miércoles, 19 de mayo de 2010

Moore says nanoelectronics face tough challenges

SAN FRANCISCO--Although many believe the future of the computing industry lies with building chips out of carbon nanotubes or other novel materials, Intel co-founder Gordon Moore predicts it won't be easy to replace silicon. "I will admit to being a skeptic to these things for replacing digital silicon," he told a gathering of reporters here Wednesday, where he also discussed artificial intelligence, Intel's future, and the early days of Silicon Valley. "We've got a cumulative couple of hundred billion dollars invested in R&D."


Although he retired several years ago, Moore will be a very visible figure during the next few months. April 19 will mark the 40th anniversary of an article he wrote for Electronics Magazine that first sketched out the idea of Moore's Law. The observation, which predicts that engineers can double the number of transistors on a chip every 24 months, has been the fundamental principle of the computing industry and paved the way for making computers and cell phones that are cheaper, faster and more powerful.

"It was a chance to look at what happened up to that time," he said of the original article. "I didn't think it would be especially accurate."

While he says he isn't up on the latest technological nuances, his skepticism about novel materials replacing silicon derives from practicality. Modern-day microprocessors contain hundreds of millions of transistors, and soon will have billions, and, to break even, manufacturers have to pop out millions of these complex devices. Although researchers have been able to produce individual nanotube transistors, the ability to mass produce hasn't been shown.

Still, continuing to produce chips on silicon has its problems too. Designers have been able to put more transistors on chips for decades by shrinking the size of the transistors, but they are now at the point where some structures inside chips are only a few atoms thick.

"Any material made of atoms has a fundamental limit," Moore said. The solution? Make the chips bigger. Carbon nanotubes, he added, wouldn't be completely left out. They could be used to replace the metal interconnects between the transistors.

Rereading the article 40 years later yielded some surprises, he admitted. For one thing, he noticed that he predicted home computers.

"I also talked about electronic watches. Unfortunately, Intel tried that once," he laughed, referring to the company's failed foray into wristwatches years ago.

Moore also made it clear that computer scientist Carver Mead dubbed the observation Moore's Law, a lofty label that took him about 20 years to get used to.

Among other topics Moore discussed:

• He gave his approval to Intel's approach to building platforms, rather than individual chips. "The recent reorganization of Intel is to an extent a reflection of how (incoming CEO) Paul Otellini wants to work in the future. I think it is a very appropriate change," he said. "Paul is different in that he is the first CEO of Intel that isn't a Ph.D. or scientist, but he is more technical than I am at this stage in the game."

• William Shockley, who invented the transistor, helped foster the Silicon Valley by driving Moore, Intel co-founder Robert Noyce, Eugene Kleiner and the rest of the "traitorous eight" up the wall at Shockley Semiconductor.

"He was a brilliant physicist, but he had very peculiar ideas about working with people," he said. "We got along reasonably well because I was a chemist, so he didn't feel that he had to know everything I did."

The eight engineers went to the company's financial backers to take Shockley out of active management. At the last minute, the backers refused. Kleiner's father knew an investment banker named Arthur Rock, who then helped form Fairchild Semiconductor.

"Fairchild was developing technology faster than it could be exploited," Moore recalled. They also were mired in a management mess, so Moore and Noyce left to found Intel while others went on to start other companies.

• Computers, as they are built now, will never think like humans. "Human intelligence in my view is something that is done in a dramatically different fashion than Von Neumann computers," he said. The brain processes "in a highly parallel and relatively sloppy" fashion, but one well-suited for its purpose.

• China is going to be a major fact of life for the United States. "The impact of China is just beginning to be felt. China is producing 10 times as many engineers as we are," he said. "Silicon Valley is still a great place to start a company, but it so expensive, especially the housing."

• Progress in the industry may also slow to the point where the number of transistors on a chip, which let designers increase performance and/or integrate new capabilities, double only every three to four years. Still, the industry has always blown past barriers in the past.

He also noted that some of the analogies from Moore's Law are a bit farfetched. Once, he extrapolated that if the car industry followed the same rules of progress, cars would get 100,000 miles per gallon, travel at millions of miles per hour and be so cheap that it would cost less to buy a Rolls-Royce than park it downtown for a day.

And as a friend pointed out, Moore said, "it would only be a half-inch long and a quarter-inch high."

No hay comentarios:

Publicar un comentario