RealTime IT News

Moore's Law Relevant But Not Forever

Next week marks the 40th anniversary of Moore's Law and yet another round of debate in the semiconductor community about its relevance.

Intel and its friends will celebrate the first publication of Moore's article Cramming More Components Onto Integrated Circuits in a 1965 edition of Electronics magazine.

Moore, who was then head of R&D at electronics giant Fairchild Semiconductor , said integrated circuits would lead to such wonders as home computers or at least terminals connected to a central computer. Moore also envisioned the number of transistors per square inch on these integrated circuits doubling every year.

In subsequent years, the pace slowed down a bit, but data density has doubled approximately every 18 months, and this is the current definition of Moore's Law, which Moore himself has blessed.

Most experts, including Moore himself, expect Moore's Law to hold for at least another two decades.

"Countless 'experts' have been predicting the imminent demise of Moore's Law for at least 20 years," George Scalise, Semiconductor Industry Association president, said in a statement. "They have all been wrong, thanks to a relentless commitment to extending the frontiers of science by the industry, university researchers, and the federal government."

But others find Moore's Law closer to retirement than immortality. Obstacles like pure materials science, systems architecture, and manufacturing are becoming larger and more difficult to work around.

"We got spoiled over the past two decades because we could make transistors that were smaller, switched faster and used far less power," said Nathan Brookwood, an analyst with Insight64. "Now we can only get two out of three advantages. That doesn't mean the law doesn't apply, it just means chip designers have to think differently."

Currently, PC and server chips are made using 130-nanometer and 90-nanometer technology with plans for 65-nanometer manufactured chips expected by the end of this year. Compare that to a decade ago when chips were made using 500-nanometer processes.

Nanotechnology has been widely seen as a way to prolong Moore's Law, but increasingly powerful software originally created a need for more powerful PCs, which, in turn, led to software that was more powerful still. Many believe that cycle and the IT purchasing decisions based on it are no longer relevant.

"Moore's Law isn't a proclamation, legislation, or fiat. It's an observation about the intersection of physics and economics. I'd like it if the Speed of Light was more flexible, too, but it's about as likely to happen," Jonathan Eunice, an analyst at IT market research firm Illuminata, said.

Eunice also agrees that Moore's Law in general has become a frequently misused term in the industry. He said people use it to justify that the idea that levels of chip performance, functionality, chip frequency, system performance, etc. will continue to increase exponentially.

"It doesn't say those things at all. The broadly held misunderstanding is that many desirable attributes -- performance, functionality -- will increase at similar rates. That is much closer to Raymond Kurzweil's generalization, which is also known as The Law of Accelerating Returns."

Some of the biggest gains from Moore's Law have come from new ways of making smaller chips. Lithography tools, resist-processing tools and materials, and etch tools have all been vastly improved from the early days of using a camel's hair brush back in 1957 to paint on hot wax for the mesa transistors.

In his essay entitled 40 Years of Moore's Law, VLSI Research CEO Dan Hutcheson said Moore's Law also has a wider impact than just shrinking chips and speeding performance.

"The economic value of Moore's Law is that it has been a powerful deflationary force in the world's macro-economy," Hutcheson said. "Inflation is a measure of price changes without any qualitative change -- so if price per function is declining, it is deflationary. This effect has never been fully accounted for in government statistics."

Hutcheson warns however that the costs of the research to keep Moore's clock ticking are rising with each node.

"I fear the day that it becomes too expensive for the private sector, and the clock stops," Hutcheson said. "In part because of the many conveniences, but mostly because of the dramatic effect it has had in driving America's productivity and thus its leadership in the global economy. When Moore's clock stops, the consequences to the economy should be obvious."