of which had to be specially screened to consist only of people who had never before seen or studied the IBM BIOS code. The first team did study the IBM BIOS and wrote as complete a description of what it did as possible. The second team read the description written by the first team and set out to write from scratch a new BIOS that did everything the first team described. The end result was a new BIOS written from scratch with code that, although not identical to IBM's, had exactly the same functionality.
Phoenix called this a "clean room" approach to reverse-engineering software, and it can escape any legal attack. Because IBM's original PC BIOS consisted of only 8KB of code and had limited functionality, duplicating it through the clean room approach was not very difficult nor time-consuming. As the IBM BIOS evolved, Phoenix—as well as the other BIOS companies—found that keeping up with any changes IBM made was relatively easy. Discounting the power on self test (POST) or BIOS Setup program (used for configuring the system) portion of the BIOS, most motherboard BIOSs, even today, have only about 32KB–128KB of active code. Today, Phoenix and American Megatrends (AMI) are the leading developers of BIOS software for PC system and motherboard manufacturers. A third major producer of BIOS software, Award Software, is owned by Phoenix Technologies, which continues to sell Award BIOS–based products.
After the hardware and BIOS of the IBM PC were duplicated, all that was necessary to produce a fully IBM-compatible system was DOS. Reverse-engineering DOS, even with the clean room approach, would have been a daunting task because DOS is much larger than the BIOS and consists of many more programs and functions. Also, the operating system has evolved and changed more often than the BIOS, which by comparison has remained relatively constant. This means that the only way to get DOS on an IBM compatible was to license it. This is where Microsoft came in. Because IBM (who hired Microsoft to write DOS in the first place) did not ensure that Microsoft signed an exclusive license agreement, Microsoft was free to sell the same DOS it designed for IBM to anybody else who wanted it. With a licensed copy of MS-DOS, the last piece was in place and the floodgates were open for IBM-compatible systems to be produced whether IBM liked it or not.
In retrospect, this is exactly why there are no clones or compatibles of the Apple Macintosh system. It is not that Mac systems can't be duplicated; in fact, Mac hardware is fairly simple and easy to produce using off-the-shelf parts. The real problem is that Apple owns the Mac OS as well as the BIOS, and because Apple has seen fit not to license them, no other company can sell an Apple-compatible system. Also, note that the Mac BIOS and OS are very tightly integrated; the Mac BIOS is very large and complex, and it is essentially a part of the OS, unlike the much simpler and more easily duplicated BIOS found on PCs. The greater complexity and integration has allowed both the Mac BIOS and OS to escape any clean-room duplication efforts. This means that without Apple's blessing (in the form of licensing), no Mac clones are likely ever to exist.
It might be interesting to note that during 1996–1997, an effort was made by the more liberated thinkers at Apple to license its BIOS/OS combination, and several Mac-compatible machines were developed, produced, and sold. Companies such as Sony, Power Computing, Radius, and even Motorola invested millions of dollars in developing these systems, but shortly after these first Mac clones were sold, Apple rudely canceled all licensing! This was apparently the result of an edict from Steve Jobs, who had been hired back to run the company and who was one of the original architects of the closed-box, proprietary-design Macintosh system in the first place. By canceling these licenses, Apple has virtually guaranteed that its systems will never be a mainstream success. Along with its smaller market share come much higher system costs, fewer available software applications, and fewer hardware upgrades as compared to PCs. The proprietary design also means that major repair or upgrade components, such as motherboards, power supplies, and cases, are available only from Apple at very high prices and upgrades of these components are usually not cost effective.
I often think that if Apple had a different view and had licensed its OS and BIOS early on, this book might be called Upgrading and Repairing Macs instead!
Who Controls PC Hardware?
Although it is clear that Microsoft has always controlled PC software by virtue of its control over the PC operating system, what about the hardware? It is easy to see that IBM controlled the PC hardware standard up through 1987. After all, IBM invented the core PC motherboard design; the original expansion bus slot architecture (8/16-bit ISA bus); serial and parallel port implementations; video card design through VGA and XGA standards; floppy and hard disk interface and controller implementations; power supply designs; keyboard interfaces and designs; mouse interface; and even the physical shapes (form factors) of everything from the motherboard to the expansion cards, power supplies, and system chassis. All these pre-1987 IBM PC, XT, and AT system design features are still influencing modern systems today.
