The Chip at 30: Potential Still Vast

Business Technology

The Chip at 30: Potential Still Vast

By John Markoff
The New York Times

September 14, 1988

Thirty years after the invention of the computer chip, engineers see no end to the electronic marvels it can generate.

The amount of circuitry that can be etched into the silicon wafers, of which the chips are made, is growing steadily and now stands at 10 million transistors. At the beginning, the chips contained just two transistors. By the turn of the century, experts believe the number will surpass one billion transistors for the same size wafer.

The chips, which are known more formally as integrated circuits and are generally the size of a fingernail, can now perform electronic operations as fast as four-billionths of a second. A decade from now that speed is likely to increase to one operation every 200-trillionths of a second.

Costs Trimmed Sharply

This miniaturization and increased speed have caused the cost of computing power to drop a thousandfold in the last two decades, a decline that will continue unabated and perhaps even accelerate, engineers have said.

Many of the things that have grown from the integrated circuit are now familiar - personal computers, hand calculators, pocket television sets and compact disk players. Others are less obvious, like miniature computers that control automobile engines and washing machine control units.

Predicting where the miniaturization will lead is more difficult, but electronics industry experts say that advances will be rapid. Such things as machines that can listen to spoken language and translate it and devices that significantly extend man's senses are often cited. Computers of enormous power but far smaller than any present machine will probably be found performing tasks throughout society.

Although experts had questioned whether there are natural barriers to the further evolution of microelectronics, the consensus now seems to be that limitations are only distant possibilities, at best.

''Most of the people who have tried to set limits have missed pretty badly,'' said Jack Kilby, the retired Texas Instruments engineer, now 64 years old, who was one of the two inventors of the integrated circuit in 1958. ''It seems to be a much more wide-open proposition than it was 10 years ago.''

The other inventor was Robert Noyce of the Intel Corporation. The two men developed the first integrated circuit working independently. Mr. Kilby received the first patent in February 1959, but six months later Dr. Noyce obtained a patent for an improved process that ultimately proved vital to the development of the semiconductor industry.

''Virtually everything we think about today, and can't quite do, will happen,'' said Gordon Bell, a vice president at the Ardent Computer Corporation, and the chief designer of the Digital Equipment Corporation's first VAX computer. ''We will have an enormous amount of computing power to simulate everything. We will be able to teach people how to drive cars, simulate materials, chemical reactions and physical phenomenon.''

Today, no one is certain what a billion-transistor integrated circuit that performs an operation every 200-trillionths of a second will be able to do.

But the question is of far more than prosaic interest. It strikes at the very heart of modern technological development and will have a significant impact on the future innovation and the economic fortunes of corporations and countries for many years.

A long list of technologies is likely to emerge. Almost certainly there will be machines that can listen and see far better than those that exist today, as well as advances in artificial intelligence and highly capable robots.

The consensus among computer designers is that despite the remarkable list of new products and new industries that have emerged based on microelectronic technologies, the surface has barely been scratched.

In Silicon Valley, the area south of San Francisco where many microelectronics advances have occurred and which has taken its name from the silicon used in computer chips, the integrated circuit has spawned industries that have marked plateaus in the evolution of electronics: the LCD watch, then the hand calculator and finally the personal computer.

Now many computer designers believe that a new industry is percolating in Silicon Valley and other technological centers that may match or surpass the personal computer in its social and economic impact. More than half a dozen new start-up companies and existing concerns, like Apple Computer Inc. and the Hewlett-Packard Company are trying to develop a new kind of portable computer that will appear less as a computer and more as an electronic notebook.

Such a computer might use a powerful microprocessor to recognize handwriting, permitting designers to dispense with a keyboard. Such a change would make it possible to use a computer by writing instead of typing. That development would offer the power of a personal computer to a far broader audience.

Three decades after their invention, integrated circuits have advanced so greatly in design that a map of one approaches the complexity of a street map of the entire United States. Yet chip density - the number of transistors or other components on each piece of silicon - continues to grow steadily, quadrupling every three years.

Moreover, the rate of increase, although slowing slightly in recent years, shows no sign of stopping, leading to the predictions of one billion transistors by the year 2000.

Peering further into the future, most semiconductor experts believe that this tempo can be maintained virtually indefinitely as new technologies emerge to circumvent existing barriers and allow the cramming of even more components onto a single chip.

