Highly Integrated Digital Chips
Once the trend toward integrating transistors and other discrete components into larger integrated components was under way, there was no stopping it. Today, big chips contain millions of gates of logic, which implies millions of transistors, resistors, diodes, capacitors, and other basic components all rolled together.
Such highly integrated chips are generically called VLSI components. There's no specific test for a chip to be considered a VLSI device; anything more than about a million transistors probably qualifies. By that measure, most new chips qualify as VLSI components.
These large-scale digital chips take many forms, from memory chips to microprocessor chips. These are the most complex and elaborate semiconductor components ever created, and their profitability reflects their advancements. Top-of-the-line microprocessor chips can sell for more than $1,000 each. They use the most advanced silicon manufacturing technology and take teams of engineers many years to design and develop.
Memory Chips
Memory chips are the most plentiful kind of large-scale digital chips because so many products use so many of them. Your PC probably has at least eight memory chips inside, but it could have 64 or more. Microwave ovens, CD players, alarm clocks, and nearly any other electronic device you can think of, probably all have at least one memory chip inside them.
The world's semiconductor makers produce about 10 billion memory chips every year. As complex as they are, memory chips are nearly commodities, priced very aggressively and sold in enormous volumes. Prices for memory chips fluctuate daily in electronics markets, and analysts track memory prices like treasured stock portfolios. Over the years, memory production gradually moved from the United States to Japan, and then to Korea, Taiwan, and Singapore. Because memory chips are so price sensitive, their production tends to shift to the cheapest labor pool able to manufacture them.
What do memory chips do? Obviously, they remember, or store, electrical information. That's not as eerie or as complicated as it sounds. Memory chips are basically arrays of tiny capacitors, those simple analog components for storing electricity. Because the capacitors are very small and arranged in very tight rows, memory chips can hold millions of separate electrical charges at once.
Memory chips come in many different subtypes, which are covered in Chapter 7, “Essential Guide to Memory.” That chapter also explains more about how memory chips work. The two major memory classifications are RAM and ROM. RAM chips can be erased and reused to store different data, whereas ROM chips are permanent, holding data forever.
Communications Chips
Communications chips handle data communications between computers, satellites, and buildings. They were once simple “bit pumps” that had no intelligence of their own; they just squirted data down a wire under the control of a computer. Now communications chips have become like microprocessors in their own right, with the intelligence to format, buffer, and massage data before it's transmitted or unpacking data as it's received. As the world's desire for instant communication has increased, the newest generations of communications chips have risen to the challenge.
You'll find specialized communications chips atop telephone poles and in cellular base station antenna “trees.” They compress and decompress voice conversations on the fly, converting our voices into densely packed (and sometimes encrypted) data packets before sending them down copper wire or optical fiber. On the return trip, these same chips decompress (and decrypt) our friends' voices and reconstitute them into audio form. As these communications chips get more advanced, telephone companies are able to squeeze more and more simultaneous conversations onto the same wires or fibers.
With the rise of the Internet and the Web, the market for communications chips has taken off. So much data is now transferred through the Internet that intelligent and specialized chips are required at several steps along the way to make sure your e-mail gets to the right place. Some chips perform a sort of Internet triage, separating the high-priority data traffic (e.g., a streaming video broadcast) from the low-priority ones (e.g., e-mail spam). Other chips are used for security, encrypting and decrypting credit card numbers as they pass through the network. The list of communications chips is almost endless and the business of creating them is volatile and fast moving.
Graphics Chips
Graphics and 3D chips have surged to become a major part of the PC components business since the 1990s. As PCs are used more and more for games, DVD video, and other television-like tasks, the importance of graphics chips has risen. The latest graphics chips are amazingly complex and powerful, using millions of transistors and rivaling some microprocessors for performance and intricacy. In many live-action computer games, the 3D graphics chip works harder than the microprocessor in the PC. Avid PC gamers agonize over what 3D graphics chip to buy and diligently read every product review that comes out.
Apart from PCs, graphics chips are used in home electronics, such as DVD players. High-end computer systems use graphics chips to render complex information for weather forecasting; aircraft design; stress analysis; jet simulation; or nuclear, chemical, and medical research.
