Author: James T. Walker
Date: 06:10:08 03/31/99
Go up one level in this thread
On March 31, 1999 at 07:58:53, Christopher R. Dorr wrote: >On March 30, 1999 at 21:57:39, Bruce Moreland wrote: > >> >>On March 30, 1999 at 20:42:45, Christopher R. Dorr wrote: >> >>>Again, we come back to exactly what is the question? If the question is 'What is >>>the strongest PC-based program out there?' then the answer may be one thing. >>>What is the strongest program in the world? Unquestionable Deep Blue. Will you >>>disqualify DB from being the strongest because it runs on a machine that cannot >>>be afforded? Shall we say 'But if DB was running on a PII 450, then it would >>>only be rated xxxx?' If we won't say that, then why should we say that, to judge >>>Crafty, we should put it on hardware it is not optimized for. >>> >>>If we want to say 'What is the strongest PC program running on a commercially >>>available single processor system?', then fine...it is an interesting and >>>valuable question. >>> >>>But if the question is 'What is the strongest PC based program?' *period*, then >>>a strong case must be made for Crafty running on a quad. There was a person on >>>ICC this afternoon, saying he had a Quad Xeon overclocked (to 600+ Mhz), and was >>>hitting NPS's in the neighborhood of 2,000,000 (2 million NPS) with Crafty 16.6. >>>If this is the case, I'll take this Crafty vs. *any* PC program running on >>>anything. >>> >>>I think, perhaps, we may be looking at this question differently. >> >>You are responding to Uri but you may as well be responding to me with this last >>sentence. >> >>You're mixing things up in this post. You are talking about "the strongest PC >>based program", you think. But what you are really talking about is a chess >>system including software and hardware. > >How do you seperate the two? When we talk about Fritz or Rebel, we are talking >about them running on some kind of minimum system. Fritz5 isn't going to run on >a 286. Perhaps I am talking about something different (which was the point of my >previous post) than some others here are. And not only am I having some >arguments with the terminology, but with some concepts as well. When you talk >about 'the strongest' program, are you talking about best in computer-computer? >That's very likely different from computer-human or problem set testing. I think >we need to be more specific in our definitions. And I don't think this is a >pedantic request either. Some people are very interested in computer-computer >testing, others (like me) are much more interested in results against humans. >And against humans, I'm interested in both platform equal performance, and it >overall performance. If Crafty on a quad is the best in the world at speed, then >I'll spend some time watching it on ICC simply because it is the best, >regardless of it's performance on a single processor. > >> >>I don't have any problem saying that Crafty on a quad 450 would be extremely >>strong, but I wouldn't call that a "program", I would call that a "system". >> >>I think that it is very hard to talk to you about this since your comparisons >>and your conclusions aren't very typical. >> > >Perhaps not typical, but still valid. But what I'm really asking is the above >question. If we are going to ask 'What is the strongest program on a single >processor system, as measured by computer-computer testing?', then that gives us >some important and useful information. But different information from the >question I stated in the last paragraph. Asking for 'the strongest' program is >like asking about 'the fastest' car. There are many caveats, and if they aren't >taken into account in the formulation of the question, then the question becomes >virtually meaningless. > >>What was the account of the guy with the overclocked Xeon? > >I don't remember the account name. He was chatting in the computer channel. He >said he was going to start running Crafty on it on ICC. Perhaps someone else >remembers his name, but if he really is getting 2 million NPS with Crafty, I >think we'll all be hearing about him shortly. Don't know why someone would >overclock an $8,000 machine like that (from 450 to over 600), but it should be a >killer before it fries itself. > >Chris > > >> >>bruce Hello again Chris, I agree with Bruce to some extent. You are talking more about a "system" than the program itself. You also seem to be ignoring the "Multi-processor" Fritz which does exist even if it does not play on ICC and is not available. Your assumption that Crafty would be the best is mostly based on the idea of running it on a Multiprocessor system where no other program has that ability. You simply have no data on Fritz running on the same system. Also this multiprocessor thing is not all that new. Fidelity sold a version of the Elite Avante Garde which was based on two 68000 processors and they claimed a 70% increase in speed (Year 1991). Besides the original post was "Why does Fritz beat Crafty". And of course the poster was not implying that Fritz on a single processor would beat Crafty on a Quad. Also I took it as being asked in a way as to invoke responses that might point out some ideas for improvement in Crafty which would reverse the outcome. Some people seem to feel necessary to come to the defense of Crafty rather than help provide useful information which would improve Crafty. I believe that Crafty should come at a cost to all who use the code to learn programming or get ideas from it which improve their own programs. I think that cost should be that everyone "Using" Crafty in that way should have to come up with at least one original idea for improving Crafty. Bob of course could accept the suggestions or flush them down the toilet at his discretion. But there is always the possibility one person might see something usefull which Bob might take and improve on and make Crafty better. Then all would benefit. By the way Crafty does not need defending. It can stand on it's own merits. I also believe that Crafty may be better against humans than it's computer-computer testing shows. I also believe that Version 16.6 has made an improvement which is aimed more at computers than humans (IMHO). The last 134 games at game/10 played against Junior 5 shows Junior 5 won by only 71.5-62.5. This indicates only 25 points difference. However this performance improvement does not seem to show up against Fritz or Nimzo99. I'm not sure what this means but just by improvimg play against one opponent makes the overall strength rise. I think Bob could make Crafty stronger against computers if that were his primary goal. Jim Walker
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.