By Michael Kanellos
Monday, April 04 2005 09:19 AM
URL: http://www.zdnetasia.com/news/hardware/0,39042972,39224422,00.htm

Forty years ago, Electronics Magazine asked Intel co-founder Gordon Moore to write an article summarizing the state of the electronics industry.

The article outlined what became known as Moore's Law, the observation that the number of transistors--tiny on/off switches that churn out electrical signals that get represented as 1s and 0s--on a chip can be doubled in a short period of time. Adopted as a yardstick by the tech industry, the concept is one of the reasons the industry evolved into a high-growth, but high-risk, affair.

This FAQ explains the impact and consequences of the principles set down in the April 19, 1965, article.

What is Moore's Law?
When writing the article, Moore noted that the number of devices (which then included transistors and resistors) inside chips was doubling every year, largely because engineers could shrink the size of transistors. That meant that the performance and capabilities of semiconductors was growing exponentially and would continue to. In 1975, Moore amended the law to state that the number of transistors doubled about every 24 months.

When the paper first came out, chips sported about 60 distinct devices. By contrast, Intel's latest Itanium chip comes with 1.7 billion silicon transistors.

As monumental as the article has become, it wasn't a big deal then. It started on page 114 of the magazine.

"It wasn't something you expected to join the archives," Moore said in a recent gathering with reporters. "I didn't think it would be especially accurate."

Why is it possible?
It's the miracle of industrial chemistry. Silicon is a good semiconductor (which means it can conduct electricity, but in a manner that can be controlled), and the crystalline structure remains intact despite shrinkage.

Is the law now dead?
No, though various analysts and executives have incorrectly predicted its demise. It will, however, likely begin to slow down to a three-year cycle in the next decade and require companies to adopt alternative technologies.

Some people, such as Stan Williams and Phil Kuekes of HP Labs, say the ability to shrink transistors will start to become problematic by around 2010. That should prompt manufacturers to adopt alternatives, such as HP's crossbar switches, to control electrical signals.

Others, such as Intel's director of technology strategy, Paolo Gargini, paint a more gradual picture. Around 2015, they say, manufacturers will start to move toward hybrid chips, which combine elements of traditional transistors with newfangled technology such as nanowires. A full conversion to new types of chips may not occur until the 2020s.

From a theoretical point of view, silicon transistors could continue to be shrunk until about the 4-nanometer manufacturing generation, which could appear about 2023. At that point, the source and the drain, which are separated by the transistor gate and gate oxide, will be so close that electrons will drift over on their own. When that happens, transistors will lose their reliability, because it will be impossible to control the flow of electrons and hence the creation of 1s and 0s.

(The nanometer measurement refers to the average feature size on a chip. A nanometer is a billionth of a meter. Current chips are made on a 90-nanometer process, while experimental devices about 6 nanometers long have been produced.)

What happens then?
Hard to say. If alternatives to silicon transistors never materialize, Moore's Law stops. If alternatives emerge, progress could accelerate under similar principles.

What's the best alternative?
Who knows? Carbon nanotube transistors, silicon nanowire transistors, molecular crossbars, phase change materials and spintronics are mostly now lab experiments.

Silicon, though, won't go easy. Manufacturers and designers love it. Chances are, silicon will continue to be incorporated into these new devices in some fashion.

"I view (silicon) technology as a fundamental way for bringing out complex microstructures and materials," Moore said.

Who said what?
California Institute of Technology Professor Carver Mead was the one who dubbed it Moore's Law, a lofty title Moore said he was too embarrassed to utter himself for about 20 years. David House, a former Intel executive, extrapolated that the doubling of transistors doubles performance every 18 months. Actually, performance doubles more like every 20 months. Moore emphatically says he never said 18 months for anything.

The rule also doesn't apply to hard-drive densities or to the growth of other devices. "Moore's Law has come to be applied to anything that changes exponentially, and I am happy to take credit for it," Moore joked.

What does doubling do?
The impact can be summed up as follows: faster, smaller, cheaper. Under Moore's Law, chip designers essentially shrink the size of transistors--which are now measured in nanometers--and then fill up the resulting empty space on the chip with more transistors. More transistors let designers add features, such as 3D graphics, that used to exist on separate chips--thereby cutting costs.

The designers can also choose to dedicate more transistors to speeding up how the chip performs its usual functions. Despite the extra transistors, these enhanced chips cost about the same as the old ones, because they take up the same surface area of silicon.

As an added bonus, smaller transistors mean electrons don't have to travel as far, boosting performance. Though chip designs vary widely, manufacturers try to get some or all of these advantages.

How does that affect products?
Put into practice, Moore's Law spells out a way for companies to enhance their products at a rapid clip. Eighteen years ago, Michael Douglas, in the movie "Wall Street," spoke on a cell phone that was about the size and shape of a brick. Shrinkage and integration has lead to phones with television tuners, 7-megapixel cameras and MP3 players. Declining costs have also put them in the hands of billions of people.

More-powerful, cheaper chips have in turn allowed software makers to develop applications such as instant messaging, 3D games and Web browsers that would have been cumbersome only a few years before they were invented. Consumers and analysts regularly complain that progress outstrips their needs, but rarely does anyone revert.

What are the technical problems?
Getting electricity to transistors is difficult, and dissipating the heat generated by these transistors is just as challenging. Some transistor structures, such as the gate oxide, are only a few atoms thick, so they leak electricity.

Where does it go next?
The trend now is to put silicon where it isn't. In the coming years, various start-ups hope to embed sensors in walls, household appliances and even wild animals. Microfluidics chips will let doctors quickly harvest large amounts of patient data with less lab equipment.

What's the economic impact?
Very few industries are this lucky. Car manufacturers have to entice customers with new cup holders or different body types, because engine performance doesn't change that rapidly. Moore noted that if car manufacturers had something like this, cars would get 100,000 miles to the gallon and it would be cheaper to buy a Rolls Royce than park it. (Cars would also be only a half an inch long.)

The fear now is that the treadmill will slow.

"Replacement has always been predicated on our industry's ability to come up with neat new things to buy. That in turn has been predicated on greater integration, allowing richer features due to the progress of Moore's Law," wrote Dan Hutcheson of VLSI Research in his newsletter "The Chip Insider." "Slow it down, and end users are likely to slow their replacement rate. Slow this, and the market slows with it."

Are the economics healthy?
Yes and no. One of the unsavory consequences of tracking Moore's Law is called Rock's Law, named after venture capitalist Arthur Rock. It states that the cost of fabrication facilities doubles every four years. Now fabs cost billions of dollars and the cafeterias in them generally cost more than the old fabs, noted Craig Barrett, Intel's CEO. Most chip companies now do not own their own factories because of the costs.

Wall Street analysts, conference futurists and even some chip executives have regularly declared that the outrageous expenses will lead to the end or a slowing of Moore's Law. Most companies, though, never take the advice: Falling off the pace would just ensure extermination by faster-moving competitors anyway. The chip industry remains a multibillion-dollar industry.

What else did Moore predict?
Re-reading the paper after 40 years, Moore noticed that he also predicted the home computer and electronic wristwatches.

In the early '70s, in another article for Electronics Magazine, Moore also forecast the growth of Ovonics Unified Memory, a type of memory made from a similar material as CD disks. In February, Intel said it may come out with Ovonics memory in a few years.

Not everything he's said, however, has come true. He once predicted that wafers, the round disks out of which chips are harvested, would measure 56 inches in diameter about now. They measure 300 millimeters, or 12 inches.