Author: Mike Byrne
Date: 18:05:59 03/01/05
Go up one level in this thread
On March 01, 2005 at 20:09:03, Steve B wrote:
>In 1989 Fidelity released the only Dual Processor Chess computers ever made
>the Elite A/G version 5(Dual 68000) and the version 8(Dual 68020)
>exactly 50 units of each were produced
>
>in going over the manual for the dual processor Elites i came across an
>interesting factoid
>
>a rather lengthy and technical addendum is inserted in the manual written and
>signed by the programming team of Dan and Kathe Spracklen
>
>they discuss how the dual processors do not provide a 2x speed up which would be
>ideal,but rather a 1.7 speed up
>
>they attributed this loss of speed up to several "overhead" factors such as:
>Communication Overhead
>Synchronization Overhead
>Divided Search Overhead
>and Split Hash Overhead
>
>they provided a new feature for the computer whereby a user could turn off the
>dual processors leaving the one
>they then suggested to try different positions to see what actual speed ups
>could be achieved
>
>the maximum speed up they saw was 1.7 times
>
>i dont know if a current "off the shelf" Dual Processor PC allows the user to
>turn off one of the processors,but if so,i wonder..for todays dual processor
>PC'S what the speed up achieved is?
>i imagine it could not be the ideal 2x as there must be something lost to some
>operating processes
>
>
>for example
>
>using the two processors how long does the computer take to achieve say a 15 ply
>search as opposed to the one processor?
>
>Double Best Regards
>Steve
It varies by position and by hashize. I would say Crafty has a =>75% efficiency
rating on average. Meaning that if were to examine all the positions that occur
in the course of the game, on average Crafty would reach the same depth in .
For example, if crafyy reaches a certain depth xx in 200 seconds, with the
second processor, he would reach that position on average in 114 (200/1.75)
seconds or better.
[d]r3r1k1/5p1p/bqn1pPpB/3pP3/ppnP4/P2Q1N2/1PB2PK1/1R5R w - - 0 1
single
time=5:30 cpu=99% mat=-1 n=683420646 fh=87% nps=2.07M
ext-> chk=48311578 cap=1384742 pp=6532966 1rep=9947453 mate=135870
1
predicted=1 nodes=683420646 evals=153441359 50move=0
endgame tablebase-> probes=0 hits=0
SMP-> split=0 stop=0 data=0/64 cpu=5:29 elap=5:30
mate in 8 moves.
White(1): Bg7
time used: 5:30
1.75 factor would indicate a time of 3:08
with dual
11-> 2:38 Mat08 1. Bg7 Ne3+ 2. fxe3 Qxd4 3. Rxh7 Qg4+
4. Kf2 Qxf3+ 5. Kxf3 Nxe5+ 6. Kg3 Kxh7
7. Rh1+ Kg8 8. Rh8#
time=2:38 cpu=197% mat=-1 n=618055851 fh=80% nps=3.89M
ext-> chk=32809285 cap=1397492 pp=8041413 1rep=6527154 mate=145069
3
predicted=0 nodes=618055851 evals=186176453 50move=0
endgame tablebase-> probes=0 hits=0
SMP-> split=726 stop=135 data=7/64 cpu=5:13 elap=2:38
mate in 8 moves.
White(1): Bg7
time used: 2:38
in this case, Crafy was very efficient - showing more than 2x faster. But
this is unusual case.
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.