Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: HW based Crafty (Boule's thesis)

Author: Robert Hyatt

Date: 21:36:03 04/04/02

Go up one level in this thread


On April 04, 2002 at 22:53:53, Tom Kerrigan wrote:

>On April 04, 2002 at 13:25:59, Robert Hyatt wrote:
>
>>>On April 03, 2002 at 17:49:58, Robert Hyatt wrote:
>>>>ASICS used in the original DB hardware...  3 micron ASICS in DB only ran at
>>>>24mhz.  An FPGA can certainly be run at that clock speed...
>>
>>I didn't write very clearly.  Chiptest and deep thought (the original deep
>>blue hardware) was, I believe, 3 micron.  DB1 and DB2 were complete re-designs
>>using smaller die sizes.  This is mentioned in his book, when/if it comes
>>out...
>
>It's not that what you wrote wasn't clear, it was simply wrong. "3 micron ASICS
>in DB..."

It wasn't really wrong.  At one point, there _was_ a thing called "deep blue
prototype" running on 3 micron ASICs...  This was the thing Fritz beat in 1995
in Hong Kong, in fact...

And it is pretty easy to interpret "early DB hardware" as either deep thought
hardware or DB1 hardware...  Since the whole "line" was finally known as "deep
blue"...



>
>>>be slower. But if you take the optimistic 50MHz number and the very overly
>>>optimistic 2-cycle-per-node number, you get 25M NPS, not "30M NPS easily."
>>All I really know about the DT chip was that it took 10 cycles per node...
>>although some cycles were skipped (fast vs slow eval for example)...  I based
>>the 30M nps value on statements made by Hsu...
>
>Statements made by Hsu about ASICs, not FPGAs. I cannot stress enough how
>completely different the two are.
>
>IIRC, DB could do a fast eval of a node, which took 2 cycles, and a slow eval,
>which took 8 cycles.

Something similar, although the "fast eval" might have been incremental..  and
took no time...



>
>It occurs to me that the DB chips were running at ~25MHz with the same feature
>size as the 200MHz PPro. With CPUs running at ~100MHz in FPGAs, you might expect
>FPGA chess logic to be half as fast as the DB ASICs, i.e., ~1M NPS. So while
>FPGA chess is a very interesting project, I'm not sure I'd expect it to be any
>faster than chess software.
>
>-Tom


I think it would _clearly_ be faster than software...  IE DB took ten clocks
per _node_.  What software program comes anywhere near that.  More like
3000 instructions per node, which turns into 4-5000 clocks most likely.




This page took 0.02 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.