Chip technology

The system unit is the box containing the CPU and other goodies (such as the speaker, power supply, and memory). If you unscrew that box and pry it open to see the circuitry inside, youíll see a green plastic board, on which is printed an electrical wiring diagram.

Since the diagramís printed in copper (instead of ink), the diagram conducts electricity; so it isnít just a diagram of an electrical circuit; it is an electrical circuit!

The green plastic board ó including the circuit printed on it ó is called a printed-circuit board (PC board). Each wire thatís stamped onto the PC board is called a trace.

The typical computer contains several PC boards.

Motherboard & babies

In your computer, the largest and most important PC board is called the motherboard. It lies flat on the bottom of the system unit.

The other PC boards are smaller. Those little baby boards (about the size of a postcard) are called PC cards.

The typical motherboard has several slots on it. Into each slot, you can put a PC card.

PCMCIA cards

If you buy a modern notebook computer, youíll see the caseís right-hand wall has a special slot in it. You can shove a card into that slot without opening the notebookís case.

The kind of card that fits into that special slot is small and thin ó the size of a credit card. That kind of card was invented by the Personal-Computer Memory-Card International Assocation (PCMCIA) and therefore called a PCMCIA card. That slot is called a PCMCIA slot.

People have trouble remembering what "PCMCIA" stands for. Cynics say it stands for "People Canít Memorize Computer Industry Acronyms". Since "PCMCIA" also stands for "Politically Correct Members of the CIA", computerists pronounce "PCMCIA" in two breaths: they say "PCM", then pause, then say "CIA".

Some PCMCIA cards are very thin. Other PCMCIA cards are slightly thicker, so they can hold extra circuitry. A PCMCIA card and its slot are called Type 1 if their thickness is 3.3 millimeters, Type 2 if 5 millimeters, Type 3 if 10.5 millimeters, Type 4 if 18 millimeters.

Caterpillars

On each PC board, youíll see black rectangles. If you look closely at a black rectangle, youíll see it has tiny legs, so it looks like a black caterpillar. (Though farmers think it looks like a "black caterpillar", city folks think it looks more like a "yucky roach". Kids call it just "a black thingy with legs".)

The "caterpillars" come in many sizes. In a typical computer, the shortest caterpillars are three-quarters of an inch long and have 7 pairs of legs; the longest are two inches long and have 20 pairs of legs.

Though each black caterpillar has legs, it doesnít move. Itís permanently mounted on the PC board.

Each leg is made of tin and called a pin.

Sadistic hobbyists play a game where they yank the caterpillars from a PC board and throw the caterpillars across the room. That gameís called "tin-pin bowling".

Hidden inside the caterpillar is a metal square, called a chip, which is very tiny. The typical chip is just an eighth of an inch long, an eighth of an inch wide, and a hundredth of an inch thick! On that tiny metal chip are etched thousands of microscopic electronic circuits! Since all those circuits are on the chip, the chipís called an integrated circuit (IC).

Four purposes

Each chip serves a purpose. If the chipís purpose is to "think", itís called a processor chip. If the chipís purpose is to "remember" information, itís called a memory chip. If the chipís purpose is to help devices communicate with each other, itís called an interface chip. If the chipís purpose is to act as a slave and helper to other chips, itís called a support chip.

So a chip is either a processor chip or a memory chip or an interface chip or a support chip ó or itís a combination chip that accomplishes several purposes.

How chips are designed

To design a chip, the manufacturer hires an artist, who draws on paper a big sketch of what circuits are to be put onto the chip. It helps if the artist also has a degree in engineering ó and knows how to use another computer to help draw all the lines.

After the big sketch is drawn, it is photographed.

Have you ever photographed your friend and asked the photography store for an "enlargement"? To produce a chip, the chipís manufacturer does the opposite: it photographs the sketch but produces a "reduction" to just an eighth of an inch on each side! Whereas a photo of your friend is made on treated paper, the tiny photo of the chipís circuitry consists of metal and semiconductors on treated silicon so the photoís an actual working circuit! That photographic process is called photolithography (or photolith).

Many copies of that photo are made on a large silicon wafer. Then a cookie cutter slices the wafer into hundreds of chips. Each chip is put into its own caterpillar.

The caterpillarís purpose is just to hide and protect the chip inside it; the caterpillarís just a strange-looking package containing the chip. Since the caterpillarís a package that has two rows of legs, itís called a dual in-line package (DIP). That DIPís only purpose is to house the chip.

Computer hobbyists are always talking about chips & DIPs. Thatís why computer hobbyists, at parties, serve chips & dips. And thatís why computer hobbyists are called "dipchips".

Buying chips

If you ask a computer dealer to sell you a chip, the dealer also gives you the chipís DIP (the entire caterpillar). Since youíve asked for a chip but also received a DIP, you might get confused and think that the caterpillar (the DIP) is the chip. But that caterpillarís not the chip; the chip hides inside the caterpillar.

The typical caterpillar-and-chip costs $3. You might pay somewhat more or somewhat less, depending on how fancy the chipís circuitry is.

If the circuits in a chip are defective, itís called a "buffalo chip". Folks who dislike that tacky term say "potato chip" or "chocolate chip" instead, like this: "Hey, the computerís not working! It must be made of chocolate chips!"

You can get chips from these famous mail-order chip suppliers:

Chip supplier Address Phone

JDR Microdevices 2233 Samaritan Dr., San Jose CA 95124 800-538-5000 or 408-559-1200

Jameco 1355 Shoreway Rd., Belmont CA 94002 415-592-8097, 24 hours

ACP 1310 E. Edinger, Santa Ana CA 92705 800-FONE-ACP

The following chip suppliers are newer and often charge less:

Chip supplier Address Phone

Memory Express 15140 Valley Blvd., City of Industry CA 91744 800-877-8188 or 818-333-6389

LA Trade 22825 Lockness Ave., Torrance CA 90501 800-433-3726 or 310-539-0019

Wordwide Tech 437 Chestnut St., Philadelphia PA 19106 800-457-6937 or 215-922-0050

Chip Merchant 9541 Ridgehaven Ct., San Diego CA 92123 800-426-6375 or 619-268-4774

How chips chat

The chip inside the caterpillar acts as the caterpillarís brain. The caterpillar also contains a "nervous system", made of thin wires that run from the brain (the chip) to the legs (the pins). The wires in the caterpillarís nervous system are very thin: each wireís diameter is about half of a thousandth of an inch.

If one caterpillar wants to send electrical signals to another caterpillar, the signals go from the first caterpillarís brain (chip) through the caterpillarís nervous system to its legs (pins). Each pin is attached to a trace (wire) on the PC board. The signals travel through those traces, which carry the signals across the PC board until the signals reach the second caterpillarís pins. Then the signals travel through the second caterpillarís nervous system to that caterpillarís brain (chip).

Binary code To communicate with each other, the caterpillars use a secret code. Each code is a series of 1ís and 0ís. For example, the code for the letter A is 01000001; the code for the letter B is 01000010; the code for the number 5 is 101; the code for the number 6 is 110.

