Author: Gregor Overney
Date: 12:53:31 05/03/99
Go up one level in this thread
This boils down to the question if Deep Blue's performance is mainly possible due to advances made in hardware design, software (including firmware), or both. (Probably both.) My first guess is that Deep Blue's power is mainly based on its multiprocessor architecture and extremely high bandwidth when compared with PC's. Once a plug-in board is able to achieve what Deep Blue was able to in 1997, the PC hardware has again gained a lot in strength. This would ultimately narrow the gap between such a future "plug-in version of Deep Blue" and Junior running on a future PC architecture. Dedicated hardware is always a great idea, if you have an unlimited budged (or you expect your customer to have it). - A low-cost version that is 1000 times faster in playing Chess than today’s PC's is not possible with our current, best thin-layer processes (0.1 micron, 3 GHz clock rate, 128-bit architecture) that could be produced in higher volume. As of today, the "packing" of over 256 high-speed processors does not fit on a single AGP-board (or PCI for that matter). What do you expect a new PC architecture to look like in 5 years? A Merced running at 3000 MHz? A mutiplrocessor Alpha with a 128-bit architecture? Gregor
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.