Author: Robert Pawlak
Date: 12:38:55 02/15/04
Go up one level in this thread
Bob, Isn't multi-hydra redundant ;-) ? Sure, a multihydra is easily done. Get a passive backplane system, and write the code :-). Personally speaking, I would like to see this approach succeed (I am speaking of FPGAs in general, not just the multihydra). However, I think there are various commercial barriers to something like this actually becoming successful. My feeling is that this might be able to exist as a high-end niche product, and be successful for Chessbase as a marketing tool. So they won't be able to make big bucks on the product, but it might be a really neat marquee. Put the thing in an external USB enclosure, with a bunch o' blinkin LEDs on it, and match some GM. Disclaimer: I've never written a chess engine in my life. I believe that people dismissing the FPGA approach out of hand are ignoring some differences between Hydra and DB. First off, deep blue was constructed when the hardware was relatively expensive. Now it is much cheaper. And today, there are a number of really good development tools for VHDL. Furthermore, changing logic is equivalent to changing code, and can be done nearly instantaneously with VHDL. Not so with an ASIC, which requires a design change. This means that hydra could have continuously updated firmware upgrades, downloadable over the internet. Not so with DB. The military is going into FPGAs in a really big way, since they provide a number of supportability benefits. This in turn, will drive costs down even further. Is it the wave of the future? Who knows?
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.