Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: to bob re:hsu's chip

Author: Robert Hyatt

Date: 07:08:31 04/28/01

Go up one level in this thread


On April 27, 2001 at 23:49:24, Robert Hyatt wrote:

>On April 27, 2001 at 11:52:26, Joshua Lee wrote:
>
>>I think i am missing something important or we all are? those chips were
>>supposed to be 200Mhz PowerPC's in Deep Blue wouldn't they use the same but
>>faster in DB2?
>
>
>Wrong chips.  DB played on an IBM SP2 which used 32 PPCs as the base
>cpus.  Connected to each PPC was a pair of VME bus cards with 16 of Hsu's
>custom chess chips on each board.  _those_ are the chips we are talking
>about when we say 20mhz and 24mhz..
>


That is wrong.  Each PPC processor communicated with only 16 chess procesors,
not 32.  Not sure where I got 32 other than it was late and I was tired...




>
>> For a 20 or 24Mhz chip at .60 Microns to perform a millions of
>>nodes per second you would have to have a RISC design
>
>
>
>Wrong idea.  The chips are not general purpose CPUS (the chess chips).  They
>are ASICs that were designed to specifically implement the chess functions
>needed (included a hardware alpha/beta search).
>
>
>
>> and be at .10 then you
>>would need more than i can figure as far as i can tell at 20Mhz if it were like
>>a regular chip in a home computer it would be running at 10,000 Nodes per second
>>(on my 500Mhz P3 i get tops 400Knps i figure on a 100Mhz you would get 50Knps
>>and on 50Mhz you would get 40Knps and on a 20Mhz you would get 10Knps)but that's
>>just it it's not a regular chip it Only runs a chess program  see what i am
>>saying that is just too big so without saying much further we need a real
>>comparison actual results with the "Chess Specific chip "
>>We can't compare a Quad Xeon which does millions of calculations to keep widows
>>from crashing, the taskbar running, time of day, Kernel32.dll, yadda yadda.
>>The processors we have in our computers are not even 50% efficient probably not
>>even that when it comes to something that is designed to do 1 thing and do it
>>really really well.  So probably if our computers were doing one thing at their
>>.18 or even .60microns they would be hitting millions of nodes per second if not
>>1 billion.  But this is just my guess as i am not an engineer.



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.