computer chips should be able to keep boosting chip performance for at least 15
years despite some serious technical hurdles, the industry's trade group
The Semiconductor Industry Association noted that chip manufacturers may face
problems in as soon as five years in maintaining improvements at current rates.
But the group has identified enough possible solutions to those bottlenecks to
predict that the industry will stay on the path of Moore's Law, named after
Intel Chairman Emeritus Gordon Moore, who predicted in the 1960s that chips
would double in performance every 18 months.
technology strategist at Intel
Corp. in Santa Clara, Calif., and one of the organizers of the industry group's
technology "road map," said memory chips will consist of 64 billion transistors
by the year 2014, 1,000 times the 64 million today for standard memory chips
used in personal computers. He also said that microprocessors will reach speeds
of 3.6 gigahertz, or 3,600 megahertz, compared with today's typical speeds of
500 megahertz to 733 megahertz.
The road map is a
guidepost to the $144 billion chip industry, whose progress has been closely
linked to Moore's Law for three decades. While some experts have predicted a
slowing in the ability of manufacturers to keep shrinking chip circuitry, and
thus chip capacity or speed, progress has sometimes come faster than the SIA has
In 1997, for example, the group predicted that chip
makers by 2001 would use a manufacturing technology that could create circuits
that were 0.15 micron in width, or 1/800th the width of a human hair. Now it
appears that chip makers by then will be using even smaller technology, at 0.13
And with each challenge, the industry has planned ahead and come up with
alternatives, said Robert Doering, a senior fellow at Dallas-based Texas
Instruments Inc. This time, one of the biggest obstacles is in finding a
replacement for silicon dioxide, a key insulating material in most chips,
The problem is that today's manufacturing technologies are accurate enough to create a uniform layer of silicon dioxide that is only eight atoms in thickness. Once that gets down to one or two atoms of thickness, the uniformity may be spotty and the layers can't be stretched any thinner.
U.S., foreign cooperation
Gargini said he was optimistic that the silicon dioxide problem can be overcome within five years, partly because U.S. and foreign companies are jointly working on a range of alternatives. In addition, the chip industry is expected to pour $600 million in funding over 10 years into special research centers at U.S. universities to help overcome barriers.
As if on cue, researchers at the University of California, Berkeley, announced a chip breakthrough Monday. Chenming Hu, a professor of engineering, said his team created a kind of transistor that is 400 times smaller than typical transistors today. The so-called FinFET transistor has a different structure that allows it to retain an electrical current even though it is much smaller than other transistors.