Thatís called the binary code, because each digit in the code has just two possibilities: itís either a 1 or a 0. In the code, each 1 or 0 is called a binary digit.

A binary digit is called a bit. So in the computer, each bit is a 1 or a 0.

When a caterpillar wants to send a message to another caterpillar, it sends the message in binary code. To send a 1, the caterpillar sends a high voltage through the wires; to send a 0, the caterpillar sends little or no voltage through the wires.

So to send the number 5, whose code number is 101, the caterpillar sends a high voltage (1), then a low voltage (0), then a high voltage (1). To send those three bits (1, 0, and then 1), the caterpillar can send them in sequence through the same leg (pin); or for faster transmission, the caterpillar can send them through three pins simultaneously: the first pin sends 1, while the next pin sends 0 and the third pin sends 1.

The speed at which bits are sent is measured in bits per second (bps).

Bipolar versus MOS

Chips can be manufactured in two ways:

The old wayís called bipolar.

The new wayís called metal-oxide semiconductor (MOS, which is pronounced "moss"). Itís more popular because it costs less, consumes less electricity, and can hold more circuitry inside the chip.

Microcomputers use just MOS. Minicomputers and maxicomputers use mainly MOS chips but also contain a few bipolar chips, because bipolar chips have one (and only one) advantage over MOS chips: bipolar chips work faster.

The most popular kind of MOS is called negative-channel MOS. (Itís also called n-channel MOS or NMOS, which is pronounced "en moss".) The main alternative, called complementary MOS (or CMOS, pronounced "sea moss"), consumes even less electricity but canít hold as much circuitry inside the chip. CMOS chips are used in simple-minded battery-operated computers (such as digital watches, pocket calculators, pocket computers, and notebook computers) and in some parts of larger computers.

CPU

The part of the computer that thinks ("the brain") is called the processor (or central processing unit or CPU).

In a maxicomputer or minicomputer, the processor consists of several chips, which are processor chips.

In a microcomputer, the processor is so small that it consists of just a single chip, called a microprocessor. It sits on the motherboard. Yes, in a typical microcomputer, the part that does all the thinking is just a tiny square of metal, less than ľ" on each side!

Intelís designs

In the IBM PC and clones, the microprocessor uses a design invented by Intel. Iíll explain Intelís microprocessors now (and discuss competitors later).

In the original IBM PC (and in the IBM PC XT), the microprocessor was the Intel 8088. IBM computers (and clones) containing that chip are called XT-class computers.

Later, Intel invented an improved version, called the Intel 80286. Since "80286" is too long a number for us humans to remember, most of us just call it the Intel 286. IBM used it in the IBM PC AT computer. Thatís why computers containing that chip are called AT-class computers.

After inventing the Intel 286, Intel invented a further improvement (called the Intel 386), then an even further improvement (called the Intel 486).

In 1993, Intel began selling an even further improvement, which ought to be called a 586; but Intel calls it the Pentium instead, so Intel can trademark the name and prevent companies from copying it. Itís the first computer chip that sounds like a breakfast cereal: "Hey, kids, to put zip into your life, try Penti-yumms. They build strong bodies, 5 ways!"

While inventing the Pentium, Intel gave it this secret code-name: "P5". Many folks still call that chip the P5.

So altogether, IBM microcomputers and clones come in five popular classes:

Chip Invented Transistors on chip

8088 1979 29,000 transistors

286 1982 134,000 transistors

386 1985 275,000 transistors

486 1989 1,200,000 transistors

Pentium (P5) 1993 3,100,000 transistors

Some programs run okay on any chip; but many modern programs require a 286, 386, 486, or Pentium and wonít run on an 8088.

To run modern programs FAST and use all the modern features, you need a 486, or Pentium.

Most computers built today contain a Pentium. A few computers built today contain a 486, which is slightly slower than a Pentium.

The 8088, 286, and 386 chips are found just in pocket computers, used computers, and old computers that liquidators try to unload. Many homes and offices still have old 8088 computers, bought many years ago. Folks who still use those ancient computers restrict themselves to running very old-fashioned programs.

Megahertz

In an army, when soldiers march, theyíre kept in step by a drill sergeant who yells out, rhythmically, "Hup, two, three, four! Hup, two, three, four! Hup, two, three, four!"

Like a soldier, the microprocessor takes the next step in obeying your program just when instructed by the computerís "drill sergeant", which is called the computer clock. The clock rhythmically sends out a pulse of electricity; each time the clock sends out a pulse, the microprocessor does one more step in obeying your program.

The clock sends out millions of pulses every second, so the microprocessor accomplishes millions of steps in your program every second!

Each pulse is called a clock cycle. The clockís speed is measured in cycles per seconds.

A "cycle per second" is called a hertz (Hz), in honor of the German physicist Heinrich Hertz. A "million cycles per second" is called a megahertz (MHz).

When Intel invented the Pentium chip in 1993, the Pentiumís clock did 60 million cycles per second. Thatís 60 megahertz! Intel also invented a faster Pentium, at 66 megahertz, then even faster Pentiums at 75, 90, 100, 120, 133, 150, 166, 200, 233, 266, and 300 megahertz. For example, a 200-megahertz Pentium thinks twice as fast as a 100-megahertz Pentium.

A 60-megahertz Pentium is called a Pentium-60. A 200-megahertz Pentium is called a Pentium-200.

Slower than a Pentium

The Pentium is an amazing chip: while it thinks about one part of your program, it simultaneously starts getting the next part of your program ready for processing. That chipís ability to do several things simultaneously is called parallel processing. The Pentium is smarter than earlier chips (the 8088, 286, 386, and 486): the Pentium can perform more tasks simultaneously; it performs more parallel processing. Earlier chips seem slower: too often during a clock cycle in earlier chips, part of the chip "does nothing" while waiting for the other part of the chip to catch up. Those earlier chips therefore accomplish less useful work during a clock cycle than a Pentium.

During a clock cycle, a 486 accomplishes half as much useful work as a Pentium. We say the 486ís usefulness factor is Ĺ.

During a clock cycle, a 386 accomplishes a quarter as much useful work as a Pentium, so the 386ís usefulness factor is ľ. A 286ís usefulness factor is 1/5. An 8088ís usefulness factor is 1/20.

Youíve seen that those early chips accomplish less useful work during a clock cycle than a Pentium. Moreover, they accomplish fewer clock cycles per second than a Pentium; they have fewer megahertz:

Chip Megahertz Usefulness

Intel 8088 4.77, 7.18 1/20

Intel 286 6, 8, 10, 12 1/5

Intel 386 16, 20, 25, 33 1/4

Intel 486 20, 25, 33, 50, 66, 75, 100 1/2

Pentium 60, 66, 75, 90, 100, 120, 133, 150, 166, 200, 233, 266, 300 1

For example, suppose you buy an Intel 486 going at 100-megahertz. Since it suffers from a usefulness factor of Ĺ, it accomplishes just Ĺ as much useful work per cycle as a 100-megahertz Pentium, so it acts about as fast as a 50-megahertz Pentium. A 20-megahertz 386, which suffers from a usefulness factor of ľ, acts about as fast as a 5-megahertz Pentium. A 10-megahertz 286, which suffers from a usefulness factor of 1/5, acts about as fast as a 2-megahertz Pentium.

