Author: Keith Evans
Date: 18:46:22 07/11/02
Go up one level in this thread
On July 11, 2002 at 18:04:21, stuart taylor wrote: >On July 10, 2002 at 13:51:22, Keith Evans wrote: > >>On July 10, 2002 at 10:00:09, stuart taylor wrote: >> >>>Obviously that is what would be nice to know, and I assume that some people here >>>have some idea, which I don't yet. >>> Would it be atleast the equivelent of 10 Ghz. of a proccessor? >>>If not more, then it would be the same as might anyway be in about 3-4 years >>>from now. >>>I would hope it was atleast that, in which case a nimzo program WOULD just about >>>trounce anything today for a single proccessor, though maybe, not >>>overwhelmingly. But enough for me (a poor person) to what to spend 3-400 dollars >>>for one! >>> But then again, wouldn't others compete with even better ones (hardware), and >>>even with better software also? >>>S.Taylor >> >>If you're really skilled then you might get the power of one Deep Blue chip out >>of a $1000+ FPGA based board. This is basically a SWAG based on how well the >>move generator would fit into an FPGA. My conclusion is that you would be better >>served by spending $500 more to get an SMP system than $1000+ to get some >>special purpose hardware. >> >>If you did buy such a board and this became popular, then there would probably >>be guys buying $10,000 boards for competition which would either run faster, or >>have more chess coprocessors much like Deep Blue. >> >>I have heard rumors from some friends about some reasonably priced >>reconfigurable computing boards loaded with high-end FPGAs under development, >>but until I see real pricing I'm a bit skeptical. If there turns out to be a >>"killer app" for this type of hardware, then this could definitely help to >>reduce the price. >> >>Regards, >>Keith > >By "one Deep Blue chip" you mean everything that Deeper Blue had? >That sounds quite great, if so. That means that Deep(er) blue that beat Kasparov >will have dropped in price to about $1000! >S.Taylor Oh no - I didn't mean to imply that. Deep Blue I had 216 chess chips, and I believe that Deep Blue II had about twice that many. My belief based on looking at the implementation of a move generator and thinking a little bit about evaluation is that it will take an expensive FPGA to fit the equivalent of what _one_ of the Deep Blue chess chips had. You might be able to make it more effective by supporting a small hash table or null move, but I don't know how much more effective it would be than Hsu's chip. The problem with FPGAs is the interconnect - there is a ton of interconnect needed in a chess chip. This is all a guess of course. Brutus did well in its most recent competition, but I don't know what he's doing in hardware and how much better its going to get. If you search for information on the board that he's using it's quite expensive. Regards, Keith
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.