Despite such confidence, semiconductor designers will have to overcome enormous barriers to make such predictions come true.

An important barrier will be the increasing difficulty that chip designers face in contemplating circuits that contain millions of transistors. ''The most difficult thing for a very-large-scale circuit designer is getting his mind around the system,'' said a Silicon Valley computer scientist.

Errors that creep into chip design because of the complexity of the integrated circuits can be costly. Soon after Intel began delivering its 80386 microprocessor last year, a user noticed that the processor sometimes made a subtle math error. The trouble was traced to one improperly designed transistor out of more than a quarter of a million. It took two weeks of round-the-clock work to amend the design and produce revised chips, and months to replace faulty chips that had been shipped.

A worrisome problem is that progress in advanced design tools has lagged behind chip density increases in recent years. It is easier to design memory chips, which are repetitive circuits. But other kinds of chips, like microprocessors, where circuitry is laid out in an irregular fashion, will require a new generation of sophisticated computer software that will aid human engineers in the design of new chips.

A second barrier is that as designers approach fundamental size limits, increases in speed will be increasingly harder to achieve. As this happens, computer designers will have to turn in other directions, searching for better circuit designs and new materials.

To speed up integrated circuits, some manufacturers have turned to alternative materials, like gallium arsenide. Transistors made from it switch much faster than traditional silicon-based circuits. Cray Research Inc. is betting on gallium arsenide technology to speed its next generation of supercomputers. But semiconductor makers believe that silicon will continue as an industry mainstay except in specialized areas like military and communications applications, where high speed is essential.

Another promising but still speculative technology concerns superconducting materials, through which electricity can move with no resistance, thus increasing speed. Researchers have deposited superconducting materials on top of silicon wafers and dramatically increased the speed of experimental integrated circuits.

''If we're going to get faster computers, we're going to have to get them in other ways that the tremendous advances of speedup that integrated circuitry has delivered for the past three decades,'' said David Patterson, a University of California computer scientist.

To build denser chips, semiconductor engineers have also begun to build vertically as well as horizontally along the chip surface.

''The analogy is to building skyscrapers,'' said James D. Meindl, senior vice president for academic affairs and provost at Rensselaer Polytechnic Institute in Troy, N.Y. ''We ran out of real estate for one-story buildings. In a sense there were no more lots left on the chip.''

For example, I.B.M.'s most advanced memory chip, capable of storing four million bits of information, uses a device called a trench capacitor. The capacitor is essentially tipped up on its side on the chip to save space, according to Michael Attardo, a vice president at I.B.M. and president of the company's general technology division.

He said that current manufacturing technology had the potential to produce chips containing as much as 64 million bits of information, or the contents of several encyclopedia volumes. Beyond 64 million bits, chip makers will have to turn to more exotic manufacturing techniques that would use X-rays or other approaches. That technology is still under development.

There is also an effort afoot to increase the area of the current fingernail-size chips, which would advance the quest to pack more transistors onto the surface of a single semiconductor.

Ever-faster and denser chips are likely to cause disruptions in industry and the workplace as the cost of computing power falls and it becomes economical to use them widely. For example, take the computer industry itself. What will happen to the giant mainframe computer, the venerable standard of the computing industry?

Executives at I.B.M. argue that despite the increasing power of the single-chip microprocessor, the mainframe computer, a multimillion-dollar machine based on thousands of chips, will continue to serve a useful role.

But at many Silicon Valley computer and semiconductor companies computer scientists and engineers say they believe that one of the important consequences of the advance of semiconductor technology will be to make the mainframe computer obsolete.

''It's not going to make sense to create these dinosaurs,'' said John Moussouris, a former I.B.M. physicist who co-founded the Mips Computer Company in Sunnyvale, Calif. ''Over the next 10 years it will be impossible to build a machine that's faster by building a machine that is bigger. If you attempt to build traditional machines you will find yourself building bigger, more expensive and slower machines.''

Such arguments leave Mr. Kilby, the chip pioneer, bemused. He said he often has trouble drawing parallels between his early invention and today's very large-scale integrated circuits. ''In a peculiar way, the things that people talk about as modern miracles are very different from what I did 30 years ago,'' he said.

GRAPHIC: graph of increasing performance and falling prices of computer chips (Sources: Siemens AG: I.B.M. Corp and Gartner Group) (pg. D1); photos of the I.B.M. memory chip; Jack Kilby of Texas Instruments (pg. D10)