The slowest chip is a 4.77-megahertz 8088. Since it suffers from a usefulness factor of 1/20, it acts about as fast as a 0.2385-megahertz Pentium. Thatís 1258 times slower than the fastest Pentium, which goes at 300 megahertz. Yes, the fastest IBM-compatible computers think over 1200 times faster than the slowest ones! Thatís progress!

The "usefulness factor" is just an approximate average. During a cycle, for example, a 486 accomplishes about Ĺ as much useful work as a Pentium, on the average; but on certain tasks a 486 accomplishes more than "Ĺ as much", and on other tasks it accomplishes less.

Variant chips

The Intel 8088 comes in two versions. One version (called simply the "8088") goes slightly slower than the other version (called the 8086).

The Intel 386 comes in two versions. One version (called the 386SX) goes slightly slower than the other version (called the 386DX).

The Intel 486 comes in two versions. One version (called the 486SX) goes slower than the other version (called the 486DX). Moreover, the 486DX comes in three varieties: the original 486DX, the 486DX2, and the 486DX4.

Intelís invented four versions of the Pentium:

The Pentium classic is the oldest and slowest kind of Pentium. Invented in 1993, itís the kind of Pentium found in most computers built from 1993 through 1996.

The Pentium MMX is slightly faster. Invented in January 1997, itís the kind of Pentium found in most computers built in 1997 and afterwards. It runs most programs about 15% faster than a Pentium classic; for example, a 200-megahertz Pentium MMX runs programs about 15% faster than a 200-megahertz Pentium classic. Thatís because the Pentium MMX is designed slightly better than a Pentium classic and contains twice as much internal level-1 cache memory (an extremely fast form of memory that holds a copy of whatís coming from other memory). Itís called MMX because it also understands 57 extra instructions (called MultiMedia eXtensions), which can theoretically increase the speed of multimedia (video & sound) dramatically; but no important programs have been invented yet to make good use of those 57 extra instructions. Those 57 extra instructions just duplicate some of the intelligence found on fancy video-&-sound cards anyway. Intelís official name for this chip is "Pentium with MMX Technology", but most folks say just "Pentium MMX".

The Pentium 2 is the newest and fastest kind of Pentium. Invented in May 1997, it runs most programs about 30% faster than a Pentium MMX. Like the Pentium MMX, it understands the 57 multimedia instructions. Itís too expensive for most folks; just rich folks buy it. Intelís official name for this chips is "Pentium II"; but to avoid Roman numerals Iíll write "Pentium 2".

The Pentium 2 replaces an old 1995 expensive version, called the Pentium Pro, which ran some programs fast but ran other programs slowly (even slower than a Pentium classic!) and lacked MMX. The Pentium Pro was nicknamed the 686 or P6; the Pentium 2 now inherits those same nicknames.

Hereís how many megahertz are available:

Intel chip Megahertz

8088 4.77, 7.18

8086 8, 10

286 6, 8, 10, 12

386SX 16, 20, 25, 33

386DX 16, 20, 25, 33

486SX 20, 25, 33

486DX 25, 33, 50

486DX2 50, 66

486DX4 75, 100

Pentium classic 60,66,75,90,100,120,133,150,166,200

Pentium Pro 150, 166, 180, 200

Pentium MMX 166, 200, 233

Pentium 2 233, 266, 300

Here are some prices:

Intel chip Megahertz Price

486SX 33 $35

486DX2 50 $45

486DX2 66 $69

486DX4 100 $75

Pentium classic 75 $88

Pentium classic 100 $92

Pentium classic 150 $107

Pentium classic 166 $146

Pentium MMX 166 $161

Pentium MMX 200 $266

Pentium MMX 233 $396

Pentium 2 233 $520

Pentium 2 266 $630

Pentium 2 300 $938

That chart shows the price charged by discount dealers (Memory Express and Worldwide Technologies) for a single chip when this book went to press in November 1997. By the time you read this, prices might be lower, since Intel drops prices 3 or 4 times per year. If you buy 1000 chips at a time directly from Intel, you pay even less.

Math coprocessor

Each Pentium chip includes math coprocessor circuitry, which handles advanced math fast. That circuitry can multiply & divide long numbers & decimals; it can also compute square roots, logarithms, and trigonometry.

Primitive chips ó the 8088, 8086, 286, 386SX, 386DX, and 486SX ó do not include such circuitry.

To make a primitive chip do an advanced math problem, you must feed the chip a program that teaches the chip how to break the advanced problem down into a series of simpler problems. That program runs slowly ó nearly 100 times slower than if a math coprocessor were present!

Youíll be very annoyed at the slowness if youíre a scientist trying to do advanced math ó or if youíre artist trying to rotate a picture, since the computer computes the rotated imageís new coordinates by using trigonometry. For example, if you draw a 3-D picture of a house and then tell the computer to show how the house looks from a different angle, you need a math coprocessor to avoid a long delay.

On the other hand, if you use the computer just as a souped-up typewriter (to record and edit your writing) or as an electronic filing cabinet (to record names and addresses on a mailing list), youíll never notice the lack of a math coprocessor, since youíre not doing advanced math.

Each 486DX chip (and 486DX2 and 486DX4) includes math-coprocessor circuitry; the 486SX does not. So hereís the only difference between a 486DX and a 486SX: the 486SX lacks math-coprocessor circuitry.

Intel invented the 486DX, then later invented the 486SX by using this manufacturing technique: Intel took each 486DX whose math coprocessor was faulty and called it a 486SX. So a 486SX was just a defective 486DX.

If you buy a 486SX today, you get a 486DX whose math coprocessor is either defective or missing.

If your CPU lacks math-coprocessor circuitry (because your CPU is an 8088, 8086, 286, 386, or 486SX), hereís how to do math quickly: buy a supplementary chip, called a math coprocessor chip. Put it next to the CPU chip on the motherboard. It contains the math-coprocessor circuitry that the CPU lacks. Instead of buying a math coprocessor chip made by Intel, you can buy an imitation made by Cyrix or ULSI:

CPU Which math coprocessor to buy

8088, 8086 Intel 8087 ($45)

286 Intel 287 ($49), Cyrix 287XL ($35)

386SX Intel 387SX ($49), Cyrix 83S87 ($44), ULSI 3S87 ($39)

386DX Intel 387DX ($49), Cyrix 83D87 ($48)

486SX Intel 487SX ($183)

Chart of details

Here are more details about how the various chips differ from each other:

