Author: Ed Schröder
Date: 04:56:17 09/05/99
Go up one level in this thread
On September 05, 1999 at 07:53:57, Ed Schröder wrote: >On September 05, 1999 at 07:33:40, Andrew Dados wrote: > >>On September 05, 1999 at 03:53:35, Ed Schröder wrote: >> >>>On September 05, 1999 at 02:54:08, Mark Young wrote: >>> >>>>On September 05, 1999 at 00:11:40, Robert Hyatt wrote: >>>> >>>>>On September 04, 1999 at 23:25:27, Mark Young wrote: >>>>> >>>>>>On September 04, 1999 at 22:17:37, Robert Hyatt wrote: >>>>>> >>>>>>>On September 04, 1999 at 18:35:32, James T. Walker wrote: >>>>>>> >>>>>>>>I feel bad for Ed and the Rebel team. It appears that ED had a corrupted >>>>>>>>version of Rebel loaded and it crashed 10 times during the match. It was still >>>>>>>>a good game in which the GM IMHO played very good.(As GM's are prone to do). I >>>>>>>>think his opening selection was perfect for him. He seemed to be in control >>>>>>>>most of the way after gambiting a pawn. >>>>>>>>Jim Walker >>>>>>> >>>>>>> >>>>>>>No.. the problem was the overclocked AMD machine. If you overclock, you ask >>>>>>>for trouble. And in general, if you ask often enough, you receive what you >>>>>>>ask for... >>>>>>> >>>>>>>I'm not counting this game in the "Is the computer a GM?" rating calculation, >>>>>>>as obviously Rebel was broken by the hardware. >>>>>>> >>>>>>>For the record, overclocking AMD is _bad_. I have had several horror stories >>>>>>>from Crafty users about this. It seems that the AMD engineers know what their >>>>>>>chips can run at, and going beyond that gets you into deep water. Intel seems >>>>>>>to be more conservative rating their chips... >>>>>> >>>>>>I heard that Ed said it was a bad version of Rebel that caused the >>>>>>problem....but I know your statement is true about AMD in general. >>>>> >>>>> >>>>>All I can say is what Ed said to me during the game... he also reported that he >>>>>had bogus problems in Paderborn... illegal moves in the PV, etc. It is >>>>>definitely possible that the program was broken, however. But I have seen log >>>>>files from crafty run on a misbehaving overclocked AMD, and they were pretty >>>>>amazing. Bogus scores, bogus moves, all sorts of nonsense... >>>>> >>>>>One person only had the problem after several consecutive games, indicating some >>>>>sort of thermal problem. Others just got bogus values, but things were fine >>>>>when the clock was set back to normal... >>>> >>>>He is what Ed posted on the Rebel Board. >>>> >>>>____________________________________________________________________________ >>>>Rebel lost. After 10 crashes the moves of the game are lost too. Rebel >>>>apparantly played with a wrong version. After 10 crashes I loaded the >>>>Sorin/Rohde version. No crashes since then but then it was already too late. I >>>>am sorry, it never happened in my whole life I played with a corrupted version. >>>>Expensive though! So most likely it was not the hardware. I have some homework >>>>to do! >>>>Ed >>>>____________________________________________________________________________ >>> >>>Some quick online notes, I will report more precisely later. >>> >>>I take the above back, it was a premature conclusion. After some more >>>testing after the game the version that played Sorin/Rohde also crashed. >>> >>>In the game Rebel played 40.Rg4 showing a score +2.30, that is ridiculus >>>of course. I can not reproduce this with any Rebel version I have. >>> >>>In any case yesterdays game was meaningless for the statistics and the >>>discussion if micro's are at GM level on tournamemt time control. Of >>>course the result remains including the financial consequences. >>> >>>Ed >> >> >> It is my understanding reading rules for GM Challenge that independent third >>party (Enrique Irazoqui) was supposed to verify Rebels moves during game. Why >>didn't he give you a clue that something is unreproductible/grossly wrong? >>(Meaning - obvious hardware failure while both program copies were identical). > >This was done in game one against GM Rohde. But then from game-2 and on I >switched to the special KryoTech machine. Enrique now checks the game >afterwards because the KryoTech machine runs twice as fast as a PII-450. Correction: because the KryoTech machine runs twice as fast as *his* PII-450 Ed Schoder >>-Regards- >> Andrew
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.