Back to the Duncans Home Page
From the mailbox of
Dr. Gaylen Duncan,
President & CEO, ITAC,
Moore and Future Chips
Like clockwork for the past 40-odd years, transistor density on computer chips has been doubling roughly every eighteen months. That's Moore's Law.
In 1965, Gordon Moore of Intel observed a striking trend when he graphed data about the growth in memory chip performance [which is directly proportional to transistor density on said chip], namely: each new chip contained roughly twice as much computing capacity as its predecessor, and each chip was released within 18-24 months of the previous chip.
From this, Moore concluded that computing power would rise exponentially over a relatively brief period of time while costs would either remain static or decline.
Obviously, planners and forecasters like to have a handle on something with the apparent dependability of Moore's law. But Moore's Law may be running out of steam. Over the past number of years, those who thought that Moore's Law could indefinitely drive computing developments have learned that the upper physical limits of computer chip construction technologies [essentially, limits on X-ray lithography] will sooner rather than later spell the demise of the trend. As the upper limits of chip manufacture are approached over the next decade or so, costs of development will continue to rise with only periodic increases in performance. Faster chips are possible and will no doubt be built and sold, but costs and time to release will continue to increase, so development and chip speed will eventually top out. (For further information on this point see: Roy Brander, "The End of Moore's Law: Thank God!",
So as the bearings begin to heat up and scream under our roller-coaster seats, the last examples of 'the fastest chip ever' will in seriatim tumble out into the market.
But wait. Is Brander correct or not?
What will the upper limits of atomic level chip construction really be?
And how will the last of the fastest chips be employed?
Read the following from today's (07 Feb 02000) NewsScan
A NEW BEGINNING: THE AGE OF GIGAHERTZ CHIPS
In the second half of this year both IBM and Intel plan to begin producing complete microprocessors with processing cycles of a billion times a second (1 gigahertz). At first, they will probably be used in Web servers, and then in voice recognition and video systems.
Randall D. Isaac of IBM says,
"The message here is a simple one. The gigahertz era has arrived, and it looks like we have room to move up to the 3 or 4 gigahertz range very rapidly.
Everyone talked about these limits, the end of Moore's Law, everything was going to slow down. But everything seems to be speeding up. The pace is just breathtaking."
And Albert Yu of Intel explains that the gigahertz chip is just the very beginning of a completely new era:
"Silicon, from a technical point of view, basically has no limitations down to the atomic level, and we're still far away from the atomic level." (New York Times 7 Feb 2000)
Want even more to chew on?
The following is also from the NY Times ... Atomic-Scale Computing Closer. IBM shows how circuitry millions of times smaller than today's may be achieved.
IBM physicists announced a breakthrough Feb 03 that demonstrates atomic-scale circuitry -- millions of times smaller than today's computer microprocessors -- is feasible, and may eventually render modern electronic circuitry obsolete. If futurists prove correct, this nanotechnology will ultimately pack the power of a supercomputer into a device so small that they could be woven into garments powered by body heat, or injected into a person's bloodstream as super-intelligent diagnostic probes. It may also create computers that would never need to be plugged in and carry batteries that would never run out of power.
Of course you can turn to the December 1999 edition of Scientific American to peruse their futures-oriented articles on computational intelligence and the portions of our lives in which we might find such power embedded.
So, it's been one heckuva ride with Moore as the pilot, but it's not over quite yet. The end of Moore's Law is almost certainly in sight, but the capacities of computational devices that will arrive in our midst over the next decade or two will be spectacularly great, bordering on the incomprehensibly fast, ubiquitous, interlinked and powerful, and they will affect our lives and our futures in ways which chip designers back in 1965 might have only imagined in their wildest dreams.
"Gordon Moore just plain got it right ... I should also mention that Moore's Law has also given rise to Machrone's Law, which was true for many years, namely, that the machine you want always costs $5,000." -Bill Machrone
Back to the Duncans Home Page
order to be sure to give you personal attention,
This page last modified November 05, 2007
Copyright © 1992-2001, Duncans, All Rights Reserved.
"In the knowledge economy (1995-2020), the best strategy is to over-invest in connecting power. Competitive advantage accrues to those who invest more than their competitors to connect to more people and share knowledge faster and farther."
--Jim Botkin, Smart Business: How Knowledge Communities Can Revolutionize Your Company