SAN JOSE, Calif. – Moore’s Law won’t last forever. At some point, the physical laws that govern the behavior of transistors will make it impossible to keep doubling the number of transistors on a single chip every couple of years.
The industry is trying to postpone the day of reckoning for at least 15 more years. But many experts believe silicon will eventually have to be replaced as the key material of computer chips to keep processing power growing at rapid rates.
“Any exponential extrapolated far enough predicts disaster,” said Intel co-founder Gordon Moore, who first made the prediction known as Moore’s Law. “The fact that materials are made of atoms will be a fundamental limit. We can’t go smaller than that.”
If the industry fails to find a solution and Moore’s Law falls off track, the consequences could be devastating for the $1 trillion electronics industry, which relies on chips getting cheaper and more powerful to drive improvements in the capabilities of everything from televisions to cell phones.
The roots of the problem lie in the nature of silicon-based transistors, the tiny switches that turn the flow of electrons on and off in a circuit, creating the ones and zeroes that are the basis of electronic information.
Now that there are more than a billion transistors printed on a half-inch-square silicon chip, it’s getting harder to lay them down in the microscopic patterns that drive the chip’s functions. The lines separating the circuits are already smaller than the point of the sharpest tools available to draw and shape them.
Chip makers are also having trouble creating the insulation layers that separate the wires within transistors, since those layers are now only just a few atoms thick. And the packed-in transistors generate a lot of heat, potentially melting the entire chip.
The end of Moore’s Law has been forecast before, and each time the collective ingenuity of the $213 billion-a-year chip industry has come to the rescue with a technological solution to extend the capabilities of silicon chips.
But Moore and others admit they are getting more worried because there are more physical barriers standing in the way as manufacturing gets close to manipulating individual molecules. At that level, a few atoms out of place could ruin an entire chip.
Bernie Meyerson, an IBM vice president, said alternatives to silicon will have to be found by 2020. Nanotechnology, a radically different approach to manufacturing on an atomic scale, offers the best potential for both the tools and materials that could enhance or replace silicon.
Traditionally, chip makers take a bulk material and winnow it down into a structure that they want. With nanotechnology, scientists envision starting with individual molecules and getting them to assemble themselves into larger electronic structures that can do the equivalent of what a chip does, which is store data or turn a switch on or off. It’s the opposite of taking a bulk material like silicon and winnowing it down to the right structure.
Governments are pouring $10 billion a year into nanotechnology research, but it may not be enough. “If we want to be in business by 2020, we have to start on it now,” said Paolo Gargini, a technical fellow at Intel.
In the short term, the industry will turn to a new type of insulator, known as a high-k dielectric insulator, to improve the power efficiency of silicon. Chip makers are also working on a way to write smaller patterns on chips. One research consortium has spent several hundred million dollars on extreme ultraviolet lithography, which will use mirrors and other methods to bend light in a way that makes writing patterns possible at tiny dimensions.
Looking out to 2010 or 2015, offshoots of nanotechnology could become useful to extend silicon’s usefulness. For instance, since 1991, scientists have been exploring carbon nanotubes, which are long sheets of graphene rolled into tube-like wires. These nanotubes can conduct electricity and are so thin they could replace today’s chip wiring, but so far, they are difficult to control.
In the long term, around 2020, true alternatives to silicon could materialize from current research into optical computing, which harnesses light to do calculations. Other promising experimental technologies include quantum dots, which are tiny switches that can be manipulated with a single electron, and spintronics, which exploits magnetic technology to do computing.
Moore himself is skeptical that the traditional methods for manufacturing chips will be replaced anytime soon.
“You can clearly make a tiny little transistor with potentially a very high frequency,” he said at a recent news conference. “But can you connect a billion of them together? That’s really the problem.”
If the cost is too high, economics may kill off Moore’s Law before technology does, said Bill Joy, a co-founder of Sun Microsystems and now a venture capitalist. Chip factories cost as much as $3 billion, as much as a nuclear-powered aircraft carrier.
Jeff Hawkins, who wrote the book “On Intelligence,” argues that the answer is to invent a whole new kind of computing that works more like the human brain. He has started a new company, Numenta in Menlo Park, Calif., to pursue that dream.
“The brain is much better at tasks, such as vision, than computers are today,” he says. “That’s because of the way it is organized, not because of its processing power.”
Carver Mead, the former professor at the California Institute of Technology who coined the term “Moore’s Law,” said a slowdown might not be a bad thing.
“We have not begun to use the capability of what we already have,” he said. “I think it’s healthy for the industry. It means we have to start thinking again.
—
(c) 2005, San Jose Mercury News (San Jose, Calif.).
Visit MercuryNews.com, the World Wide Web site of the Mercury News, at http://www.mercurynews.com.
Distributed by Knight Ridder/Tribune Information Services.
AP-NY-04-13-05 0620EDT
Comments are no longer available on this story