Read this before you bet on the Outer Limits of Computing.
Chip technology mutates by megatrends and each new generation of chips instantly antiquates the most recent state-of-the-art version. Intel's new Pentium packs millions of transistors onto a single chip, transforming your computer into the rival of many a mainframe. Had automobile technology advanced at a similar pace over the past 20 years, your car would travel 500,000 miles an hour, get a million miles to the gallon, and only cost a measly $1,000.
But is there life after Pentium? Despite the dizzying evolutionary pace, chip technology may be running into a silicon wall. Semiconductor circuits presently incorporate tiny threads of precious metals, some as small as 1 micron wide - that's a millionth of a meter. But the current technology may be approaching a barrier of .25 microns that will be difficult to overcome. This barrier will limit the number of transistors that can be placed on a single chip when using traditional design and manufacturing techniques.
A superchip breakthrough is imminent, though, and promises to offer all you digitally attuned folks a sustained tempo of mind-boggling technological treats. In addition, superchips will do more than enhance speed; they'll open doors to new computing (and existential) possibilities, such as multimedia creations (audio and video), speech recognition, artificial intelligence, synthetic secretaries, and virtual reality. A single superchip could be programmed to simulate a diverse range of hardware devices. Modems, fax modems, network cards, sound cards, and video cards may fall the way of the slide rule.
Two trends support this prognostication of sustained rapid evolution: First, the granddaddy of chips, Gordon Moore, predicted in 1975 that the maximum possible number of transistors - the basic building block of computer chips - would double every two years. His claim was ridiculed at the time, but he turned out to be almost right: The number of transistors per chip has increased about 40 percent each year.
Secondly, by making transistors smaller, one can make chips that run faster while using less power. This increase in the number of transistors on a chip has souped-up operating speeds to supersonic levels. Today's densest CPU chips, which are clocked at up to 100-Mhz and have 32-bit data paths, contain just a few million transistors per chip and run at speeds nearing 100-million instructions per second. Even within the limits of current technology, the year 2000 could bring us 500-Mhz, 64-bit chips packed with more than 100-million transistors that could cruise at more than 2-billion instructions per second.
RAM, the basic component of computer memory, will undergo a similar growth in capacity. Semiconductor manufacturers claim that they will be able to produce gigabit (a billion bits) RAM chips within the next 10 years. By comparison, today's RAM chips just recently hit the 16-Mbit, or 16-million-bit, level. (An interesting aside: one of those manufacturers also predicts that by the year 2000, most RAM will go into HDTV sets, not traditional computers.)
But even those estimates may be conservative. One near-term breakthrough aims to develop a new breed of superchip that crams several separate chips into one package. This technique can be used to create either multi-layered 3-D chips or multi-chip modules.
The multi-layered approach stacks chips on top of one another. This type of 3-D cube configuration provides symmetrical multiprocessing at a relatively low cost. It's cheaper to attain 400-MHz performance levels by stacking four 100-MHz processors in a single chip, for example, than it would be to produce a single 400-MHz processor - something that can't be done with current technologies anyway.
Multichip module technology melds a variety of different chips into a single package. This technique could help eliminate bottlenecks in system performance by placing items such as RAM, video, and input/ouput (I/O) support in the same chip package as the processor, bypassing any slowdowns imposed by the computer's motherboard circuitry or expansion bus.
A far more radical approach to chip development argues for the abandonment of electronics in favor of photonics - a technology based on the use of photons, the basic particles of light. Digital photonic processors have already been demonstrated by scientists at AT&T's Bell Laboratories, but commercial applications are still many years away. AT&T's experimental photonic processor uses the world's smallest laser to send light through a chip composed of many microscopic lenses and mirrors etched into quartz.
Photonic processors outshine electronic chips in several ways. Most important, light can carry more information than electricity, and more quickly. AT&T scientists estimate that photonic processors will process more than 1,000 times as much information as today's most powerful supercomputers.
And because light beams can pass through each other without affecting the physics or path of the beam, they can be packed more densely than electronic circuitry. Should photonic processors become available, the idea of today's 16-bit or 32-bit buses would appear quaint compared to a photonic data bus that could be as much as 10,000 bits wide.
With photonic processors, one can also expand the capabilities of the chip's data paths - a profoundly important feature. The biggest performance bottleneck of future systems will not be the speed of the processor, but rather the speed of getting information on and off the chip. With photonic processors, a computer could have an optical backplane that provided more than 1,000 I/O channels, each running at speeds of more than a gigabit per second, compared with the 5-Mbyte-per- second speed of standard personal computer buses.
And for a quantum leap forward in chip technology, expect to see major breakthroughs in quantum mechanic electronics, which offer awesome potential for making one's chips fly. Quantum mechanics is the branch of physics that deals with the minute particles that compose atoms. The electronics part of quantum mechanics exploits the unique properties of subatomic particles, which can be packed a thousand times more densely than the precious metal circuits used in contemporary chip-making. By manipulating subatomic particles, designers hope to create terabit (one trillion bit) RAM chips as well as teraherz processors.
In order to build quantum electronic chips, scientists are developing nanotechnology, a process that creates minuscule machines from individual atoms. These "molecular machines" will be able to manipulate other atoms and particles to create almost anything imaginable: from other molecular machines to quantum chips, new chemicals, or tiny robots that could roam inside our bodies to fight disease and perhaps reverse aging by reprogramming our DNA. Incorporated into chip technology, these molecular machines will enable computer users to someday reach levels as yet "measureless to man."
