Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Chess pc program on super computer

Author: Vincent Diepeveen

Date: 04:57:26 08/04/05

Go up one level in this thread


On August 04, 2005 at 07:43:04, Bernhard Bauer wrote:

>On August 04, 2005 at 06:18:35, Mark jones wrote:
>
>>On August 04, 2005 at 05:22:58, Bernhard Bauer wrote:
>>
>>>On August 04, 2005 at 02:37:20, Mark jones wrote:
>>>
>>>>Can you imagine how would Junior,shredder,fritz would have played if they where
>>>>deployed on A super computer like this:
>>>>http://www.top500.org/sublist/System.php?id=7605
>>>>
>>>>If this were possible not only it would kill all the humans I think it would
>>>>have crushed Hydra to...
>>>>What do you thisk about it? And was there an attempt to deploy an pc program
>>>>on asuper computer?
>>>
>>>Your link gives:
>>>
>>>Processor:
>>> PowerPC 440 700 MHz (2.8 GFlops)
>>>
>>>so  Junior,shredder,fritz  would perform quite normal.
>>>Please note that the word supercomputer is misleading. Supercomputers are
>>>special purpose computers. They favore floating point operations.
>>>
>>>Kind regards
>>>Bernhard
>>I dont think the word "normal" is the right word.
>>I mean,Can you Imagine one of this programs run on 65536 proccesors?!
>>And I belive its not so unrealistic. I know IBM willing to rent this
>>supercomputer for 10,000$ a weak.
>
>As I said Junior,shredder,fritz  would perform quite normal.
>IMHO they use about zero floating point operations.
>They are not good at many processors (>100).
>For diep you may read Vincent's comment in that thread.
>BTW comments like yours come up regularly. May be people like to dream.
>
>Kind regards
>Bernhard

There is a 12288 processor IBM machine with 700Mhz processors in the NOrth of
this country. Someone suggested i could get system time at it. That was someone
with big influence there.

These 700Mhz processors are ugly weak for chess. IBM processors always are.

Their cell processor is going to be even more ugly weak for chess. It has no
branch prediction and like 35 cycles penalty for every branch. Each of the 8
'cores' (processing elements) has only 256 KB ram to run the entire program
and will have to work at a block of data, so hashtable references are not
possible (well theoretical you can ask for hashtable but it will cripple the cpu
completely), as you have to ship blocks of data to each cpu and cannot ship just
16 bytes of hashtable information to it (diep and most chessprograms stores a
chess position in 16 bytes).

A very polite calculation of mine at the back of a paper suggested i would need
754 of these 700Mhz processors to get the same searchdepth worst case like at a
quad.

Of course i'd take the supercomputer blindfolded just for publicity reasons if i
could test at it without needing to deal with 8 burocratic organisations.

Vincent








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.