Author: Slater Wold
Date: 11:08:28 11/29/01
Go up one level in this thread
On November 29, 2001 at 13:41:13, Robert Hyatt wrote: >On November 28, 2001 at 19:05:19, Slater Wold wrote: > >>On November 28, 2001 at 17:53:43, Uri Blass wrote: >> >>>On November 28, 2001 at 17:15:49, Slater Wold wrote: >>> >>>>On November 28, 2001 at 16:34:12, Uri Blass wrote: >>>> >>>>>On November 28, 2001 at 16:03:42, Slater Wold wrote: >>>>> >>>>>>On November 28, 2001 at 15:23:32, K. Burcham wrote: >>>>>> >>>>>>> >>>>>>> >>>>>>>does anyone post here that works for chessbase? >>>>>> >>>>>>Yes. But they will never answer. I've written TO chessbase, and never got a >>>>>>decent answer. >>>>> >>>>>I remember that they answered to me few years ago about other things when I had >>>>>Fritz5. >>>>>The fact that they did not answer also does not prove that they will never >>>>>answer. >>>> >>>>Example of Chessbase Customer Service: >>>> >>>>When I purchased HIARCS 7.32 (several years ago) it came with very few TB's. I >>>>wanted to know if I could use the TB's on Hyatt's FTP machine, so I went to >>>>Chessbase's website, and filled out a form to ask. >>>> >>>>A week later I had no response, so I e-mailed Bob. Within 1 hour, I had an >>>>answer. Not from Chessbase, but from Bob. >>>> >>>>I believe it was about 3 weeks before I got an e-mail that simply said "yes" in >>>>it, from Chessbase. >>>> >>>>I'd venture to say Hyatt gets 2x the e-mails a day than Chessbase does. And >>>>he's just one man. >>>> >>>>I've learned to expect this. Not because I dislike Chessbase, or dislike their >>>>products. I've learned to expect this, because this is what they have shown me >>>>they will do. >>>> >>>>>> >>>>>>>has anyone talked to Frans Morsch or Mathias Feist about the difference in kns >>>>>>>when using one processor or when using two processors with deep fritz. >>>>>> >>>>>>Not that I know of. But I don't know anyone who talks to them. >>>>>> >>>>>>>has anyone talked to chessbase about the low kns when deep fritz is on two >>>>>>>processors? >>>>>> >>>>>>I have tried. They don't have very good customer service. >>>>>> >>>>>>>has anyone discussed a patch? or is this low kns for deep fritz what the >>>>>>>programmers intended when running the deep version? doesnt make sense to me, >>>>>>>based on other deep programs kns increase. >>>>>> >>>>>>It's just bad programming. And I will continue to think so, until someone from >>>>>>Chessbase can explain to me otherwise. >>>>> >>>>>I prefer not to call it bad programming but inferior programming. >>>> >>>>Ok. So sugar coat it. Whatever. Still the same thing. >>>> >>>>>It is not easy to teach chess programs to use more than one processor in a >>>>>correct way. >>>> >>>>Not. It really isn't easy. Bob has said it took him years to perfect his SMP >>>>code. >>>> >>>>Not easy to build a car though either. Go buy a brand new car, and have it fall >>>>apart on you. How would you feel? >>> >>>I did not say that your complain against chessbase is not justified and "it is >>>not easy" was only to defend the programmers. >>> >>>The problem is not the programmers but the fact that chessbase did not tell >>>correct information about Deep Fritz. >>> >>>Uri >> >>I'd bet any amount of money that Chessbase probably quoted Frans with that >>"..80% nonsense.." >> >>If they are allowing anyone to sell their product that does not work in the >>fashion it is being advertised, there is little you can do to defend them. >> >>Frans and Mattias are great (chess) programmers. No doubt about it. But the >>SMP in DF doesn't work. Plain and simple. > > >Here is a question for you: can you do a one-processor search for exactly 1 >minute and see how many nodes it searched? (ignore knps stuff). Then do a >two-processor test and again see the total nodes searched? Compare those two >numbers. If SMP is working, they should be almost exactly doubled. (note >that this is not 2x faster because some of those nodes are searched but were >not necessary). > >I have a hell of a time getting my NPS value to display right on windows. I >don't know why. But my nps value is always screwed up, probably due to the >elapsed time that I use to compute the number... I have posted that result twice on this forum. It will be a 1.2x speedup or so. When I get home I will run a test; 1 position using 1 and 2 CPU's for about 10 minutes. I will post them here. I promise a speedup of no more than 1.2x. And if I can find the position, I will use the one where 2 CPU's is slower than 1 CPU. Should be around 6pm central.
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.