Internal External Math

Chip accum. data path Address MMX copr. Internal MHz External Mhz

8088 16-bit 8-bit 20-bit no no 4.77, 7.18 same as internal

8086 16-bit 16-bit 20-bit no no 8, 10 same as internal

286 16-bit 16-bit 24-bit no no 6, 8, 10, 12 same as internal

386SX 32-bit 16-bit 24-bit no no 16, 20, 25, 33 same as internal

386DX 32-bit 32-bit 32-bit no no 20, 25, 33 same as internal

486SX 32-bit 32-bit 32-bit no no 25, 33 same as internal

486DX 32-bit 32-bit 32-bit no yes 25, 33, 50 same as internal

486DX2 32-bit 32-bit 32-bit no yes 50, 66 one-half of internal

486DX4 32-bit 32-bit 32-bit no yes 75, 100 one-third of internal

Pentium classic 64-bit 64-bit 32-bit no yes 75 50

Pentium classic 64-bit 64-bit 32-bit no yes 60, 90, 120, 150 60

Pentium classic 64-bit 64-bit 32-bit no yes 66, 100, 133, 166, 200 66

Pentium MMX 64-bit 64-bit 32-bit yes yes 166, 200, 233 66

Pentium Pro 86-bit 86-bit 36-bit no yes 150, 180 60

Pentium Pro 86-bit 86-bit 36-bit no yes 166, 200 66

Pentium 2 86-bit 86-bit 36-bit yes yes 233, 266, 300 66

Here are more details about what the chart means.Ö

Internal accumulator Each chip contains registers. Each register can hold a binary code number (such as 01000001). The chipís main register is called the accumulator.

If the accumulator is wide enough to hold 32 bits inside it (such as 10000110111001111110010101010101), the accumulator is called 32-bit; the chip is said to contain a 32-bit accumulator and be 32-bit internally.

If the accumulator is narrower and holds just 16 bits, the accumulator is called 16-bit. In that case, the chip can handle code numbers that are 16 bits long but not code numbers that are 32 bits long. If you try to feed that chip a 32-bit code number, the chip wonít understand it.

The typical program uses just 16-bit instructions. (Instead of using a 32-bit instruction, it uses a pair of 16-bit instructions.)

But a few fancy programs use 32-bit instructions. To run those 32-bit programs, you must buy a chip thatís 32-bit internally. The chart shows that to run the fanciest programs (32-bit), you must buy at least a 386SX.

External data path The column marked "external data path" tells how many of the chipís pins transmit data.

As you can see from the chart, the 386SX is "32-bit internal, 16-bit external". That means the 386SX contains a 32-bit accumulator but has just 16 data pins. To transmit the accumulatorís 32 bits, the chip sends out 16 of the bits (on the 16 data pins), then sends out the next 16 bits by using those same pins.

That technique of using just a few pins to transmit many bits is called multiplexing. Computerists say the 386SX is "a 32-bit chip multiplexed onto 16 pins"; they say the 386SX is a multiplexed 386DX.

Thatís why the 386SX is slightly slower than the 386DX: to transmit the 32 bits, the 386SX must send out two bursts of 16 bits, whereas the 386DX can send out a single burst of 32 bits all at once!

Notice that the 386SX is just as smart as the 386DX ó it understands the same 32-bit codes ó but it transmits them more slowly (as 2 bursts of 16, instead of 1 burst of 32). So the 386SX is smart but a slow communicator ó like Einstein with his mouth full and trying to talk through a narrow drinking straw.

The 8088 is a multiplexed 8086. Like the 8086, the 8088 thinks about 16 bits; but the 8088 must send them out in two 8-bit bursts.

Address The computerís main memory (which consists of RAM chips and ROM chips) is like a city: each location in it has an address. If the main memory is large enough to hold lots of info, it has lots of addresses.

A city has addresses such as "231 17th Street, Apartment 501". In the computerís main memory, each address is a binary code number instead, such as 01000101010111101010.

For an 8088 or 8086, each address must be brief: just 20 bits long. An 8088 or 8086 therefore canít handle a big main memory ó and canít handle big programs.

A 286 can handle longer addresses (24-bit) so it can handle the big main memory required by modern big programs. To run modern big programs, you must buy at least a 286.

Though 24-bit addresses are long enough to handle all popular programs sold today, the chart shows that the fanciest chips permit even bigger addresses (32-bit or 36-bit), to prepare for the bigger programs of the far future ó and to handle computers that are networked together and share a gigantic big RAM.

External megahertz If a chip is simple (an 8088, 8086, 286, 386SX, 386DX, 486SX, or 486DX), it communicates at the same speed as it thinks. For example, a chip that thinks at 20 megahertz communicates at 20 megahertz.

The laws of physics make it difficult & expensive to manufacture a motherboard that communicates faster than 33 megahertz, and very difficult & expensive to manufacture a motherboard that communicates faster than 66 megahertz. Any CPU chip thinking faster than 66 megahertz must therefore slow down when communicating with the other chips on the motherboard.

For example, suppose you buy a Pentium chip thatís advertised as being "133-megahertz". That means that the chip performs 133 million cycles per second while thinking; but when the chip wants to transmit its answers (or questions) to other chips on the motherboard, the chip must perform the transmission at just 66 megahertz. So the chip thinks quickly but talks slowly ó like lawyer smart enough to talk slowly to a stupid jury.

As you can see from the chart, Pentium chips communicate at 50, 60, or 66 megahertz.

A 486DX2 chip communicates half as fast as it thinks. For example, a 50-megahertz 486DX2 chip communicates at 25-megahertz. That chip is said to be 50 megahertz internally, 25 megahertz externally. Since that chip thinks twice as fast as it communicates, itís called a clock-doubled chip.

A 486DX4 chip communicates a third as fast as it thinks. For example, a 75-megahertz 486DX4 chip communicates at 25-megahertz. Since that chip thinks three times as fast as it communicates, itís called a clock-tripled chip. Since the chip is clock-tripled, it ought to be called a "486DX3"; but Intel calls it a "486DX4" instead because Intel wants to pretend that itís better than "486DX3" chips manufactured by IBM.

Imitations

Intelís competitors have imitated Intelís chips. Some of the imitations go faster than Intelís originals!

8088 The most popular imitation of the 8088 is the V20 chip. Itís made by Nippon Electric Company (whose abbreviation is NEC, which is pronounced "neck"). At 10 megahertz, itís faster than Intelís original!

8086 The most popular imitation of the 8086 is NECís 10-megahertz V30 chip.

286 Imitations of the 286 are made by Harris and come in 16-megahertz and 20-megahertz versions.

386 Imitations of the 386SX and 386DX are made by Advanced Micro Devices (AMD) and come in 40-megahertz versions.

486 AMDís imitations of the 486 are excellent and come in 66-megahertz, 80-megahertz, 100-megahertz, and 120-megahertz versions.

Cyrix and IBM make awful 486 imitations that go much slower than Intelís originals and ought to be called "386Ĺ" instead of "486". Cyrixís imitation of the 486SX is called the 486SLC; Cyrixís imitation of the 486DX is called the 486DLC. IBMís imitation of the 486DX is called the Blue Lightning (BL).

Pentium classic Imitations of the Pentium classic are made by AMD and Cyrix.

They work much slower than Intelís Pentium and should be called "486Ĺ" instead. For example, a 133-megahertz AMD 586 goes about as fast as an 80-megahertz Pentium would go.

Pentium Pro Cyrix makes a Pentium Pro imitation. Cyrix calls it the 686, but cynics call it the "586Ĺ".

Pentium 2 Imitations of the Pentium 2 are made by AMD and Cyrix.

