November 4, 2005 1:00 PM PST

Moore's Law lives

All things eventually come to an end.

Or do they?

For years technologists have debated the prospects for longevity of Moore's Law, an observation made by Intel co-founder Gordon Moore in 1965 that the number of transistors per square inch on an integrated circuit doubles every year to two years.

Click here to Play

Video: Life span of Moore's Law even amazes the author
--Intel co-founder Gordon Moore

Over the past 40 years this observation, first published in an article for Electronics magazine, has become a defining principle of computing that has driven the industry to build computers and cell phones that are cheaper, faster, more powerful and, very often, more compact.

At a symposium hosted on Friday by the Marconi Society in New York City, where Gordon Moore is being honored with a lifetime achievement award, scientists and engineers discussed just how long Moore's observation will hold.

Moore himself said he is amazed that his prediction has held true even this long.

"I never expected it to be especially accurate," he said during a panel discussion. "It's amazing that a throwaway article has established a history of its own."

The success of Moore's predictions has created some problems. After decades of doubling transistors, a single chip now contains several million transistors. Multibillion-dollar factories have been built to produce these increasingly complex chips. Shrinking the size of transistors and the copper wires that connect them to fit more densely on a chip has also led to problems like electric leakage, increased power consumption and processors that generate a fair amount of heat.

Not to mention the fact that physics at some point limits how much transistors can be shrunk.

"My intellect tells me that it will end at sometime," said Leonard Kleinrock, professor of computer science at UCLA and creator of the basic principle of packet switching. "The size of an atom, the size of my fingers, and the capability of my eyes will at some point get in the way."

Transistors consist of four basic parts: a source (which stores electrons), a drain (where they go to create a "1" signal), a gate and a gate oxide (which controls the flow from the source to the drain). After several shrinks, the gate oxide is only about 10 atoms thick, in some cases, meaning further shrinking is not possible without an arduous chemical or architectural makeover.

But many scientists in the industry today believe that new materials and new methods will be developed to extend Moore's Law. Gradually, chips could move away from silicon to other materials such as carbon nanotubes.

Another strong possibility is that silicon continues as the medium of choice, but new materials get added to it. Some technologists, such as Federico Faggin, who developed the method for manufacturing metal oxide semiconductors, said that chips may be built in three dimensions, with components layered on top of each other like skyscrapers.

"We've never used a third dimension," he said. "So we will have to learn to build up."

Other technologists say that Moore's Law will have to continue, because society demands it.

"Can the end of Moore's Law really happen?" said Robert Lucky, Marconi Society chairman and former director of Bell Laboratories. "We all live off this trillion-dollar industry, so what would happen to the industry if it ended? We all demand more. Whether that means going to carbons nanotubes or 3D, I don't know."


Join the conversation!
Add your comment
Tunneling Current at Room Temperature
We're going to hit the wall for room-
temperature silicon CMOS pretty soon.
I'd say that 35 nanometer is about as
small as you can get without unacceptable
lowering of the operating voltage.

That means that, sometime in 2011 or 2012,
we'll have to develop a new method of
making chips bigger. 3-D stacking is one
possibility. Another is to make the dice
bigger. Perhaps wafer-scale is going to
make a come-back. . .
Posted by (139 comments )
Reply Link Flag
Back in the good ole days
I remember back in 2000, the 1Ghz machines were out. But needed refridgeration, to keep them cool.

Now the very laptop, I'm using to type this, runs a 3.2Ghz with no problems at all.

I don't know where this will all end. But I suspect architectural changes will help lower voltages.

But if we go 3D, perhaps processor chips/blocks will have to eventually be built around a cooling source - built around a fan, or including liquid cooling.
Posted by (409 comments )
Link Flag
Moore's Law good but not limiting
Limitation means destination not brokeness.

It explains how atoms are the limit or last
horizon but once you control the atom with
nanotechnology and multi-dimensional techniques
(3D chips etc.) you have the speed you need to
retain at or about even the speed of light
That means instantaneous operations to all.
i think the question here is what do we want? We
want full organic simulation per perspective.
That means we need a processor per perspective or
person to achieve this I think.

Already computer graphics are easy on the eyes at
3 gigs speeds but they don't go deep like to the
cell or atom mainly for more technical reasons.
We need Petaflop speeds to emulate organics on a
computer and maybe even upload ourselves to a
virtually prefect environment and that is not far
off if not here already with IBM's Blue Gene at
around 1/3 a Petaflop. These speeds emulate and
fix organic deterioration issues and limitations.

If the new light speed technology or Nano-tech
comes it will probably allow for petaflop speeds
in your back pocket and that means a fully
organic digital world for each individual to
Posted by Blito (436 comments )
Reply Link Flag
Other Substrates and Strategies
One of the problems with Silicon is the on chip heat at the transistor junction can melt the silicon substrate. In a typical Intel clocked chip all transistors are receiving power all the time, whether working or not. The asynchronous design, where only thos transistor arrays needed for current work are powered, can help alleviate this problem.
A better way for all concerned may be a higher melting point substrate, and recent innovations in manufacturing cheap diamond slices which can be doped to carry a current, can carry existing designs forward, while work on other carbon based technology like nanotubes and gated quantum diodes (transistors) develops into something useful.

Although we all concentrate on transistors when speaking about Moore's Law, little is said of the other electronic components and their size that make up the chip's population of parts, so a parallel development for the size and performance of these must also be a priority for chip designers.

Apart from young people with 20-20 vision, size of hand held devices need not shrink too much in an aging population, so that new innovations in chip design could be focused towards functionality of the interface, like what is happening with the Cell chip, and performance for vertical function like what the Rock chip is promising.

As is seen in the development of graphics chips, it is the length of the registers that let instruction set engineers let us have complex functions in a minimum of chip cycles, so that CPU with say 256 bit instructions and data paths will also enable faster processing and complex funtionality.

Another architectural change which can enable these features is a return to the asynchronous 32 bit multi chip systems of the 70s to 80s workstation era, where each I/O and vector function was performed autonomously by a different chipset, the CPU only getting involved in traffic jams.

So apart from helping Intel sell "moore" of their very out of date 80's technology 32 bit chips, Moore's Law is not the only way to increase speed and, "moore" importantly, funtionality of computing devices.
Posted by Stomfi (52 comments )
Reply Link Flag

Join the conversation

Add your comment

The posting of advertisements, profanity, or personal attacks is prohibited. Click here to review our Terms of Use.

What's Hot



RSS Feeds

Add headlines from CNET News to your homepage or feedreader.