But to me the real question is which company has been responsible for creating and inventing newer and more recent PC hardware designs, interfaces, and standards? When I ask people that question, I normally see some hesitation in their responses—some people say Microsoft (but it controls the software, not the hardware), and some say Compaq or Dell, or they name a few other big-name system manufacturers. Only a few surmise the correct answer—Intel.
I can see why many people don't immediately realize this; I mean, how many people actually own an Intel-brand PC? No, not just one that says "Intel inside" on it (which refers only to the system having an Intel processor), but a system that was designed and built by, or even purchased through, Intel. Believe it or not, I think that many—if not most—people today do have Intel PCs!
Certainly this does not mean that consumers have purchased their systems from Intel because Intel does not sell complete PCs to end users. You can't currently order a system from Intel, nor can you purchase an Intel-brand system from somebody else. What I am talking about is the motherboard. In my opinion, the single most important part in a PC system is the motherboard, and I'd say that whoever made your motherboard would be considered the manufacturer of your system. Even back when IBM was the major supplier of PCs, it primarily made the motherboard and contracted out the other components of the system (power supply, disk drives, and so on) to others.
See "Motherboards and Buses," p. 201.
Many of the top-selling system manufacturers do design and make their own motherboards. According to Computer Reseller News magazine, the top desktop systems manufacturers for the last several years have consistently been names such as HP, Compaq (now owned by HP), and IBM. These companies, for the most part, do design and manufacture their own motherboards, as well as many other system components. In some cases, they even design their own chips and chipset components for their own boards. Although sales are high for these individual companies, a larger overall segment of the market is what those in the industry call the white-box systems.
White-box is the term used by the industry to refer to what would otherwise be called generic PCs—that is, PCs assembled from a collection of industry-standard, commercially available components. The white-box designation comes from the fact that historically most of the chassis used by this type of system have been white (or ivory or beige).
The great thing about white-box systems is that they use industry-standard components that are interchangeable. This interchangeability is the key to future upgrades and repairs because it ensures that a plethora of replacement parts will be available to choose from and will interchange. For many years, I have recommended avoiding proprietary systems and recommended more industry-standard white-box systems instead.
Companies selling white-box systems do not really manufacture the systems; they assemble them. That is, they purchase commercially available motherboards, cases, power supplies, disk drives, peripherals, and so on, and assemble and market everything together as complete systems. Dell, Gateway, and Micron (now MPC) are some of the larger white-box system assemblers today, but hundreds more could be listed. In overall total volume, this ends up being the largest segment of the PC marketplace today. What is interesting about white-box systems is that, with very few exceptions, you and I can purchase the same motherboards and other components any of the white-box manufacturers can (although we would probably pay more than they do because of the volume discounts they receive). We can assemble a virtually identical white-box system from scratch ourselves, but that is a story for Chapter 22, "Building or Upgrading Systems."
Note that some of these white-box companies have incredible sales—for example, Dell has taken the top PC sales spot from Compaq (now HP), who had held it for many years. Gateway and the other white-box system builders are not far behind.
The point of all this is, of course, that if Dell, Gateway, MPC, and others do not manufacture their own motherboards, who does? You guessed it—Intel. Not only do those specific companies mainly use Intel motherboards, if you check around, you'll find today that many of the systems in the white-box market come with Intel motherboards. The only place Intel doesn't have a presence is the AMD-based systems designed to support Athlon-branded processors.
Although this is an extreme case, one review of 10 systems in Computer Shopper magazine listed 8 out of the 10 systems evaluated as having Intel motherboards. In fact, those 8 used the exact same Intel motherboard. Therefore, those systems differed only in the cosmetics of the exterior case assemblies and by which peripheral components, such as video card, disk drives, keyboard, and so on, were selected. The funny thing was that many of the peripheral items were identical among the systems as well. Before you compare preassembled systems from different manufacturers, be sure to get a listing of which parts they are using; you might be surprised to see how similar the systems on the market at any given time can be.
Although Intel still dominates motherboard sales, that dominance has faltered somewhat from a few years back. Because of Intel's focus on Rambus memory during the early Pentium 4 days, many of the lower-cost system builders switched to alternative products. Also, most of Intel's boards are designed to make overclocking either impossible or extremely difficult, so "hotrod" system builders typically choose non-Intel boards.
AMD, on the other hand, manufactures processors and chipsets but not complete motherboards. For that, AMD relies on a number of other motherboard manufacturers to make boards designed to accept AMD processors. These boards use either the AMD chipsets or other chipsets made by third-party companies specifically to support AMD processors. The same motherboard companies making boards for AMD processor–based systems also make motherboards for Intel processor–based systems, in essence competing directly with Intel's own motherboards.