AMDís imitation is called the K6 MMX. The K6 MMX is slightly slower than a Pentium 2 but much faster than a Pentium MMX. Its price is ridiculously low: its quantity-1000 price is just $109 for 166MHz, $189 for 200MHz, $289 for $233MHZ. The K6 MMX 166MHz is an amazing bargain: its price is nearly as low as a Pentium classic 166MHz, but it performs better than a Pentium MMX 200MHz. Its ridiculously low price makes the K6 MMX the best value of chips made today: it gives you the most power per dollar. All reviewers love it and recommend it.

Half-assed systems

For a chip to accomplish anything at all, you must give it some work to do! If the chip must wait for you to tell it what to do, the chip accomplishes nothing useful during the wait: it just mumbles to itself.

So to make full use of a fast Pentium, make sure you know what commands to give the computer and make sure you help the chip reach its full potential by buying quick RAM, quick disk drives, and a quick printer. Otherwise, the Pentium will act as idiotic as if itís in the army: it will just "hurry up and then wait" for the other parts of the system to catch up and tell it what to do next.

A mind is a terrible thing to waste! To avoid wasting the computerís mind (the CPU), make sure the other computer parts are fast enough to match the CPU and keep it from waiting.

If you get suckered into buying a computer that has a fast Pentium chip but a slow RAM, slow disk drives, and a slow printer, youíve bought a computer thatís just half-fast; itís half-assed.

Total cost

When you buy a microcomputer, its advertised price always includes a microprocessor, motherboard, and other goodies. Pay for the microprocessor separately just if youíre inventing your own computer, buying parts for a broken computer, or upgrading your computer by switching to a faster microprocessor and motherboard.

Though the microprocessor is cheap, the computer containing it can cost thousands of dollars. Thatís because the microprocessor is just a tiny part of the computer. In addition to the microprocessor, you need memory chips, interface chips, and support chips; you also need PC boards to put the chips on; you also want I/O devices (keyboard, screen, printer, speaker, and mouse), disks, and software.

Discount dealers sell IBM clones for these prices:

Chip Complete computer

8088 or 8086 $100

286 $200

386 $300

486 $600

Pentium $1200

Those prices include nearly everything you need (such as the CPU, memory chips, disks, keyboard, and a screen that displays lots of colors) but do not include a printer or software. Those prices are approximate; the exact price you pay depends on the CPUís speed (how many megahertz) and on the other componentsí speed, quality, and size.

Notice that a 286 computer costs $200, which is $100 more than an 8086 computer. Thatís because a 286 computer includes a better CPU chip and also comes with a better keyboard, better screen, better memory chips, and better disks.

Motorola

Intelís biggest competitor is Motorola. It manufactures the 6809E microprocessor, the 68000 (which is faster and understands advanced commands), several souped-up versions of the 68000, and the Power PC:

Chip Price Computers that use it

6809E $3 Radio Shack Color Computer

68000 $9 Mac, Mac Plus, Mac SE, Mac Classic, Amiga (500, 600, 1000, 2000), Atari ST

68020 $45 Mac LC, old Mac 2, Amiga 1200

68030 lots Mac (SE/30, Classic 2, LC 2, LC 3), new Mac 2, Amiga 2500 & 3000

68040 lots Mac Centris, Mac Quadra, and Amiga 4000

Power PC lots Power Mac

Motorolaís microprocessors are not Intel clones. They use different commands than Intel and require different software.

When fed the proper software, they work as fast as Intelís microprocessors:

Motorolaís 6809E is about as fast as Intelís 8080 (which was the predecessor to the 8088)

Motorolaís 68000 is about as fast as Intelís 8086

Motorolaís 68020 is about as fast as Intelís 286

Motorolaís 68030 is about as fast as Intelís 386

Motorolaís 68040 is about as fast as Intelís 486

Motorolaís Power PC is about as fast as Intelís Pentium

Whatís the Power PC? Motorolaís fastest microprocessor, the Power PC, was invented by a team of researchers from three companies (Motorola, Apple, and IBM), all working together. Thatís why itís called the love-triangle chip. It was invented to prevent Intel from monopolizing the microcomputer marketplace.

The first version of the Power PC, called the Power PC 601, was manufactured just by IBM. Later versions (the Power PC 603, 604, and 604e) are manufactured by both Motorola and IBM.

The Power PC is used in Appleís fastest computer, the Power Mac.

Intel emulation Suppose your computerís microprocessor is made by Motorola, but somebody gives you software written for Intel microprocessors instead. You can run that software on your computer if you feed your computer an Intel emulator (software that makes Motorola microprocessors imitate Intelís). But Intel emulator software runs slowly. To accomplish tasks faster, buy software that runs directly on Motorola microprocessors without needing an Intel emulator.

Math coprocessor Want a Motorola math coprocessor? For the 6809E CPU, no math coprocessor is available. For the 68000 or 68020, buy the 68881 math coprocessor ($49). For the 68030, buy the 68882 math coprocessor ($69). The 68040 comes in two versions: the standard version (called the 68RC040) includes math-coprocessor circuitry; the stripped-down version (called the 68LC040) does not. The Power PC includes math-coprocessor circuitry.

In Motorola & Mac jargon, a math coprocessors is called a floating-point unit (FPU).

Accelerator cards To make an old Mac go faster, insert a CPU accelerator card, which contains a faster CPU chip. Here are your choices:

CPU chip included on card Cardís price

68030 33 megahertz $99

Power PC 604 150 megahertz $179

Power PC 604e 166 megahertz $239

Power PC 604e 180 megahertz $289

Power PC 604e 200 megahertz $469

Power PC 604e 225 megahertz $729

Power PC 604e 233 megahertz $799

You can get those cards from Sonnet Technologies (800-786-6260, 714-261-2800) and Source Technologies (800-900-4599, 954-725-9777) Before buying one of those cards, ask the salesperson whether itís compatible with your Mac.

Classic microcomputers

Primitive old microcomputers contain microprocessors invented by Zilog and MOS Technology. Theyíre not Intel clones.

Zilog, which was owned by Exxon, made the Z-80A microprocessor, which was super-cheap: it cost just $2! Itís in many obsolete computers, such as the Radio Shack TRS-80 models 1 & 2 & 3 & 4 & 12, the Kaypro 2 & 4 & 10, the Epson QX-10 & Geneva, the Timex-Sinclair 1000 & 1500, and the Coleco Adam.

The 6502 microprocessor was invented by MOS Technology, which became part of Commodore. It was also manufactured by other chip makers, and you could get souped-up versions that understood extra commands and went faster.

Chip Price Computers that use it

6502 $2 Apple 2 & 2+ & old 2e, Atari 800

65C02 $7 Apple 2c & 2c+ & new 2e

6510 $15 Commodore 64 & 128 & Vic

65C816 $17 Apple 2GS

The 65C02 and the 65C816 are made of CMOS; thatís why their names contain the letter C. The other chips in that table are traditional: theyíre made of NMOS.

How many pins?

A cheap microprocessor (such as an 8088, 8086, Z-80, 6502, or 6809E) comes in a DIP (caterpillar) that has 40 pins (20 pairs of pins).

