Author: Robert Hyatt
Date: 10:17:00 07/25/02
Go up one level in this thread
On July 24, 2002 at 19:31:45, Vincent Diepeveen wrote: >On July 23, 2002 at 11:36:36, Robert Hyatt wrote: > >>On July 23, 2002 at 10:40:00, Uri Blass wrote: >> >>>On July 23, 2002 at 10:06:18, Robert Hyatt wrote: >>> >>>>On July 23, 2002 at 04:18:46, Bo Persson wrote: >>>> >>>>>On July 23, 2002 at 00:32:44, K. Burcham wrote: >>>>> >>>>>> >>>>>> >>>>>>Some say that Deep Blue could analyze 200,000kns in some positions. >>>>>> >>>>> >>>>>This is more the *average* speed of the system. I have seen figures of up to 1 >>>>>billion nodes per second, in favourable (for speed) positions. >>>> >>>> >>>>I have a paper written by Hsu and Campbell that says their peak speed during >>>>the 1997 match was 360M for a single move. Of course, the machine was capable >>>>of bursts to 1 billion nodes per second, which I am sure it hit at times >>>>since that only required that all 480 chess procesors be busy at the same >>>>instant. >>>> >>>> >>>> >>>>> >>>>>This doesn't say, of course, what the search speed would have been in your test >>>>>position. Could have been 150M nodes/s, could have been 950M nodes/s... >>>>> >>>>> >>>>>The comparison of speed is also somewhat flawed by the fact that Deep Blue was >>>>>explicitly designed to be fast (as in nodes per second), which most of the other >>>>>programs are not. >>>> >>>> >>>> >>>>I don't agree with that. DB was a two-fold design: (1) fast, due to special- >>>>purpose hardware (2) good, due to adding whatever they though necessary into >>>>the hardware. >>>> >>>>Software programs don't have the option (2) available to them, so to keep >>>>option (1) viable, they compromise. DB didn't have to. >>> >>>In order to have good evaluation function you need to know what is good. >>>Programmers today know better than what they knew in 1997. >>> >>>Uri >> >> >>No we don't. Chess has been chess for 100 years. 20 years ago we knew what >>we "needed". And we knew what we could "afford". Today, what we "need" has >>not changed one iota, but what we can afford has changed drastically. DB just >>got to this point way before we got there... > >I always have agood laugh about this. Deep Blue had about 40 patterns, >some of which they considered complex (i don't doubt it). Where do you get this? Hsu's last paper said that over 50% of the total chip area was devoted to evaluation. It also says that they recognize over 8,000 "features". (page 3). > >Those were indexed all with arrays, which gives a couple of thousands >of 'adjustable' parameters which were adjustable. > >For a project which focussed more upon nodes a second than any other >thing (that's the only thing IBM ever made marketing with too, they >didn't care for the rest), and in order to get more nps , they even >used a lazy evaluation, which today gets used less and less by >the active programmers. They didn't use a lazy evaluation. They used the same fast/slow evaluation as used in Belle, which is a _long_ way from "lazy evaluation". Bruce still uses this, if you recall ( true lazy evaluation ) and it works just fine for him. And for me... > >I assume that when Brutus' evaluation gets improved he'll soon have to >do without lazy evaluation too. Of course the words 'lazy evaluation' >and 'quick evaluation' are basically meaning the same here. > >It proves how much DB was focussed upon getting more nodes a second. >They sure managed that part. 126MLN nodes a seconds with peeks up to 300+ >in endgames is pretty impressive. > >Of course that's the only good thing about the whole machine, and we >will forgive them for that. It was already hell of a work to get so >many nodes a second. > >If i remember well, somewhere was posted that in order to get more >nps, even OLD processors (with bugs in eval probably) were mixed with >newere processors of faster speed. Just to get more nodes a second. Not in deep blue 2. And no bugs. They just had some runs of the chip that would only run at 20mhz, while others would run at 24mhz. They did mix those. > >I would be completely sick if i would run with my latest diep in a tournament >and then in order to show SGI/NWO a bit more nodes a second i would add a bunch >of processes from old diep versions in order to get more nodes a second :) This didn't happen with DB 2... > >Best regards, >Vincent
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.