Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: By Christmas simply choose between the P4/3066 and the Athlon XP 3000+

Author: Jesper Antonsson

Date: 06:53:37 10/27/02

Go up one level in this thread


On October 26, 2002 at 17:10:27, Jorge Pichard wrote:

>In your case, since you have a lot of patience I recommend to wait until the
>Christmas of 2010, probabbly by then an Intel microchip might be as fast as the
>Deep Blue that play against Kasparov in 1997.

That may be a bit too optimistic (but not much). If the NPS doubles every 18
months, and today's single chip maximum is 1 millon NPS (perhaps it is closer to
two?), then we have the following progression, in MNPS:

2002	1,0
2003	1,6
2004	2,5
2005	4,0
2006	6,3
2007	10,1
2008	16,0
2009	25,4
2010	40,3
2011	64,0
2012	101,6
2013	161,3
2014	256,0
2015	406,4
2016	645,1
2017	1024,0

I've read that Intel expect to sustain Moore's law at least until 2017. I'm a
bit sceptical, but if true, we may reach one billion nodes per second on a
single procesor machine somewhere around that point. What's a typical branching
factor with null move? Three? Then we may be able to reach 6 ply deeper. That
depth will make today's programs look like they're playing much more
"positional" chess, even without enhancements.

Perhaps even more frightening, in 2017, the machines will reach 1-2 ply deeper
in bullet games than they reach with standard tournament time controls today.



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.