Author: Robert Hyatt
Date: 08:26:08 06/13/99
Go up one level in this thread
On June 13, 1999 at 05:27:59, Micheal Cummings wrote: > >On June 13, 1999 at 00:54:42, Robert Hyatt wrote: > > >>>Read College and University papers, and see how many of these types of comments >>>are made and written on various subjects, and see how many actually come into >>>reality. >>> >> >>I have no idea what you mean... He has built the machine... So all you >>are hung up on is whether he can deliver it to the end-user or not? Why >>not just wait rather than making statements that are insulting? Either he >>will sell them or he won't. I suspect the former. He generally doesn't >>put things into writing and then walk away from them.. >> > >So you are tellimg me he has already built the chip, and it is just waiting for >someone to make and market it. Because if you are talking about the DB machine >that won against Kasparov. That is a long way from a small chip to put into a >computer. You are wrong. The chip _was_ used in the DB machine. 480 of them if I recall the IEEE micro article correctly. _that_ is the chip he is talking about. He said that refabbing it on a newer process (.18 micron I think) could speed it up by a factor of 10 or more. So yes, the 'chip' is the same as the deep blue chip design... No one says he is going to sell an IBM SP machine... rather that he will take this chip, build a PCI interface for it, and go... > >And just wondering, how do you know if he has made this chip already. Or are you >just trusting what he said. I watched this 'chip' beat Kasparov... > >I want to see some hard data on this chip, I have seen many invention fall by >the way side. And I am not going out to insult anyone. I am just wondering why >so many people want to jump on the band wagon of this. Send me some hard data, >not just words from Hsu and I will eat my words. > Hard data: Beat kasparov in 1997 using 480 of thse things. Current version of the chip searches at 2-2.4M positions per second per chip. Using an old fab process. Commercial version (.18 micron I think) will search at an estimated 36M nodes per second per chip. He could possible (for high-end users) put maybe 4-8 of these chips on a PCI card. Which would be about as fast as the old deep blue hardware. Whether the PCI bus could sustain the bandwidth to handle multiple cards is unknown to me... but with search tuning it could possible work... >I am just not as gullable as some, especially in computer related area, so many >promises, so many time people fail to live up to their statements Would you care to give an example? IE HSU promised a machine that could beat Kasparov. He delivered. I promised a parallel search. I delivered. Eugene promised uncompress-on-the-fly endgame databases with better compression than anything currently existing. He delivered. Etc....
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.