The performance and capabilities of graphics chips have risen dramatically over the years. Early graphics chips were little more than a D/A converter and some memory. The computer put data into the memory and the D/A converter turned it into an analog electrical signal for the screen. Now 3D graphics chips rival the microprocessor they're serving. They render 3D images from mathematical models, generate millions of colors, calculate angles of refraction based on light sources and intensities, hide overlapping objects behind one another, and fade distant objects into the background. The mathematics involved in creating 3D scenes is complex and arcane, and the circuitry involved in making all this happen 60 times per second is impressive. It's a lot of technology all devoted to entertainment.
Peripherals
Peripheral chips are so named because they are used around the outside edges, or periphery, of computers. Peripheral chips are the gatekeepers or customs agents of a computer, channeling incoming data where it belongs and sending outgoing data on its way.
Peripheral chips control printers, keyboards, speakers, mice, joysticks, or anything else you might plug into your computer. Each peripheral chip specializes in one type of plug-in, so a well-equipped computer will include several different peripheral chips. In addition to computers, peripheral chips are also used in industrial and scientific systems, or anywhere there's a computer connected to something outside in the real world.
The number and type of peripheral chips change over time as people invent new ways to use computers. Keyboard chips have been available for a long time, but chips for controlling joysticks came later. A few years ago, chips for universal serial bus (USB) and FireWire, two new PC connections, entered the market. Although peripheral chips are fairly complicated, they're also pretty inexpensive. They're considered little better than commodities by computer makers, so their prices tend to be low and their life spans short.
Custom Chips: ASICs and ASSPs
Big teams of engineers working at huge semiconductor companies design most large-scale chips. Chip design is difficult work, and the costs are high. However, over the past 10 years or so, the barriers to chip design have crumbled, little by little. Designing a chip is still not easy by any means, but it's not the mysterious black art it once was. It's now within the realm of possibility for a medium-sized company to consider designing its very own chip.
Why would you want to? Well, maybe none of the major chip-making companies produce just the chip you're looking for. Maybe you want a memory chip with a certain capacity, or a communications chip that transmits data a certain way, or a security chip that encrypts information using your special password or code system. There are as many reasons for making a custom chip as there are companies considering it.
The term for these custom chips is application-specific integrated circuit (ASIC). An ASIC is a general term for any chip that is custom designed, as opposed to a chip that's mass-produced by one of the major chip companies. You don't have to sell your custom ASIC to anyone. In fact, that's usually the point. An ASIC gives a competitive advantage to the company that makes it; it's a secret weapon inside the television, microwave oven, or antilock brake system.
What does “application-specific” mean? The word application in this sense means product. It's another unfortunate case of a common English word that's been perverted somewhat by the technical community. Application-specific means the chip is intended for some particular type of product, like elevators or steam whistles. It also implies the chip isn't very useful outside of that specific market.
ASICs by their nature are custom, so every one is different. An ASIC can be big or small, complex or (relatively) simple. It might include any combination of components from the family tree near the beginning of this chapter. An ASIC might have one or more microprocessors inside, some memory, some analog components, some programmable logic, or nearly anything else.
Designing and building an ASIC isn't cheap, so companies don't undertake the task lightly. It can easily cost $1 million in engineering time, materials, and services before the first chip is ready. Companies want some guarantee that the benefits of their ASIC will outweigh the risks and costs of its development. For starters, that means ASICs are created only for high-volume products like digital cameras, cell phones, automobiles, and so forth.
When an ASIC is designed for sale, as opposed to internal use, it's called an application-specific standard product (ASSP). The “standard product” part implies that the chip is available to anyone who wants to buy it, a sort of noncustom custom chip. There's no technical difference at all between an ASIC and an ASSP; the distinction is purely economic.
In addition to ASIC and ASSP, custom chips are sometimes called a system-on-a-chip (SoC). This term doesn't have any specific meaning; it's used to vaguely describe a particularly large or complex ASIC that includes everything the final product will need, all in a single chip. The world of ASICs, ASSPs, and SoCs is mapped out in more detail in Chapter 8, “Essential Guide to Custom and Configurable Chips.”
