Author: Dann Corbit
Date: 01:43:30 07/17/00
Go up one level in this thread
On July 17, 2000 at 02:55:02, blass uri wrote: >I guess that 100 times faster hardware could save Deep Junior in these games. > >In the game against kramnik Junior blundered by Kh8 and so the problems few >plies after it. > >I read that in the game against piket Junior could win by 8.h4 h6 9.e6 > >If this is correct I guess that faster hardware could help Junior to find these >moves. > >I guess that faster hardware could not help it to understand the position later >but it does not need to understand the position if it does not get it. Your hypothesis is easily testable. Even a 500 MHz machine will do for the test bench. Choose any failed move and set up the board position. Let's say that the DJ hardware was about 6 GHz throughput. That means you need twelve times as long to equal the same strength with 500 MHz. So, if you let the program run over the weekend, you could get around 50 Hrs. Did it make the right choice? If so, you can calculate the point where hardware improvement would be enough. Just look at the log and find the time point. Divide by the scaling factor and you will know the hardware that is needed. Of course, if it does not find the right choice, you will know you need a more powerful test bench, but you will also know how much power would _not_ be enough.
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.