Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why is SMP not standard in chessprograms?

Author: Robert Hyatt

Date: 14:44:15 09/25/01

Go up one level in this thread


On September 25, 2001 at 14:16:29, Dann Corbit wrote:

>On September 25, 2001 at 14:06:14, Robert Hyatt wrote:
>[snip]
>>Hold on to your hat.  I'd expect that within the next _five_ years you will
>>see single chips with two processors on them.  _that_ will change the way
>>computing looks, once and for all...
>
>Why stop at two?  If you put 64 CPU's on a single die, you can eliminate the
>external switching circuitry, and greatly reduce the distance the signals must
>travel.  At super high switching rates, a distance of 10cm is very significant.
>(See Seymour Cray's "nanosecond of wire" for an illustration).
>
>If we have trouble breaking (say) 10GHz, then just have 1024 CPU's at 10GHz
>which would give a few terahertz [though not ten certainly].
>
>I'll bet that would play a pretty snappy game of chess.  Might even make blitz
>worth watching.


More will come.  As transistors shrink, more and more are going to be stuffed
onto a chip.  Ncube did a 4-cpu chip several years ago.  After the duals, the
quads and up will come over time.



This page took 0 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.