Author: Robert Hyatt
Date: 07:09:33 04/13/02
Go up one level in this thread
On April 13, 2002 at 09:20:59, Nick Wilson wrote: >Has any chess engine programmer looked at using the internet to create a >massively parallel chess engine? > >Deep Blue is 1000 times (or whatever) faster than Fritz, but if 10,000 or even >100,000 PCs could be harnessed, surely that processing power would be a killer? > It isn't so easy to do. Internet latency would cause massive problems when trying to approach DB's speed. >If every internet-connected chess player allowed spare processing power to be >used, or even if a large corporation let its entire 10,000PCs be used overnight, >then surely even the might of Deep Blue would be put in the shade... It would not be easy to use _all_ that power. First, there will be significant efficiency lost due to issues in parallelizing alpha/beta. And then there is the network latency that reaches tens of seconds or even minutes at certain times... > >Ok, the task of managing all that would be <ahem> tricky... maybe an engine >working that way could only work on tournament time limits. > >Is this realistic? Anyone tried it on a smaller scale? > >I know in the UK, there was a screen saver that used 'spare' processing power to >scan extra-terrestrial radio signals, and transmitted findings back via the >internet. But this is obviously a far simpler exercise than playing chess.. This was done by NASA...
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.