Author: Robert Hyatt
Date: 06:33:14 10/20/00
Go up one level in this thread
On October 20, 2000 at 00:13:18, Dave Gomboc wrote: >On October 19, 2000 at 21:08:06, Robert Hyatt wrote: > >>On October 19, 2000 at 16:51:49, Albert Silver wrote: >> >>>On October 18, 2000 at 09:56:52, Robert Hyatt wrote: >>> >>>>On October 18, 2000 at 05:44:36, Andrew Williams wrote: >>>> >>>>>My apologies if this is old news. >>>>> >>>>>There was an account called DeepBlueJr watching Kramnik-Kasparov, game 6 >>>>>on ICC last night. The finger notes said the owner was Murray Campbell >>>>>(a couple of the admins stated that it was indeed DeepBlueJr). I asked >>>>>him what exactly he was using and he said that it was a 24-processor >>>>>version attached to a R/6000. He said the processors were the same ones >>>>>that ran in the Deeper Blue that beat Kasparov. I asked him what NPS he >>>>>was getting and he said "looks like 28M" (!). I also asked him what sorts >>>>>of depths it was searching, but he didn't answer. Another thing he didn't >>>>>answer was my suggestion that he join CCT2 :-) >>>>> >>>>>The program wasn't kibitzing automatically like crafty does, but he was >>>>>occasionally cutting and pasting the analysis into channel 211. This was >>>>>around the time that Kramnik was on top, before the win seemed to slip >>>>>away from him. >>>>> >>>>>Andrew >>>>> >>>>>PS Just in case anyone was wondering, I asked him if he minded my reporting >>>>>this conversation here and he was quite happy for me to do so. >>>> >>>> >>>>I was surprised, too. But I did chat with him a good bit and am convinced that >>>>it was Murray. I will try to bug him a bit about CCT2, but I have a suspicion >>>>that IBM won't allow public exhibitions like that... >>> >>>Really.... Wow. When he posted that, I was convinced some joker had set up the >>>pseudo as a prank. (Sorry about that Andrew) >>> >>>Did you ask him about the tuning? Have they done any work on refining the eval >>>or is it unchanged from the time of the match? 28 million NPS... Wonder how much >>>that cuts in to the depths compared to its bigger brother. >>> >>> Albert >> >> >>I didn't understand the 28M number, but my screen was so danged busy scrolling >>all the nerdy comments from the 1000 people observing the game, that I didn't >>get a chance to ask him. 24 processors at >= 2M nodes per second per processor >>should be 48M. I assumed he typoed when he said 28M. >> >>It (DB Jr) was a holy terror against GMs in the many exhibitions they played >>prior to the final DB match. This machine would be less than 10x slower than >>the real machine. I would think it would be a handful for anybody. > >24 processors? Hmm... 16? > >Dave He definitely said 24. Of course he might have meant 16. 24 caught my eye as that would be three 'modules' of DB hardware. No reason why it couldn't be right now, as the workstation he was using might be faster than the one they had when they were touring the world with DB Jr. And a faster host would need more chess processors to maintain the right "balance".
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.