Chip makers brace for slower pace in Moore's Law

[lovuian]

http://news.yahoo.com/s/nm/20050916/tc_nm/summit_asia_moore_dc

TOKYO (Reuters) - The journey to ever smaller, faster and cheaper chips is slowing down and may put a big dent in sales and profits of the semiconductor sector and even the economy, industry players and analysts said this week.

ADVERTISEMENT

Until recently, chip makers doubled the capacity of their products at the same size and cost every 18 to 24 months, helped by miniaturization and scale advantages, a phenomenon known as Moore's Law, named after the Intel co-founder Gordon Moore who predicted this trend as early as in 1965.

As Moore predicted, a transistor that cost about $1 in 1968 dropped to just under 10 cents in about five years. The cost again fell 10-fold every four to five years until 1985 when the cycle lengthened to seven years, according to Intel and Dataquest data.

Technically, his trendline continues, with lithography machines steadily shrinking the detail on a chip from 130 nanometers in 2001 to 90 nm in 2003 to 65 nm 2005 and set to move to commercial chips with 45-nm detail in 2007.

more...

How small can we get???

[Maple]

1. What's the smallest thing in existence?

That we know of so far?

That small.

[lovuian]

2. Good Point Maple!!!

itty bitty!!! takes on new meaning!!!

[Bernardo de La Paz]

4. Bush's Scruples? Approaching the Planck Length

The Planck Length is named after the physicist Max Planck and is the smallest measurable length for anything because below that length, quantum effects take over and everything becomes stochastic. It is about 5 x 10^-44 meters.

Electrons are pretty small.

[Sialia]

25. The Planck Length

The Planck length is way, way below the scale at which quantum effects start to become significant. Quantum effects are important at about the size of a molecule (roughly 10^-9 m). The Planck length is the scale of full force unification (i.e. quantum gravity). That is, at the Planck scale none of the four fundamental forces we see in the current universe (gravity, electromagnetism, weak nuclear, strong nuclear) are distinguishable.

Semiconductors are already inherently quantum devices, though the overall behavior of the chips can be described classically. What is happening is that at too small a scale, thermal fluctuations start to overwhelm the signal. Plus power consumption goes up as the dies shrink. Intel has already stated that they're going to start concentrating on reducing power consumption rather than increasing clock speed.

[sakabatou]

13. Well someone did make a watch with 1 nm gears

[Delphinus]

15. The boy king's heart ... black as it is? n/t

[Tesha]

24. Bush's brain? (NT)

[chicagiana]

3. You can only go so small ...

At some point, the circuits will become subject to quantumn effects.

Fear not, at that point our CPUs will probably sport a dozen or more cores and a gigs of of internal cache. The name of the game will be parallell technology and multi-threading.

[Maple]

5. Actually...it will be something quite different

so you won't need to worry about 'quantum effects' and all the rest of it.

[petgoat]

11. Photon computers have the potential to continue the Moore paradigm;

multiple-core processors can add functionality (if they can solve the heat problem!).

We don't know how to use the computer power we've got now. Hardware's
not the limitation; software is behind hardware; user education (and I'm
talking about myself here) is so far behind both that it's got to be the
limiting factor.

I look forward to the time when a basic wireless internet-ready machine
can be built for $200 and people across the world can learn to read from
internet tutors.

[chicagiana]

19. It depends on the application ..

We know how to make dedicated software that performs a well defined task set. For example, the people using super-computers could utilize all the CPU and I/O power you can give them.

Software gets messy when you start making it more general and more inter-operable.

When it comes to AI, that's largely a software problem. AI looks at intelligence from completely the WRONG angle. And it is very possible that what is required is a completely different HARDWARE paradigm that emulates the processing and memory operation of the human brain. This means making machines that by definition are VERY prone to error. But this is what intelligence IS.

[Maple]

21. Operating on light

or liquid.

