Author: Robert Hyatt
Date: 19:14:53 08/29/98
Go up one level in this thread
On August 29, 1998 at 15:14:46, Robert Henry Durrett wrote: >This bulletin is a generalization and continuation of the following threads: >(A) The Future of Computer Chess . . . >(B) Mr. Iacovoni's idea: How to program it. >(C) Threads related to Deep Blue and Hsu > >Related Questions: >(a) What is the trend in home "power-user" chess computers? Will the >dual-processor computers commonly used today for server applications find their >way into the home? this is already happening. IE two programs on ICC (fitter and cataclysm) are running on dual PII/400 systems, and the machines are at home, not being used as servers. The entire Linux SMP software development project is being done by people that have duals and quads *at home*. It isn't extremely common just yet, but there are far more duals at home that I would have guessed... because you can put together a PII/400 dual system for way under 2500 dollars, including big disks, lots of memory, and a monitor... pretty amazing... >(b) As the faster microprocessors [Merced, etc.] become available, will we still >see dual-processor servers, or will triple- or multiple- servers become the rule >rather than the exception? Or, will there be a return to single-processor >computers? Will these processors all have extremely high processing speeds? quad servers are already quite common, although only the pentium-pro can be used, until the new PII/xeon starts shipping. Several vendors have announced quad and 8-processor xeon servers already... >(c) Will chess-playing programs eventually see greater and greater >specialization? As it is now, "books" are used for openings, conventional >engines used for middlegames, and end-game programs used for endgames. Where >will this trend toward greater software specialization lead? >(d) Will the processors used for home chess applications be different, i.e. >specialized for chess, from those found in general-purpose home computers? this is doubtful. The cost of the PC is so low because of volume. Chess processors would still be pretty pricey at the volumes they would sell at, compared to a machine that can do *everything*. >(e) Will the computers used for playing chess for facilities like ICC, FICS, >etc., be big huge monsters which noone could afford to own in their personal >residences? this can happen. A 16 cpu alpha would be a monster playing chess, and a monster to pay for too... > >Discussion: >It seems to me that there are really the following theoretical issues here: >(1) Specialization in the hardware itself: Different processors designed for >different purposes. Each processor having a different hardware configuration, >(2) Specialization in identical multiple processors, all processors to work >simultaneously, and >(3) Specialization in the software itself. Different subroutines, subprograms, >modules, or even different programs available in a huge software package >[rivaling Microsoft's Office 97 in size] where only one small part [or a few >parts] of the total program works at any given time. >(4) A hybrid, such as different parts of the software package working >simultaneously on different processors. > >Extreme software specialization could see individual subroutines, subprograms, >modules, programs being optimized only for special cases. For example, one >might be specialized just to play bishop and pawns versus knight and pawns >games. > >QUESTION: Is this really all that "far out"? Where ARE we going? I think the next step is multiple threads in an engine, because these machines are becoming affordable, and may become a standard in the future. Special-purpose is going to always be a very small "niche" market because of the cost of buying the stuff and the cost of upgrading from time to time, compared to upgrading a GP machine..
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.