Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Introducing "No-Moore's Law"

Author: Matt Taylor

Date: 23:29:47 02/28/03

Go up one level in this thread


On February 28, 2003 at 17:52:48, Jeremiah Penery wrote:

>On February 28, 2003 at 11:46:24, Matt Taylor wrote:
>
>>The newest Intel process is 90 nanometers. I remember them talking about ways
>
>90nm isn't in production yet.

No, but it soon will be.

>to
>>hit 30 nanometers. They are -fast- approaching a width of 1 atom.
>
>30nm isn't going to happen in the next few years.  And 30nm still isn't close to
>the width of a single atom.  IBM has demonstrated 6nm transistors.

Let's do a little math. I pulled my Chemistry textbook off the shelf, and it
says the width of an atom of silicon is 118 pm (0.118 nm). If we are using a 130
nm process now, that's 1102 atoms wide. The 90 nm process is 763 atoms wide.
Now, following Moore's Law, that's about 14 years we have left at most on
silicon.

There are other materials, yes, but are they smaller than 118 pm?

This is not even considering financial and other practical limitations that
others have pointed out. AMD and IBM are working together to research smaller
SoI processes. Undoubtedly they both stand to gain from making AMD more
competitive with Intel. (Not to imply that they aren't competitive right now.)
However, it is evidence that the financial burden grows increasingly large.

>>If you have read much about Quantum computing, it is useless for many
>>applications. Quantum computing is useful only for highly parallel problems.
>
>For many problems, parallel algorithms can be devised to replace the serial ones
>that run well on today's computers.

Yes, like chess and encryption cracking.

>>Quantum computers run at ridiculously low speeds right now -- a few Hz.
>
>Key words being "right now".  The forerunners of today's computers didn't run
>any faster, and technology is accelerating.

In serial applications a 2 GHz Pentium 4 is about 400,000,000 times faster than
a quantum computer. Quantum technology is going to have to break the
practicality barrier before they can even think about making it faster.

>>Aggressive estimates are viable quantum computers in about 20 years.
>
>And?

Moore's Law runs out before 20 years. Quantum computers aren't a viable
replacement because they won't be nearly mature enough in 14 years to replace
silicon. Chances are they won't be mature enough even in 30 years. Yes, they'll
be fast for parallel applications, but that's not going to help you when you
want to play Quake or type up a document and have it spell checked, etc. That
brings up another point -- there are many real-time problems that a 5 Hz
computer cannot solve because it simply can't do it fast enough.

Yes, quantum is great for solving really big problems, but it absolutely sucks
for doing many commonplace desktop activities. When a problem isn't parallel,
you might as well invest in a 6502 because it would solve the problem faster.

Many of the quantum researchers talk about pairing silicon with quantum
technology. The obvious application is encryption cracking. The quantum computer
can solve it quickly, but it still takes silicon to sift through the results.

>>DNA computing is likewise a parallel paradigm and does not address -many-
>>problems.
>
>See above.
>
>There are plenty of other options for high-performance computing that don't
>exist today.  I wouldn't be surprised to see asynchronous chips being seriously
>considered soon.  The clock-based approach is beginning to cause lots of
>problems as speed is agressively increased.

Yes, but that still does not solve the problem. It only delays it. Furthermore,
it does not deter Moore's Law in any way. Moore's Law will hit a brick wall, and
at that point the exponential increase in processing power per CPU will likewise
hit a brick wall. Some things will benefit from multiprocessing/parallel
techniques; other gains will be made with different design strategies; perhaps
manufacturing processes will still see some increases (new materials, etc.).
However, the exponential increase of processing power will end.

-Matt



This page took 0.01 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.