Programmable Logic Chips
Many companies would like to develop their own ASIC but don't have the stomach—or the deep pockets—for it. For them, there is a middle ground, a kind of semicustom chip. These chips are called programmable logic devices (PLDs). Different PLD companies use different terminology, so you'll also hear them called complex programmable logic devices (CPLDs) and field-programmable gate array (FPGA). These and other types of custom chips are covered in Chapter 8, “Essential Guide to Custom and Configurable Chips.”
Programmable logic chips are like electronic Etch-a-Sketches, ready to use but not really finished. Customers imprint their own design onto them, making their own semicustom chips. Customizing a programmable logic chip is nowhere near as difficult as designing and building a “real” ASIC chip from scratch. Like an Etch-a-Sketch, you can erase the chip and start over any time you want, so there's no risk.
Programmable logic chips look like normal chips on the outside but they're “blank” on the inside. They give designers and engineers a chance to play “what if” with theoretical chip designs. In the space of an afternoon, you can create and test half a dozen different chip designs on the same programmable logic chip. PLDs provide a blank canvas for engineers to be creative, and a washable one, too, because a single programmable logic chip can be used over and over. When the design is finally complete, the chip can be permanently installed into a product, where it will operate like any other chip.
There are downsides to all this flexibility. For one, programmable logic chips are slower than other chips. The exact same chip design will run slower in a PLD than it would in an ASIC manufactured by the normal means. Programmable logic chips are also more expensive than their ASIC counterparts. Companies that expect to ship thousands and thousands of chips shy away from using PLDs because of their higher unit price. On the plus side, there's no million-dollar development cost for a PLD. So although PLDs are definitely cheaper in the beginning, they might be more expensive in the long run.
Microprocessors
Microprocessors are the pinnacle of semiconductor development, or at least that's what the microprocessor makers will tell you. They are the “computer on a chip” so beloved by the popular press. There's no denying that microprocessors are the most complex and elaborate of all semiconductors or that they've forever altered our economic, social, and scientific worlds.
Microprocessors have enabled cheap, anywhere, anytime computing. Computers used to be large, noisy machines owned by governments and tended by teams of scientists in white lab coats. Now they're in every modern home, car, and telephone, ubiquitous to the point of invisibility. Computers can be cheaper than the batteries they run on. All of this has happened in the span of one human generation.
The first microprocessor was invented in 1971; now 1 billion are sold every few months. Surprisingly, almost none of those microprocessor chips are used in computers. Most people equate microprocessors with PCs but that's far from the whole picture. Computers are just a tiny fraction of what microprocessor chips are used for. Microprocessors have made a far greater impact in totally unrelated household, entertainment, security, and industrial uses than anyone could have predicted. The computer on a chip actually has very little to do with computers anymore.
Microprocessors have shown themselves to be amazingly adaptable. Like tiny robots, they can be commanded (programmed) to do any number of chores from the mundane to the complex. Instead of predicting the weather or calculating missile trajectories, microprocessors are used more commonly to cook food, turn on the furnace, open garage doors, play music, or blast imaginary space aliens. The average household has about two dozen microprocessors or more, without even counting home computers. The average car includes another dozen microprocessors; some have more than 60. Nearly anything that plugs into the wall or uses batteries is likely to have a microprocessor chip in it.
Microprocessors come in all shapes and sizes, and not all of them are the big, fast chips we see advertised on television. In fact, few microprocessor chips cost more than $5. The overwhelming majority of microprocessor chips are squirreled away inside everyday household appliances and toys. Microprocessor makers have started to specialize, so some chips are good for home appliances, some are for automobiles, some are for video games, and so on. Each variation has its own name or abbreviation.
There are different types of microprocessors, and they all have three-letter abbreviations like CPU, MPU, MCU, DSP, or NPU. For now we'll treat them all the same and just call them microprocessors. For more in-depth information, turn to Chapter 6, “Essential Guide to Microprocessors,” to see what these chips are all about.