Author: Robert Hyatt
Date: 15:00:48 05/17/00
Go up one level in this thread
On May 17, 2000 at 16:00:42, Joshua Lee wrote: >I was wondering the Ratings of the Different Mainframe computers of the past >like Deep Blue 1 and 2 and Deep Thought and how many ply they looked ahead? >Also What Clock Speed would a PC need to be equivalent? These below are what i >have found while the Processor speed if on a PC is my Guess. > >I have found different things on the internet. Here is my list: > Deeper Blue 1997 15 Ply 2817 (USCF) 200 Million NPS >2.048Thz The above is wrong, based on their output posted on their web site. They were searching (typically) 10-11 plies in software, plus another 5-7 plies in hardware. IE the very minimum depth they would see was maybe 15 plies, but more common numbers were 17-18. Look at the IBM web site, under the Deep Blue section, and notice the depths, which are given like this: 11(6). That means 11 plies in software + another 6 plies in hardware, for 17 plies total. > Deep Blue 1996 12 Ply (*)14 Ply 2642(FIDE?) >(*2535) 100 Million NPS 1.024Thz See above. DB2 was about 2x faster, although it had a much bigger evaluation than DB1. The depths were fairly close, with DB 2 at best going 1 ply deeper. > Deep Thought II 1991 6-7 Million NPS 64Ghz > Hiarcs 7.32 2009 6.272.000-6.308.000 >64Ghz > Cray Blitz T90 ???? 5-7 Million NPS Cray Blitz could hit this number on the T932, which was first delivered maybe 5 years ago or so. We didn't play many games, but did see search depths of 11-12 plies in the middlegame... We didn't use recursive null-move R=2 which would likely have added at least 2 plies to that figure. > Hiarcs 7.32 2008 3.136.000-3.172.000 >32Ghz > Deep Thought 1989 2480-2500 (FIDE) 2 Million NPS >20Ghz? > Hiarcs 7.32 2006 1.568.000-1.604.000 >16Ghz > Cray Blitz C90? 1995 1.5 Million NPS About right. depth of 10-11 plies. the T90 added about 1 ply. > Hiarcs 7.32 2005 784,000-820,000 8Ghz > Deep Thought 0.02 1988 2551(USCF) 2451(FIDE?) 750,000 NPS* >7.453Ghz > Deep Thought 0.01 1988 720,000 NPS > Chiptest-M 1987 500,000 NPS > Hiarcs 7.32 2003 392,000-428,000 4Ghz > Hitech 1995 200,000 NPS > Hiarcs 7.32 2001 196,000-232,000 2Ghz > Hitech 1988 2413(USCF) 2363 or 2313FIDE 175,000 NPS > Hitech 1985 2350 > Belle 1983 2263 160,000 > Hiarcs 7.32 2000 98,000-134,000 1Ghz > Hiarcs 7.32 2000 2660(LCTII)2460(FIDE EST) 87,000-123,000Nps >Middlegame > Chiptest 1985 50,000 Nodes Per Sec > Belle 1978 > > ELO and Ply Depth > 2650 15Ply > 2625 14Ply Only this deep in >the Endgame (Deep Blue in the Middlegame) This >Matches elo - Ply > 2600 13Ply > 2575 12Ply Deep Thought 1989 >this Matches elo-ply > 2550 11Ply Sometimes 11Ply in the >Middlegame > 2525 10Ply > 2500 9Ply But Usually 8 or 9 > 2475 8Ply Hiarcs Looks atleast 8 >Complete Ply in The Middlegame (this matches LCT II -200 (2690- 200 =2490) > 2450 7Ply > 2425 6Ply > 2400 5Ply > 2375 4Ply > 2350 3ply Hitech 1988 However >this doesn't match up obviously not everybody is playing with the same Ply hehe >haha... >I've heard before how much doubling clock speed results in but it was an SSDF >thing(which must be atleast 150points too high??) but against humans this is a >different story. > >Some of the things i've run across online have mentioned the experiments to >figure out ply in relation >to elo they said sometimes 1Ply deeper meant like 80-192elo >and that knowledge meant 200-300 more elo but i am not sure about this i just >want to hear other opinions. Thankyou > > > doubling the cpu speed is generally said to produce 50-60 rating points. Since the typical effective branching factor is around 3.0, every time the speed is tripled, we get another ply, and using the 60 point figure above, a ply would be about 90 rating points, roughly. But there is nothing that says that as we go deeper this doesn't taper off. Nor is there anything that says that as we go deeper, the gain doesn't actually get larger... > > > >*according to the Deeper Blue - Kasparov Match article in Chess Life >http://www.frc.ri.cmu.edu/~hpm/book97/ch3/ >http://mitpress.mit.edu/e-books/Hal/chap5/five4.html >http://www.geocities.com/SiliconValley/Lab/7378/comphis.htm >http://foxnet.cs.cmu.edu/people/spot/frg/Berliner.txt >see also Dejanews search for FAQ WWW Hypertext Edition (rgc Groups FAQ) >In 1988 DEEP THOUGHT and Grandmaster Tony Miles shared first place in the U.S. >Open championship. DEEP THOUGHT had a 2745 performance rating. This matches up > 2745-180 = 2565 > >In 1985 HITECH achieved a performace rating of 2530 >2530-180 = 2350 > >In 1983 Belle became the first computer to beat a master in tournament play and >the first computer to gain a master rating (2263) This is wrong. Belle became the first program with an established rating of over 2200. (2208 was the number.) They were awarded the title "life master" by the USCF at the 1983 WCCC event in NY. Cray Blitz beat a master in 1981 to win the Mississippi State Chess Championship. The master it beat was Joe Sentef, and the game was published in Chess Life in late 1981 or early 1982.
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.