Author: Sune Fischer
Date: 07:08:18 04/14/02
Go up one level in this thread
On April 14, 2002 at 08:07:22, David Dory wrote: >On April 13, 2002 at 05:11:15, Sune Fischer wrote: >>>All programs experience diminishing returns with greater depth. The programmers >>>work harder and harder for a smaller increase in playing strength. Nevertheless, >>>it is hardware not software, that has provided the largest increase in chess >>>program strength in the last 20 years. >> >>That is a myth. >>You only have to take one look at the SSDF list to see there are significant >>improvements from version to version on the same hardware. >> > >My point, which is not a myth, is that the improvement from better hardware has >certainly _exceeded_ the improvement from software in chess playing strength. > >My first chess playing computer ran on a Z80 at 4MHz, (an 8 bit CPU). Consider >the gain made by taking that same program (the Spracklen's Fidelity program) and >moving it (not changing it except to re-compile it for the better processor), it >to an Intel P4 at 2GHZ at 32 bits. > >Let's do a rough arithmetic: 8 bit to 32 bit, = 4 X improvement > + 2000 Mhz/4Mhz = 500 X improvement > ========================================= > Sub Total = 2,000 X improvement > >And that's just the main CPU. Then we have the faster memories (and lots more of >them. My first PC had 512K of memory, and a 20MB Hard Drive, now I've got a Gig >of memory and a 100GB HD (great for all those TB files). And of course the disk >access times have shrunk from about 22ms to about 7ms, so: > >Memory improvement = 1000Megs/0.5Megs = 2,000 X improvement >Hard Disk improvement (size) = 100,000MB/20MB = 5,000 X improvement >Hard Disk speed improvement = 22ms/7ms = 3 X improvment > ========================================= > Sub Total = 20,000,000 X improvement > >So HOW MUCH OF AN IMPROVEMENT IS THAT, SUNE? My calculator won't go that high! > >If you want to talk percent improvment for the hardware, be sure to add two >zeroes onto the end. :-) Obviously you have no idea what the h*ll you are talking about. 5,000 x 3 improvement because of increased harddisk size and speed, are you serious? Memory size, sure bigger hash helps, but a factor of 2,000? Are you serious? Besides, there has been software development into how to maximize the use of the hash, so I'd say a good deal of the improvement here is actually better software design. Going from 8 to 32 or even 64 bit is part software and part hardware, you can design better structures when they can carry more information. For instance I don't think rotated bitboards was in anyway worth considering on the 8 bit machines. I will agree to one of you numbers, the factor of 500 in speed from the very early days, this is the factor you need to put up against the development in software. Let me teach you some real math here: 500 ~= 512 = 2^9 every double is about 50 points (that is the estimate most people use disregarding DR) so that is 450 elo because of hardware. Current top engines are rated around 2700 so that would put the very early engines at 2250 elo! Where they that strong, I think not! >Naturally, no program will use all these hardware improvements 100%, but I think >anyone with a little sense can see that HARDWARE has made the largest >improvements to chess playing strength, NOT SOFTWARE. Think again please. And maybe show some respect for all the hard work that goes into Fritz and Tiger and Shredder and the others, I'd say it is _at least_ 50% software progress now a days. >Indeed, many features of the top programs wouldn't even be practical at all on >original PC hardware, let along the woeful hardware of a BORIS or CHESS >CHALLENGER. You got it, software has changed. >Dave -S.
This page took 0.01 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.