[Occulus]

12. Or we may be using a quantum computer of some sort eventually:

http://www.cs.caltech.edu/~westside/quantum-intro.html

[Bernardo de La Paz]

6. We don't yet exploit all avail. capacity due to software primitiveness.

[Bernardo de La Paz]

7. One limiting factor is the capital cost of fab plants

One limiting factor is the capital cost of building and bringing online a semiconductor fabrication plant. The most advanced plants cost about two billion dollars now.

[Maple]

8. The last thing you need

is semiconductor fabrication plants

[Kenergy]

9. Our disposable income under the Bush admin n/t

[Solo_in_MD]

10. This has been predicted for years

And then pr oven wrong. A new process or even new technology emerges that keeps the increase in performance on tract. Read the tech rags 5 years ago, they predicted the same thing back then.

The limitation on packaging (size) is more a factor of heat dissipation and power. IF you make the Ipod too small, you can't have the neato controls etc.

The current trend is toward more function at the same size and price point. For example, the $1000 desktop computer. That has been the historical price point for "good desktops". Yes there are cheaper ones, but most people spent right around 1K. What they get is more and better every six months or so.

[MGKrebs]

14. Nanotubes!

HOW WELL NANOTUBES CONDUCT ELECTRICITY depends a lot on their
environment. Hongjie Dai and his colleagues at Stanford have made
the first electrical measurements of currents flowing under high
voltage (high bias) through single-walled carbon nanotubes suspended
like miniature power lines. They discovered that in suspended form
a micron-scale-long nanotube could carry about 5 micro-amps of
current, whereas lying in the plane of a substrate the same tube can
carry about 25 micro-amps. The reason for the better in-the-plane
performance is that the substrate helps to dampen "optical phonons,"
high-energy vibrations of the nanotube atomic lattice. Dai
believes that with careful
engineering of the interface between a nanotube and a substrate,
maximum currents could be raised to higher levels than previously
possible, which might make carbon nanotubes useful for applications
in high-power transistors and even nanoscale transmission lines. To
make the kind of transmission lines you see in the countryside out
of nanotubes, you'd have to develop a process for producing
km-length carbon tubes, which is not feasible for the foreseeable
future. (Pop et al., Physical Review Letters, upcoming article)

[Gregorian]

16. We only need to be so small.

We're working with, and against the laws of nature. We're at that point in the curve where the easy things have been achieved. Now we're searching for the harder to find open doors. And remember, our pace is greatly increased from when this first began. It took centuries to get to the diode. Then decades to get to integrated chips. Then a decade to produce them efficiently. And now a year or so before we make the next jump.

And we only need to get so small. I use the human body as a reference. Yes, we want to send things into space. But we only need electronics to be so small to do what we want. Soon we'll have Dick Tracy's wristwatch. Kind of hard to believe, really. And to further my argument, computers only need to go so fast. I bought half a dozen computers from the mid 80's until 2000. Then one more, and I'm almost never thinking about the speed anymore. We can do all but the most sophisticated fluid and thermodynamics on our pc's.

We fight not against those things we see, but those invisible forces. It's like there is a war going on that we cannot see. We can't see gravity. But step off a roof and it hits you like a brick.

We have to work like mad to find the answers. It keeps us busy, that's for sure.

I can't help but think that nanotubes will be in the news again.

[daleo]

17. I got a new computer at work in March

And I still haven't gotten around to unpacking it. The minimal improvement in speed (for routine spreadsheet, database, and statistical applications) just isn't enough to warrant the loss of time in moving my data, re-loading various applications, tweaking things, etc.

I suppose I will get around to it eventually, but it is hard to get excited.

[Renew Deal]

18. Here's some info on Moore's law

Right from DU.

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=240x567

[Arkana]

20. Well, they did say it would slow down eventually.

[loudsue]

22. Nano nano nano nano nano nano

Nano technology is going to be the Mt. Everest of the computer age.

:loveya: I just love the concept! :loveya:

:kick::kick::kick:

[Sven77]

23. cheap to make

Intel chips cost $40 to make - report
http://www.theregister.co.uk/2005/09/14/intel_die_costs/