Author: Robert Hyatt
Date: 18:38:40 07/20/00
Go up one level in this thread
On July 20, 2000 at 15:19:46, Chris Carson wrote: >Dead wrong! > >You do not list the results of DT against any programs/hardware, >I agree with that. > >1988 to 1994 DT dominated programs on 386 and 486 hardware. With >me so far, not P90. Perhaps 90% against 386/486, but not P90. > >In 1995 DT played a few games against P90, lost to Fritz and drew >with Wchess. > >Your argument is that DT beat everything 90%, not true for P90 >systems. > >Your argument is that DB is 100 times faster than DT thus the >90% victory margin holds because machines today are not 100 >times faster than the machines DT beat. Not true. > >Again, the 90% score was against 386 and 486 machines and 8x700 is >400 to 900 times as fast. Thats a pretty big jump from the machines >DT beat from 1988 to 1994. The 1995 hw that fritz ran on is a P90, >DT lost and drew against those systems. That is not 90% and 8x700 >is about 100 times as fast. > >Got it. No I didn't think you did. > >I'll bet you have to have the last word, even if it makes you >look silly. :) > >I may post more on this, or I may not. Ignoring any futher >posts is not an admission that Bob has won, just that I am >tired of arguing with a fence post on this subject. :) > >Best Regards, >Chris Carson You overlook the main detail: DT got better every year. It was not a static one-off piece of hardware. They redid the chips. They added more chips. Etc. So yes, it was light-years ahead of the 486 machines. And each year as new PC cpus came out, newer and faster and better versions of DT came out as well. Right on thru 1997. They ended up peaking at 1B nodes per second. With a sustained 200M nodes per second (this included factoring in 1/3 efficiency of the alpha/beta search, while the rest of us quote raw NPS values for our parallel searches). They got better each year. The gap _never_ got narrower. At least until post- 1997. We are now catching up. We only have to get from the 2M of Djr to 1B of DB2. That's only a factor of 500 left to close the gap, speed-wise, to zero (ignores their eval hardware but so what at this difference). Ought to take only a couple of years by your reckoning for the micros to catch up? My log2() function gives a different answer, personally...
This page took 0.01 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.