Author: Slater Wold
Date: 13:39:54 03/20/02
Go up one level in this thread
On March 20, 2002 at 14:27:32, Torstein Hall wrote: >On March 20, 2002 at 12:41:50, Gian-Carlo Pascutto wrote: > >>On March 20, 2002 at 12:29:59, Jonas Cohonas wrote: >> >>>I would gladly pay 400$+ for a finished version of this hardware/engine or (how >>>much it will cost eventually) if what they say are true. >> >>Hmm, it depends on how good they can get it, of course :) >> >>2-3M nps is not very impressive with a small evaluation >>(state of the chip now) >> >>-- >>GCP > >Can they not do it somehow the "Deep Blue way" also using the PC as evaluator >and having the extra chip search fast and doing a small eval? > >Torstein The software in DB did the first 6 ply. Everything else was 100% HW. The software also of course "distibuted" all the work. DB had 480 chess chips. On a .13 micron chip, packed to the core, a PC could maybe have 20. However, Hsu estimated teh .13 micron chips to do 30M nps. 30M x 20 = 600M nps. You don't want to know the cost to get a fab to do a .13 micron chip.
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.