Fancier chips have more pins. For example, the Motorola 68000 comes in a DIP that has 64 pins.

If a chip is even fancier (such as the 68-pin Intel 286 or the 132-pin Intel 386DX), it requires too many pins to fit in a DIP. Instead of coming in a DIP, the chip usually comes in a pin grid array (PGA), which is a square having many pins underneath it, as if it were a square porcupine lying on its back.

Memory chips

Although the CPU (the computerís brain) can think, it canít remember anything. It canít even remember what problem it was working on!

Besides buying a CPU, you must also buy memory chips, which remember what problem the CPU was working on. To find out what the problem was, the CPU looks at the memory chips frequently ó about a million times every second!

The part of the computerís main circuitry that contains the memory chips is called the main memory.

The typical memory chip comes in a DIP that has 8 pairs of legs (16 pins). In a typical microcomputer, the motherboard contains lots of memory chips.

If you buy extra memory chips (so that your computer can remember extra information), and the extra memory chips donít all fit on the motherboard, you must buy an extra PC card to mount them on; that extra card is called a memory card. If the memory card comes in a cute little cartridge that you can pop into and out of the computer easily, itís called a memory cartridge.

Warning: if you buy a memory chip or card or cartridge, and want to pop it into the computer, turn off the computerís power first. If you forget, and accidentally leave the power on while youíre inserting (or removing) the memory, you might wreck your computer!

You need two kinds of memory chips: RAM and ROM. The RAM chips remember information temporarily; the ROM chips remember information permanently. Letís begin by looking at RAM chips.

RAM

If a chip remembers information just temporarily, itís called a random-access memory chip (RAM chip).

When you buy RAM chips, they contain no information yet; you tell the CPU what information to put into them. Later, you can make the CPU erase that information and insert new information instead. The RAM chips hold information just temporarily: when you turn the computerís power off, the RAM chips are automatically erased.

Whenever the CPU tries to solve a problem, the CPU stores the problem in the RAM chips, temporarily. There it also stores all instructions on how to solve the problem; the instructions are called the program.

If you buy more RAM chips, the CPU can handle longer problems and programs. If the computer doesnít have enough RAM chips to hold the entire problem or program, you must split the problem or program into several shorter ones instead, and tell the CPU to work on each of the short ones temporarily.

How RAM is measured A character is any symbol you can type on the keyboard, such as a letter or digit or punctuation mark or blank space. For example, the word HAT consists of 3 characters; the phrase Mr. Poe consists of 7 characters (M, R, the period, the space, P, O, and E). The phrase LOVE 2 KISS U consists of 13 characters.

Instead of saying "character", hungry programmers say byte. So LOVE 2 KISS U consists of 13 bytes. If, in the RAM, you store LOVE 2 KISS U, that phrase occupies 13 bytes of the RAM.

RAM chips are manufactured by a process that involves doubling. The most popular unit of RAM is "2 bytes times 2 times 2 times 2 times 2 times 2 times 2 times 2 times 2 times 2", which is 1024 bytes, which is called a kilobyte. So the definition of a kilobyte is "1024 bytes".

Although a kilobyte is exactly 1024 bytes, the following approximations are useful.

A kilobyte is about a thousand bytes. Itís about how many characters you see on the screen of a TV computer. Itís about half as many characters as you see on the screen of an 80-column monitor. Itís about a quarter as many characters as you get on a typewritten page (assuming the page is single-spaced with one-inch margins and elite type).

The abbreviation for kilobyte is K. For example, if a salesperson says the computer has a "64K RAM", the salesperson means the main circuitry includes enough RAM chips to hold 64 kilobytes of information, which is slightly over 64,000 bytes.

A megabyte is 1024 kilobytes. Since a kilobyte is 1024 bytes, a megabyte is "1024 times 1024" bytes, which is 1,048,576 bytes altogether, which is slightly more than a million bytes. Itís about how much you can fit in a 250-page book (assuming the book has single-spaced typewritten pages). The abbreviation for megabyte is meg or M.

A gigabyte (pronounced "gig a bite") is 1024 megabytes. Itís slightly more than a billion bytes.

A terabyte is 1024 gigabytes. Itís slightly more than a trillion bytes.

In honor of the words "kilobyte", "megabyte", "gigabyte", and "terabyte", many programmers name their puppies Killer Byte, Make a Byte, Giggle Byte, and Terror Byte.

Rows of RAM chips In a cheap microcomputer (such as the Commodore 64), the RAM is a row of eight NMOS chips. That row of chips holds 64K altogether. So it holds 64 kilobytes, which is slightly more than 64 thousand bytes (since a kilobyte is slightly more than a thousand bytes).

That row of chips is called a 64K chip set. Each chip in that set is called a "64K chip", but remember that you need a whole row of those 64K chips to produce a 64K RAM.

Mail-order discount dealers charge 50Ę for a 64K chip. So to get 64K of RAM, you need a 64K chip set, which is a row of eight 64K chips, which costs "8 times 50Ę", which is $4.

The most popular style of 64K chip is the TI 4164. Although that style was invented by Texas Instruments, other manufacturers have copied it.

If your computer is slightly fancier (such as the Apple 2c), it has two rows of 64K chips. Since each row is a 64K RAM, the two rows together total 128K.

If your computer is even fancier, it has many rows of 64K chips. For example, your computer might have four rows of 64K chips. Since each row is a 64K RAM, the four rows together total 256K.

64K chips didnít become popular until 1982. If your computer was built before then, it probably contains inferior chips: instead of containing a row of 64K chips, it contains a row of 16K chips or 4K chips.

During the 1980ís, computer engineers invented 256K and 1M chips. The most popular style of 256K chip is called the 41256, which you can get from discount dealers for $2. A 1M chip costs $4.

If your computer has very little RAM, you can try to enlarge the RAM, by adding extra rows of RAM chips to the motherboard. But if the motherboardís already full, you must buy an extra PC card to put the extra chips on. That extra PC card is called a RAM memory card.

Parity chip The IBM PC and some clones contain an extra chip in each row, so that each row contains 9 chips instead of 8.

The rowís ninth chip is called the parity chip. It double-checks the work done by the other 8 chips, to make sure theyíre all working correctly!

So for an IBM PC or one of those clones, you must buy 9 chips to fill a row.

SIMMs and SIPPs If your computer is ultra-modern and you want to insert an extra row of RAM chips, you do not have to insert 8 or 9 separate chips. Instead, you can buy a strip that contains all 8 or 9 chips and just pop the whole strip into the computerís motherboard, in one blow.

The typical strip of chips is called a Single In-line Memory Module (SIMM) and pops into one of the motherboardís slots. If the strip pops into a series of pinholes instead, the strip is called a Single In-line Pin Package (SIPP).

Hereís what SIMMs cost:

$6 for a SIMM that holds ľ megabyte (which is 256 kilobytes)

$7 for a SIMM that holds 1 megabyte

$13 for a SIMM that holds 4 megabytes

$25 for a SIMM that holds 8 megabytes

$44 for a SIMM that holds 16 megabytes

$88 for a SIMM that holds 32 megabytes

You get those prices from discount dealers (such as Memory Express at 800-877-8188 or 818-333-6389). SIPPs cost $5 more than SIMMs.