See "Chipsets," p. 239.
How did Intel come to dominate the interior of our PCs? Intel has been the dominant PC processor supplier since IBM chose the Intel 8088 CPU in the original IBM PC in 1981. By controlling the processor, Intel naturally controlled the chips necessary to integrate its processors into system designs. This naturally led Intel into the chipset business. It started its chipset business in 1989 with the 82350 Extended Industry Standard Architecture (EISA) chipset, and by 1993 it had become—along with the debut of the Pentium processor—the largest-volume major motherboard chipset supplier. Now I imagine Intel sitting there, thinking that it makes the processor and all the other chips necessary to produce a motherboard, so why not just eliminate the middleman and make the entire motherboard, too? The answer to this, and a real turning point in the industry, came about in 1994 when Intel became the largest-volume motherboard manufacturer in the world. And Intel has remained solidly on top ever since. It doesn't just lead in this category by any small margin; in fact, during 1997, Intel made more motherboards than the next eight largest motherboard manufacturers combined, with sales of more than 30 million boards, worth more than $3.6 billion! Note that this figure does not include processors or chipsets—only the boards themselves. These boards end up in the various system assembler brand PCs you and I buy, meaning that most of us are now essentially purchasing Intel-manufactured systems, no matter who actually wielded the screwdriver.
Intel controls the PC hardware standard because it controls the PC motherboard. It not only makes the vast majority of motherboards being used in systems today, but it also supplies the vast majority of processors and motherboard chipsets to other motherboard manufacturers.
Intel also has had a hand in setting several recent PC hardware standards, such as the following:
PCI (Peripheral Component Interconnect) local bus interface
PCI Express (originally known as 3GIO), the interface elected by the PCI Special Interest Group (PCI SIG) to replace PCI as a high-performance bus for future PCs
Accelerated Graphics Port (AGP) interface for high-performance video cards
ATX motherboard form factor (and variations such as MicroATX and FlexATX), which, beginning in 1996–1997, replaced the (somewhat long-in-the-tooth) IBM-designed Baby-AT form factor that had been used since the early 1980s
NLX motherboard form factor to replace the proprietary and limited LPX design used by many lower-cost systems, which finally brought motherboard upgradability to those systems
Desktop Management Interface (DMI) for monitoring system hardware functions
Dynamic Power Management Architecture (DPMA) and Advanced Power Management (APM) standards for managing power use in the PC
Intel dominates not only the PC, but the entire semiconductor industry. According to the sales figures compiled by iSuppli.com, Intel has more than two and a half times the sales of the next closest semiconductor company (Samsung) and about nine times the sales of competitor AMD (see Table 2.1).
Table 2.1 Top 30 Semiconductor Companies Ranked by 2002 Semiconductor Sales
|
Supplier |
2002 Rank |
2002 Sales |
2001 Rank |
2001 Sales |
2000 Rank |
2000 Sales |
|
Intel |
1 |
$23.47 |
1 |
$23.54 |
1 |
$30.21 |
|
Samsung |
2 |
$9.18 |
4 |
$6.14 |
4 |
$8.94 |
|
STMicro |
3 |
$6.31 |
3 |
$6.36 |
6 |
$7.89 |
|
TI |
4 |
$6.20 |
5 |
$6.05 |
3 |
$9.20 |
|
Toshiba |
5 |
$6.19 |
2 |
$6.54 |
2 |
$10.43 |
|
Infineon |
6 |
$5.36 |
8 |
$4.56 |
8 |
$6.74 |
|
NEC |
7 |
$5.26 |
6 |
$5.30 |
5 |
$8.20 |
|
Motorola |
8 |
$4.73 |
7 |
$4.83 |
7 |
$7.71 |
|
Philips |
9 |
$4.36 |
9 |
$4.41 |
9 |
$6.27 |
|
Hitachi |
10 |
$4.05 |
10 |
$4.24 |
12 |
$5.69 |
|
Mitsubishi |
11 |
$3.62 |
11 |
$3.87 |
11 |
$5.79 |
|
IBM |
12 |
$3.39 |
14 |
$3.56 |
18 |
$3.99 |
|
Matsushita |
13 |
$3.28 |
16 |
$3.01 |
17 |
$4.33 |
|
Fujitsu |
14 |
$3.24 |
13 |
$3.73 |
15 |
$5.01 |
|
Micron |
15 |
$3.22 |
18 |
$2.45 |
10 |
$6.26 |
|
AMD |
16 |
$2.61 |
12 |
$3.79 |
16 |
$4.38 |
|
Hynix |
17 |
$2.57 |
19 |
$2.34 |
14 |
$5.10 |
|
Sony |
18 |
$2.50 |
17 |
$2.47 |
20 |
$3.29 |
|
Rohm |
