BOSTON (06/14/2000) - People want faster computers, and they want them now. The need for speed has pushed the computer industry to develop more powerful processors, denser memory storage and faster network connections. But silicon chips have certain theoretical physical limits, and manufacturers are rapidly approaching them.
Processors used in everyday equipment like cell phones and PDAs (personal digital assistants) can be improved by making them more specialized, rather than more powerful. But for the chips used to build the fastest supercomputers, engineers are finding they must experiment with different technologies entirely, industry leaders said Wednesday at the "Beyond Silicon 2000" conference here.
"We've only just begun a new, 30-year improvement in computer technology," said Nicholas Donofrio, keynote speaker and senior vice president for technology and manufacturing at IBM Corp. "Keeping pace with Moore's Law is not a significant challenge."
Moore's Law -- more of an empirical observation than a law -- states that the number of transistors on a single integrated-circuit chip increases by a factor of four every three years. Chip makers Intel Corp., Advanced Micro Devices Inc.
(AMD) and others have enthusiastically tried to prove the law true. Intel plans to release its new Willamette chip later this year, with a clock speed of 1.4GHz. Three years ago, the top clock speed was provided by a 233MHz chip.
The increasing cost of making high-speed chips -- a corollary to Moore's Law -- has kept pace with improvements in speed as manufacturers build larger, cleaner production facilities, said Joel Birnbaum, chief scientist for Hewlett-Packard Co. and a panelist at the conference. "Silicon chip development will reach a wall ... silicon CMOS (complementary metal oxide semiconductor) will end."
Conventional answers to moving beyond that development wall proffered by the panel included changes in computer architecture, allowing computers to use the power of unused or slower processors linked in a network. As for computer chips in specialized devices like cell phones, microprocessors can be designed to better serve the specific function, rather than improving overall speed, panelists said.
However, the computer industry is also betting hundreds of millions of dollars on unconventional alternatives worthy of science fiction novels. Molecular computers, built atom-by-atom using scanning-tunneling microscope technology, could theoretically store and process hundreds of thousands of times more information than computer chips made from silicon, Donofrio said.
Quantum computers use the state of electrons as the basis for calculation, and could operate even faster. These kinds of computers "have the potential to solve incredibly difficult problems, but won't be used for general computing," he added.
Computers like this may take decades to develop, but hybrid systems combining molecular and conventional silicon designs may evolve sooner, according to HP's Birnbaum. Other more mundane problems may impede development, however.
"If there's a stone wall, it may well be lack of people to work on the problem," said Andrew Viterbi, founder of wireless communications company Qualcomm Inc. The science involved draws experts in fields outside of traditional computer development, like physical chemists and molecular biologists. "We're clearly up against it in recruiting people," he added.
IBM, in Armonk, New York, can be reached at +1-914-499-1900 or at http://www.ibm.com/.