Some computers use SIMMs containing a set of just 2, 3, or 4 chips. That set of chips is special and imitates 8 or 9 normal chips.

In old-fashioned computers, each SIMM fits into a motherboard slot by using 30 big pins. In computers that are more modern, each SIMM uses 72 big pins instead.

The typical SIMM contains chips that are fast: they retrieve information in 70 nanoseconds. (A nanosecond is a billionth of a second.) Old-fashioned SIMMs contain slower chips, requiring 80 nanoseconds; the fanciest SIMMs contains extra-fast chips, requiring just 60 nanoseconds. The newest SIMMs include circuitry called Extended Data Out (EDO), which transfers data from the SIMM to the CPU faster.

If you want to buy an extra SIMM to put in your computer, make sure you buy the same kind of SIMM as the other SIMMs that are already in your computer. Make sure the extra SIMM has the same number of pins (30 or 72?), the same number of chips on it (2, 3, 4, 8, or 9?), operates at the same number of nanoseconds (80, 70, or 60?), and uses the same technology (standard or EDO?).

Let your memory grow In a typical computer, the RAM contains several rows of chips, so that the total RAM contains several megabytes.

Hereís how much RAM you typically get altogether:

Computerís price Typical quantity of RAM

$50-$75 64K (64 kilobytes, 65,536 bytes)

$75-$100 128K (128 kilobytes, 131,072 bytes)

$100-$125 256K (256 kilobytes, 262,144 bytes)

$125-$150 512K (512 kilobytes, 524,288 bytes)

$150-$200 1M (1 megabyte, 1,048,576 bytes)

$200-$300 2M (2 megabytes, 2,097,152 bytes)

$300-$500 4M (4 megabytes, 4,194,304 bytes)

$500-$800 8M (8 megabytes, 8,388,608 bytes)

$800-$1,600 16M (16 megabytes, 16,777,216 bytes)

$1,600-$3,200 32M (32 megabytes, 33,554,432 bytes)

$3,200-$6,400 64M (64 megabytes, 67,108,864 bytes)

Mac The original Mac (nicknamed the Slim Mac) included 128K of RAM. Then came a version nicknamed the Fat Mac, which included 512K. Next came an improvement called the Mac Plus, which included 1M.

Those Macs are obsolete. All Macs sold today come with at least 4M, which is what you need to run modern Mac software.

Names of classic computers The Commodore 64 computer got its name because it contained 64K of RAM. Then Commodore invented an improved version, the Commodore 128, which contained 128K of RAM.

The Laser 128 imitates the Apple 2c. Each comes with 128K of RAM.

IBM The original IBM PC came with just 16K of RAM, but you could add extra RAM to it. Hereís how much RAM the typical IBM PC or clone contains now:

CPU Typical quantity of main RAM

8088 512K or 640K

286 640K or 1M

386 2M or 4M

486 4M or 8M

Pentium 16M or 32M

To run modern IBM PC software, you need at least 8M of main RAM. To run the FANCY modern IBM PC software WELL, you need at least 16M. Get 16M!

For computers having lots of RAM, hereís how itís divvied up.Ö

The first 640K of main RAM is called the base memory (or conventional memory). Thatís the part of the RAM that the computer can handle easily and quickly.

The next 384K is called upper memory. Itís relatively unimportant, since most programs donít know how to use it.

Those two parts (the conventional memory and the upper memory) consume a total of 640K+384K, which is 1024K, which is one megabyte.

The rest of the main RAM (beyond that first megabyte) can be either expanded or extended. Hereís the difference between "expanded" and "extended":

Expanded RAM is old-fashioned. Extended RAM is modern. (To remember that, notice that the word "expanded" comes before "extended" in the dictionary.)

Expanded RAM runs slowly. Extended RAM runs fast.

Expanded RAM can be added to any IBM-compatible computer. Extended RAM requires a modern CPU (a 286, 386, or 486) and will not run on an 8088 or 8086 CPU.

Modern programs work best if you have modern RAM (extended). Old-fashioned programs donít understand extended RAM; they understand just old-fashioned RAM (expanded). To run both kinds of programs, you need both kinds of RAM.

Some primitive programs use just the 640K of conventional RAM. They donít understand how to use expanded or extended RAM at all.

Expanded RAM and extended RAM are both built from the same kind of NMOS RAM chips. Whether a chip acts as "expanded" or "extended" RAM depends just on what other hardware and software you bought to control those chips.

If a chip acts as "extended" RAM, the CPU gets information from that chip directly and fast.

If a chip acts as "expanded" RAM, the CPU gets the chipís information by copying that information to the upper memory area. Then the CPU examines whatís in the upper memory area. That process is slow, since you must wait for the CPU to copy the chipís information to the upper memory area. That process was invented because itís the only way an 8088 or 8086 chip can handle RAM beyond a megabyte. Extended RAM is faster and simpler but requires a 286, 386, 486, or Pentium ó and is understood just by programs that are modern.

Hereís how to get expanded or extended RAM:

For an 8088 or 8086 CPU, the expanded RAM comes on an expanded RAM card. That card contains the RAM chips and the hardware necessary to control them. That card is expensive.

For a 286 CPU, you can buy an expanded RAM card, an extended RAM card (which is cheaper), or a combination card that you can switch between the two.

For a 386, 486, or Pentium, the RAM chips usually come on SIMMs. The CPU normally treats those RAM chips as extended RAM; but you can run a program that makes those RAM chips imitate expanded RAM so that old-fashioned programs can use them.

If you have a 386, 486, or Pentium and want to run even the fanciest software well, buy at least 16M of RAM. The computer will use the first megabyte for conventional RAM (640K) and the upper memory (384K). The computer will use the remaining 15 megabytes for extended RAM but make some of that extended RAM imitate expanded RAM.

A trio of companies (Lotus, Intel, and Microsoft) agreed on the technical details of how expanded memory should be handled. Their agreement is called the Lotus-Intel-Microsoft Expanded Memory Specification (LIM EMS). Expanded memory fitting their specification is called EMS memory. To manage that expanded memory, you need a special program, called the expanded memory manager (EMM).

The same trio of companies, working together with a fourth company (AST), developed an agreement on extended memory. Their agreement is called the Lotus-Intel-Microsoft-AST eXtended Memory Specification (or LIMA XMS). Extended memory fitting their specification is called XMS memory. To manage that extended memory, you need a program called the extended memory manager. The most popular extended memory manager is called "HIMEM.SYS".

The first 64K of extended memory is called the high memory area (HMA), because itís just slightly higher than the base memory and upper memory. (The rest of the extended memory should be called "even higher memory", but nobody does.)

NMOS RAM versus CMOS Most RAM chips are NMOS. The prices I quoted you were for NMOS.

If your computer operates on batteries, it uses CMOS instead, which consumes less electricity than NMOS. Unfortunately, CMOS chips cost more than NMOS. A 64K chip costs 50Ę if made of NMOS, but costs $4 if CMOS.

