Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Could Deep Blue Have done better than D.J Using the same Primergy ?

Author: Jorge Pichard

Date: 21:10:04 07/11/00

Go up one level in this thread


On July 12, 2000 at 00:02:20, Jorge Pichard wrote:

>On July 11, 2000 at 22:54:08, Dann Corbit wrote:
>
>>On July 11, 2000 at 22:22:05, Jorge Pichard wrote:
>>
>>>On July 11, 2000 at 22:01:13, Dann Corbit wrote:
>>>
>>>>On July 11, 2000 at 21:17:40, Jorge Pichard wrote:
>>>>
>>>>>Graham Laight wrote that in 1997 Deep Blue did better than Deep Junior is doing
>>>>>now; whenever it got a significant advantage, it won against Kasparov.
>>>>>
>>>>>My answer to that comparison is that if you compare the difference in hardware
>>>>>between Deep Blue Super computer to the Primergy netserver, it is like
>>>>>comparing the Space Shuttle to a Helicopter in Speed. Deep Blue could not have
>>>>>done better against this type of opponents, using the same Primergy Netserver.
>>>>
>>>>The comparison is even more pure nonesense than you are relating.
>>>>
>>>>The architecture is completely different.  Deep Blue used a hardware solution.
>>>>Deep Junior is a software solution.  Apples and Oranges if there ever were any.
>>>
>>>I was not comparing the architecture of the hardware, what I am trying to point
>>>out is that given the calculating power of the two different systems, there is a
>>>ratio of a least 200 to 1. Therefore, the advantage of using a system which is
>>>capable of calculating 200 times faster, simply indicate that Deep Blue Software
>>>was not really superior to Deep Junior, but the advantage of its calculating
>>>power will make any top Commercial software rates by SSDF play like a super GM.
>>
>>It's strictly a nonesense comparison.  Superiority or inferiority of the
>>software piece is something we will never know, either.
>>
>>Imagine a CPU that has a square root instruction.
>>Imagine a second CPU that emulates the square root in software.
>>
>>You might do things very differently when you write your programs if you know
>>you have a 20 cycle hardware square root instead of a 200 cycle software square
>>root.
>>
>>There is no meaningful comparison that can be made, and "counting nodes" is
>>little more than pure nonesense to try to prove which one is/was better.
>>
>>The hardware is completely different.
>>The software is completely different.
>>The algorithms are completely different.
>>
>>One of them is no longer available.
>
>It is obvious that most chess softwares have different algorithms unless it is a
>clone. As far as the hardware differences, nodes per seconds is not pure
>nonsense. Now here is the final logical conclusion, is you take Chess Junior 5.0
>and install it inside of a Pentium III 850 Mhz which of course it will produce a
>bigger count of nodes per second, than if you install the latest version of
>Fritz 6a inside of a 80486 50 Mhz II, you will expect Junior to win a least
>75% of the games.

>If I was to compare Deep Blue speed than I would have to compare it to an earlier 80486 to make the ratio 200 to 1.

>PS: In my previous example the softwares and the hardwares were differents, but
>the difference in speed made a big difference.
>
>Pichard.



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.