Computer Chess Club Archives

Search

Terms

Messages

Subject: Re: Hydr's KN/s

Author: Dann Corbit

Date: 11:37:46 07/05/05

On July 04, 2005 at 08:17:17, Vincent Diepeveen wrote:

>On July 04, 2005 at 00:21:38, Madhavan wrote:
>
>>On July 03, 2005 at 10:12:55, Vincent Diepeveen wrote:
>>
>>>On July 03, 2005 at 08:18:48, Amir wrote:
>>>
>>>>On July 03, 2005 at 07:21:55, Vincent Diepeveen wrote:
>>>>
>>>>
>>>>>By the way, my own experiments at supercomputers indicated that 16 cpu's is most
>>>>>easy to get a good speedup from, relative seen. 32 cpu's it is a lot hardware,
>>>>>but putting 32 cpu's to work is real easy. Yet the extra speedup you get from
>>>>>moving from 16 to 32 cpu's is not so big. Not only speedup gets bigger, also the
>>>>>worst case speedup still is utmost tiny.
>>>>
>>>>Actually, with 16 CPUs they were saying Hydra does 40 million nps .. so the
>>>>juump was really big if now they are doing 220 million nps
>>>
>>>hydra machine has new fpga cards which are more than 2 times powerful and
>>>Chrilly has been fighting long to get a node faster.
>>
>>What would be the purpose of getting node faster?Does it mean the software
>>searches deeper depths?Is Hydra perfectly programmed,was there no flaw in its
>>search?
>>
>>Does Hydra get more speed search than Deep Blue of 1997?
>
>Deep Blue searched 10-12 ply as you can see in logfile.
>Hydra searches on average 2 times deeper.

The logfile does not consider the depth on-chip at the leaves.  About 6 plies
more.  So consider it really to be 16-18 plies.

>The purpose of more nodes per second is mostly in the simple imagination of the
>mankind, who is not technical strong enough to realize that plies and nodes per
>second isn't the most important thing, but yes Hydra gets far more nps than Deep
>Blue too.
>
>An optimistic estimation from Deep Blue team was they searched on average 130
>million nodes per second.
>
>So even the current hydra machine is far stronger in all respects.

Re: Hydr's KN/s Gian-Carlo Pascutto 02:05:50 07/07/05
- Re: Hydr's KN/s Dann Corbit 10:56:04 07/07/05
  - Re: Hydr's KN/s Gian-Carlo Pascutto 11:14:36 07/07/05
    - Re: Hydr's KN/s Dann Corbit 11:37:19 07/07/05
      - Re: Hydr's KN/s Gian-Carlo Pascutto 11:51:56 07/07/05
        
        Re: Hydr's KN/s Dann Corbit 12:01:26 07/07/05
        
        Re: Hydr's KN/s Dann Corbit 12:43:03 07/07/05
        
        Re: Hydr's KN/s Vincent Diepeveen 01:05:46 07/08/05
        
        Re: Hydr's KN/s Tony Werten 00:15:26 07/08/05
        
        Re: Hydr's KN/s Vincent Diepeveen 01:44:18 07/08/05
        
        Re: Hydr's KN/s Dann Corbit 13:34:58 07/07/05
        
        deep blue parallel splitting and branching factor Vincent Diepeveen 01:23:07 07/08/05
        
        Re: deep blue parallel splitting and branching factor Dann Corbit 10:24:19 07/08/05
        
        Re: Hydr's KN/s Dann Corbit 15:24:53 07/07/05
        
        Re: Hydr's KN/s Gian-Carlo Pascutto 23:20:50 07/07/05
        
        Re: Hydr's KN/s Ricardo Gibert 13:17:41 07/08/05

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.