Dynamic versus static A RAM chip is either dynamic or static.

If itís dynamic, it stores data for only 2 milliseconds. After the 2 milliseconds, the electrical charges that represent the data dissipate and become too weak to detect. When you buy a PC board containing dynamic RAM chips, the PC board also includes a refresh circuit. The refresh circuit automatically reads the data from the dynamic RAM chips and then rewrites the data onto the chips before 2 milliseconds go by. Every 2 milliseconds, the refresh circuit reads the data from the chips and rewrites the data, so that the data stays refreshed.

If a chip is static instead of dynamic, the electrical charge never dissipates, so you donít need a refresh circuit. (But you must still keep the power turned on.)

In the past, computer designers were afraid that the dynamic RAMís refresh circuit wouldnít work, and used static RAM instead. But today, refresh circuits are reliable, and the most popular kind of RAM is dynamic NMOS. For example, the TI 4116, 4164, and 41256 are all dynamic NMOS.

Dynamic RAM is called DRAM (pronounced "dee ram"). Static RAM is called SRAM (pronounced "ess ram").

Static NMOS is still available. CMOS and bipolar are always static.

Bipolar cache In a maxicomputer, minicomputer, or fancy microcomputer, the RAM is divided into two sections. One section is huge, contains many rows of NMOS chips, and is called the main RAM. The other section is tiny, contains just a few bipolar chips, and is called the cache (which is pronounced "cash").

The cacheís bipolar chips work much faster than the main RAMís NMOS chips.

In most IBM clones containing a 486DX or Pentium, the NMOS chips retrieve information in 60 or 70 nanoseconds, and the bipolar chips take 15 or 20 nanoseconds.

The typical bipolar chip holds 32 kilobytes and costs $5. Thatís a lot to pay for just "32 kilobytes", which is 1/32 of a megabyte! Yes, bipolar chips are pricey!

The typical computer contains a few bipolar cache chips.

In the typical IBM clone containing a 486DX,

the main RAM holds 4M or 8M, but the cache chips total just 128K or 256K.

In the typical IBM clone containing a Pentium,

the main RAM holds 16M or 32M, but the cache chips total just 512K.

So the bipolar cache is a memory thatís super-fast, small, and pricey.

In the bipolar cache, the computer keeps a copy of the main RAMís info that youíve been using recently, so the CPU can grab that info again super-quickly.

ROM

If a chip remembers information permanently, itís called a read-only memory chip (ROM chip), because you can read the information but canít change it. The ROM chip contains permanent, eternal truths and facts put there by the manufacturer, and it remembers that info forever, even if you turn off the power.

Hereís the difference between RAM and ROM:

RAM chips remember, temporarily, info supplied by you.

ROM chips remember, forever, info supplied by the manufacturer.

The typical computer includes many RAM chips (arranged in rows) but just a few ROM chips (typically 6).

What kind of info is in ROM? In your computer, one of the ROM chips contains instructions that tell the CPU what to do first when you turn the power on. Those instructions are called the ROM bootstrap, because they help the computer system start itself going and "pull itself up by its own bootstraps".

In the typical microcomputer, that ROM chip also contains instructions that help the CPU transfer information from the keyboard to the screen and printer. Those instructions are called the ROM operating system or the ROM basic input-output system (ROM BIOS).

In the typical microcomputer, one of the ROM chips tells the computer how to make each character on the screen out of dots. That chip is called the character generator.

In famous old microcomputers, several ROM chips contain definitions of fundamental English words, which are called BASIC words. For example, those ROM chips contain the definitions of BASIC words such as PRINT, NEW, RUN, LIST, GO, TO, END, STOP, INPUT, IF, and THEN. Those BASIC definitions in the ROM are called the ROM BASIC interpreter.

Commodore 64 For example, letís look inside a primitive computer: the Commodore 64. It contains just four ROM chips. The first chip contains 8K, for the ROM bootstrap and ROM BIOS. The second contains Commodoreís 8K ROM BASIC. The third contains Commodoreís 4K character generator. The fourth contains ľK that tells the computer how to make the screen produce pretty colors.

IBM In the typical IBM PC or clone, the motherboard contains a ROM BIOS chip. That chip contains the ROM BIOS and also the ROM bootstrap. If your computer is manufactured by IBM, that chip is designed by IBM; if your computer is a clone, that chip is an imitation designed by a company such as Phoenix. Such a chip designed by Phoenix is called a Phoenix ROM BIOS chip. Other companies that design ROM BIOS chips for clones are American Megatrends Incorporated (AMI), Award (a smaller company), and Quadtel (which is now owned by Phoenix.)

On a special PC card (called a video display card), youíll find a ROM chip containing the character generator.

If your computer is built by IBM, some chips on the motherboard contain the ROM BASIC interpreter. If your computer is a clone, all of BASIC comes on a disk instead of in ROM chips.

Altogether, the original IBM PC contained six ROM chips: the ROM BIOS chip, the character generator, and four ROM BASIC interpreter chips. Each of those six chips contained 8K, so that the computerís ROM totaled 48K. On newer computers from IBM and clones, the total is slightly different.

Extra ROM chips Some microcomputers include extra ROM chips that tell the computer how to handle specific applications, such as word processing and accounting.

ROM cartridges If your computer attaches to a TV and is old-fashioned (such as a Commodore Vic, Commodore 64, Commodore 128, Atari 800, Atari 800XL, or Radio Shack Color Computer), you can pop ROM cartridges into the computer. A ROM cartridge is a cartridge containing a PC card full of ROM chips. Etched into those ROM chips is a program.

The typical ROM cartridge contains a program that plays a video game, such as Space Invaders or Pac Man or computer chess. You can also buy ROM cartridges that contain programs for word processing, music, art, or tutoring you. Each ROM cartridge costs about $30.

How ROM chips are made The info in a ROM chip is said to be burned into the chip. To burn in the info, the manufacturer can use two methods.

One method is to burn the info into the ROM chip while the chipís being made. A ROM chip produced by that method is called a custom ROM chip.

An alternate method is to make a ROM chip that contains no info but can be fed info later. Such a ROM chip is called a programmable ROM chip (PROM). To feed it info later, you attach it to a device called a PROM burner, which copies info from a RAM to the PROM. Info burned into the PROM canít be erased, unless the PROMís a special kind: an erasable PROM (EPROM).

To erase a typical EPROM, shine an intense ultraviolet light at it for 20 minutes. Thatís called an ultraviolet-erasable PROM (UV-EPROM).

A fancier kind of EPROM can be erased quickly by sending it a 25-volt shock for a tenth of a second. Thatís called an electrically erasable PROM (EEPROM) or electrically alterable PROM (EAPROM).

After you erase an EPROM, you can feed it new info.

If youíre a manufacturer designing a new computer, begin by using an erasable PROM (EPROM), so you can make changes easily. When you decide not to make any more changes, switch to a non-erasable PROM, which costs less to manufacture. If your computer becomes so popular that you need to manufacture over 10,000 copies of the ROM, switch to a custom ROM, which costs more to design and "tool up for" but costs less to make copies of.