Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Node frequencies, and a flame

Author: martin fierz

Date: 04:24:49 10/17/03

Go up one level in this thread


On October 16, 2003 at 23:48:18, Dann Corbit wrote:

>On October 16, 2003 at 22:48:09, Robert Hyatt wrote:
>>On October 16, 2003 at 19:11:20, Dann Corbit wrote:
>>>On October 16, 2003 at 18:49:55, Anthony Cozzie wrote:
>[snip]
>>>>1. Moore's law is NOT A LAW.  Its going to come to an end by 2020, if not
>>>>earlier.
>>>
>>>Not a chance.  It will continue to accelerate.  Of course, I could be wrong.
>>
>>It can't possibly continue to accelerate.  Everything is limited by C.  Nothing
>>can propagate faster than that.  So we are stuck with shrinking to shorten
>>distances so that C doesn't kill us.  But then we are limited by how far we
>>can shrink things.  IE we now do traces that are a few dozen atoms wide.  We
>>won't get to 1-atom widths.  And we _certainly_ won't get below that.
>
>Too many assumptions.
>
>Imagine (for instance) if we grow IC's that are 3-dimentional.  Suppose (for
>instance, that instead of making 10 nanometer traces on a 1x1 cm flat face, we
>are making 10 nm thick slices linked together in a 1x1x1 cm cube.  Now the
>compute power is suddenly 1e8 times larger.

sounds good at first, but think about this: today's processors generate
something between 10 and 100 watt of heat that you need to remove. since your
idea explicitly attempts to use today's technology, that would mean that you
also generate 1e8 times more heat. 1GW, that's about what an atomic power plant
delivers... now that will need one hell of a cooler :-)
ok, so you say you will go to lower voltages in the future, as we have done in
the past. but there is a limit there too, which is given by the band gap of
silicon. you can't go lower than that, and we are already quite close IIRC.

>Now, that's just one sort of work-around.  I imagine that there are many people
>a lot more clever than I am that can think of even better solutions.  (Using DNA
>to compute is a popular idea that may have merit).
>
>When we run out of ways to make the chip faster, why not just add more chips?
>So instead of 1 50 GHz chip, why not use 1000 10 GHz chips?

because you are increasing the distances again. and many tasks are not easily
parallelizable (e.g. chess...)

>I would be very surprised if chips fail to follow Moore's law for the next 30
>years.

i would be very surprised if they do. the main drivers of moore's law over the
years have been miniaturization, miniaturization and miniaturization. and that
is very definitely going to end in the near rather than in the far future.
moore's law is an empirical observation. the laws of physics are a bit more
solid than that :-)

you can bet your money on quantum computers, DNA computers or other fancy stuff.
IMO that's the only hope for moore's law in around 10 years time or so. and i
certainly won't bet my money on that kind of sci-fi stuff!

cheers
  martin



This page took 0.01 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

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