Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Faster, deeper and more of such...

Author: Robert Hyatt

Date: 22:23:11 09/14/00

Go up one level in this thread


On September 14, 2000 at 23:39:29, walter irvin wrote:

>On September 14, 2000 at 18:40:37, Robert Hyatt wrote:
>
>>On September 14, 2000 at 14:42:28, Dan Ellwein wrote:
>>
>>>On September 14, 2000 at 14:39:37, Robert Hyatt wrote:
>>>
>>>>On September 14, 2000 at 14:36:41, Christophe Theron wrote:
>>>>
>>>>>On September 14, 2000 at 14:33:49, Dann Corbit wrote:
>>>>>
>>>>>>On September 14, 2000 at 14:30:07, Dan Ellwein wrote:
>>>>>>[SNIP]
>>>>>>>I guess it would be impractical to run this test with opening book disabled...
>>>>>>>
>>>>>>>(startin' with the very first move have the computer think on its own)...
>>>>>>>
>>>>>>>but i wonder what the data would look like if you did...
>>>>>>>
>>>>>>>it may be that there would not be a cut-off at iteration 19...
>>>>>>
>>>>>>My guess is that in the first ten moves no program on earth can get to ply 19
>>>>>>unless it does a ludicrous amount of speculative pruning.  Even 16 plies would
>>>>>>be formidable.
>>>>>
>>>>>
>>>>>We are close.
>>>>>
>>>>>
>>>>>    Christophe
>>>>
>>>>
>>>>DB was "there" in 1997.
>>>>
>>>>:)
>>>
>>>Bob
>>>
>>>What would it take to get Deep Blue up and running and give us some data
>>>comparable to what Ed has done here...
>>
>>
>>Several million dollars, an act of congress, and divine intervention, I am
>>afraid.  The several million dollars part is the _easiest_ of the required
>>events.  :)
>>
>>We do have some data for 6 games vs Kasparov.  Someone could hand-compute the
>>above for DB using that data, and get a rough approximation of what kind of
>>'change expectancy' it had for each additional ply.
>
>i figure your as good as the deep blue teem , what sort of hardware would you
>require to compete with deep blue and what program would you use???

I think your figuring is probably inaccurate.  The DB guys were _very_
bright.  And there were several of them.  I can guarantee you I would
rather work in a setting with 4 bright chess people at my elbows, as
opposed to a setting with zero.

I could "compete" with any hardware.  But if you mean "play equally or better
than" when you say compete, then there is no hardware available today that I
would want to carry into such a match.  I can probably hit 1/10th of their
search speed, maybe even 1/5th with some _real_ sophisticated hardware.  But
I couldn't do what they did in their evaluation without dragging performance
back down by at least a factor of 10x...  so that in reality, I might hit
1/100th to 1/50th of their effective speed.  If I got lucky with hardware.



> could crafty
>be changed or would you use something like the cray blitz type program.plus how
>long would it take you to put something like that together if getting the
>hardware was not a problem .like a similar situation with IBM and DEEP BLUE
>,where money was not an issue .do you think with current mainframe tech you
>could duplicate deep blues results ????? or even do better?????

No.  General purpose hardware is always a factor of 100 or so behind the speed
of special-purpose hardware.  In 1997, I would put this advantage at 1000x for
DB over the best PC hardware available.  In 2000, I would say closer to 200x
counding the 8-way xeon machines.  Or if you go to the best of the best, the
numbers I gave are probably close.



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.