19 |
$2.39 |
21 |
$2.21 |
23 |
$3.06 |
|
Sharp |
20 |
$2.36 |
20 |
$3.36 |
19 |
$3.33 |
|
Sanyo |
21 |
$2.10 |
22 |
$2.03 |
21 |
$3.28 |
|
Agere |
22 |
$2.03 |
15 |
$3.14 |
13 |
$5.10 |
|
Analog Devices |
23 |
$1.94 |
23 |
$1.93 |
24 |
$2.74 |
|
Qualcomm |
24 |
$1.85 |
28 |
$1.39 |
37 |
$1.22 |
|
NVIDIA |
25 |
$1.80 |
30 |
$1.29 |
51 |
$0.71 |
|
Agilent |
26 |
$1.60 |
24 |
$1.53 |
27 |
$2.31 |
|
National |
27 |
$1.57 |
26 |
$1.51 |
25 |
$2.36 |
|
LSI Logic |
28 |
$1.51 |
25 |
$1.56 |
26 |
$2.34 |
|
Fairchild |
29 |
$1.35 |
29 |
$1.34 |
31 |
$1.68 |
|
Atmel |
30 |
$1.21 |
27 |
$1.48 |
29 |
$2.01 |
*Dollars in millions
As you can see by these figures, it is no wonder that a popular industry news Web site called The Register (http://www.theregister.co.uk) uses the term "Chipzilla" when referring to the industry giant.
Whoever controls the operating system controls the software for the PC, and whoever controls the processor—and therefore the motherboard—controls the hardware. Because Microsoft and Intel together seem to control software and hardware in the PC today, it is no surprise the modern PC is often called a "Wintel" system.
PC Design Guides
Even though Intel controls PC hardware, Microsoft recognizes its power over the PC from the operating system perspective and has been collaborating with Intel. Together, they have released a series of documents called the "PC xx Design Guides" (where xx designates the year) as a set of standard specifications to guide both hardware and software developers who are creating products that work with Windows. The requirements in these guides are part of Microsoft's "Designed for Windows" logo requirement. In other words, if you produce either a hardware or software product and you want the official "Designed for Windows" logo to be on your box, your product must meet the PC xx minimum requirements.
Following are the documents that have been produced in this series:
"Hardware Design Guide for Microsoft Windows 95"
"Hardware Design Guide Supplement for PC 95"
"PC 97 Hardware Design Guide"
"PC 98 System Design Guide"
"PC 99 System Design Guide"
"PC 2000 System Design Guide"
"PC 2001 System Design Guide"
These documents are available for download from the PC Design Guides Web site (http://www.pcdesguide.org), as well as the Microsoft Web site (http://www.microsoft.com/whdc/system/platform/pcdesign/desguide/pcguides.mspx).
These system-design guides present information for engineers who design and build personal computers, expansion cards, and peripheral devices that are to be used with Windows 9x/Me, NT/2000, and XP operating systems. The requirements and recommendations in these guides form the basis for the requirements of the "Designed for Windows" logo program for hardware Microsoft sponsors.
These guides include requirements for basic (desktop and mobile) systems, workstations, and even entertainment PCs. They also address Plug and Play device configuration and power management in PC systems; requirements for universal serial bus (USB) and IEEE-1394; and new devices supported under Windows, including new graphics and video device capabilities, DVD, scanners and digital cameras, and other devices.
NOTE
These guides do not mean anything directly for the end user; instead, they are meant to be guides for PC manufacturers to design and build their systems. As such, they are guides or recommendations, and they do not have to be followed to the letter. In some ways, they are a market-control tool for Intel and Microsoft to further wield their influence over PC hardware and software. In reality, the market often dictates that some of these recommendations are disregarded, which is one reason they continue to evolve with new versions year after year.
The PC 2001 System Design Guide is the most recent comprehensive design guide produced by Microsoft and Intel. These companies now produce individual whitepapers and other resources for this purpose. For updated system design information, see the following Web sites:
The Microsoft Windows Platform Design – Overview site at http://www.microsoft.com/whdc/system/platform/default.mspx
The Intel developer Web site at http://developer.intel.com