Six computer scientists take a look into the future. What's in store? Think speed.
In the world of nuclear physics, where moving at Internet speed still isn't fast enough, scientists are sending data from the CERN particle physics lab in Geneva, Switzerland, to the California Institute of Technology at the rate of a CD's worth of information every second. By 2007, they hope to double that to a gigabyte of data every second, fast enough to send a DVD movie in three seconds. Even at those speeds, it would still take 40 minutes to transfer a trillion bytes of information--the yardstick particle physicists use to measure the information their instruments spew. Scientists now collect a few terabytes of data a year, but that could increase a thousandfold by early in the next decade.
Contrast that with the technology many of us have at home today. With a dial-up Internet connection, it would take about two years to move a terabyte of data to your house, says Jim Gray, a distinguished engineer at Microsoft Research who's working with CERN and Caltech on the high-speed project. "I'm trying to get things that run in hours or days or weeks to run in seconds," says Gray, a specialist in making huge databases hum, whose resumé stretches back nearly 40 years, including work at Bell Labs, Digital Equipment, and IBM. People want answers in real time, Gray says. Slowness, "makes you much more reluctant to ask questions."
Carbon nanotubes, plastics semiconductors, and more esoteric areas of research such as using the spin of electrons are being examined as successors to silicon, IBM's senior VP and head of research Paul Horn says.
For its 25th anniversary issue, InformationWeek asked six leading computer scientists--Gray; IBM senior VP and head of research Paul Horn; Hewlett-Packard senior VP of research and HP Labs director Dick Lampman; Sun Microsystems executive VP and chief technology officer Greg Papadopoulos; Intel senior VP and CTO Pat Gelsinger; and Palo Alto Research Center president and director Mark Bernstein--to look ahead, to identify the ways the computer industry is likely to change, or needs to change, during the next decade. If any single theme emerged, it's speed--and the desire for it.
For these gentlemen and the computer industry in general, slowness stands in the way of greater future achievements. Microprocessor speeds are flattening after years of phenomenal gains. PCs can't find us the information we need fast enough. Supercomputer users thirst for faster "time to insight" from their complex machines. Meanwhile, the explosion of technology patents confounds companies, making it more challenging for them to assemble all the pieces they need to bring innovative products to market quickly. And America's universities attract fewer students interested in science and technology, as Asia and India shine in this area, a development that could slow U.S. competitiveness.
For an in-depth look at the big changes afoot in computer design, the office of the future, the wired home, intellectual property, and education and globalization, read on. Just make it fast.
What's Next For Silicon Chips? If there's a metaquestion dogging computer designers, it's how much longer the industry will keep churning out silicon-based machines that are twice as fast as last year's. The most common guess: About a dozen years.
Chips' clock speeds already are increasing more slowly: 10% to 15% a year, versus 35% to 40% per year historically. Yet computer performance keeps nearly doubling each year, as Advanced Micro Devices, IBM, Intel, and Sun Microsystems make their products more specialized and combine more computing functions on a single piece of silicon. Even so, designers are running into engineering problems that rob performance as electronics shrink into the nano scale. "Frequencies will continue to go up, but nowhere near at the rate they have in the past," IBM's Horn says. "We're going to see a sea change in the way processors are designed."
But what's the limit? The question has many practical implications. "We've been on a curve where you lead with your fastest microprocessor with the biggest cache," and you charge premium prices for those products, Sun's Papadopoulos says. "Now the world's starting to unravel."
The semiconductor industry is building products with electronics just 65 billionths of a meter wide. The next two generations--45 nanometers and 32 nanometers, each about three years apart--look OK, Horn says. "It's pretty clear nothing's going to replace silicon in that time," he says. "You go one more cycle out and, I'll tell you, it's getting pretty dicey. The problem is, there's no good alternative."
Carbon nanotubes, plastic semiconductors, and more esoteric areas of research, such as using the electronic spin of electrons or the quantum mechanical properties of atoms to perform computations, have all been posited as solutions. "All have some potential," Horn says, "but there's no clear-cut road map as a replacement for silicon."
We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.