Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Adams-HYDRA Announcement

Author: Dann Corbit

Date: 21:14:58 05/24/05

Go up one level in this thread


On May 24, 2005 at 23:47:12, Ricardo Gibert wrote:

>On May 24, 2005 at 20:45:48, Vincent Lejeune wrote:
>
>>On May 24, 2005 at 17:13:45, Dann Corbit wrote:
>>
>>>On May 24, 2005 at 16:16:35, Ricardo Gibert wrote:
>>>
>>>>On May 24, 2005 at 14:57:25, Dann Corbit wrote:
>>>>
>>>>>On May 24, 2005 at 14:51:01, GuyHaworth wrote:
>>>>>
>>>>>>
>>>>>>Channel 4 TV has a story that Michael Adams is to play HYDRA in a 6-game match,
>>>>>>? in London next month.
>>>>>>
>>>>>>"The smart money's on the chip, which is 6x more powerful than DEEP BLUE".  No
>>>>>>reference to Adams tailing off badly at the Mtel event.
>>>>>
>>>>>Deep Blue hit one billion NPS (with full eval) peak and 200M NPS sustained.
>>>>>6x that would be 1.2 Billion NPS sustained.
>>>>
>>>>There is contradictory information on this. Hsu was finally pinned down on the
>>>>issue and it turned out that the 1 billion NPS figure is a theoretical peak. The
>>>>200 million NPS figure is the practical peak with 120 million NPS the sustained
>>>>figure. Of course, it is the latter figure that is most useful, but the 1
>>>>billion figure was tailor made for hyperbole.
>>>
>>>At 120M NPS (about 1000 times faster than Shredder on a 2.2 GHz machine)
>>>Six times that would rate at 720,000,000 NPS sustained for Hydra.  Certainly,
>>>this would be rather impressive if it were the correct figure.
>>>
>>>If doubling the speed gave one more ply, it would be 12 plies deeper than what
>>>my machine can perform.
>>>
>>
>>If my memory serve me well (it's not sure ;o) ), from operator chat on playchess
>>: "hydra do 400,000,000 N/S : 200 M/S in 32 CPU + 200 M/s in 32 FPGA"
>
>Given the estimated Deep Blue EBF of 4, that should be at least an order of
>magnitude greater of what is needed to overtake Deep Blue depth-wise.

There were two sorts of branching factor with Deep Blue.  There was a "robust
search" that pretty much exhausted a sphere around the root move, and a
speculative search that went a lot deeper.  I am not sure that the depth
conclusion can be assumed.

>>>>>>Mmm:  I guess the usual discussion will start but this seems hype to me.
>>>>>>However, it confirms that the 'DEEP_BLUE' is still taken as the 'SI unit' of
>>>>>>computer chess.
>>>>>
>>>>>It's the stuff that legends are made of.



This page took 0.01 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.