Author: Uri Blass
Date: 10:37:49 11/04/00
Go up one level in this thread
On November 04, 2000 at 12:30:41, Christopher A. Morgan wrote: > > According to the latest SSDF rating list, 10/27/00, the number 16 and 17 >programs had an average rating of 2475 (16. Fritz 5.32 64MB P200 MMX 2476, >and 17. Hiarcs 7.32 64MB P200 MMX 2473). When we go up a generation in >microprocessor, double the size of the hash table, double + the speed, then the >same two programs now have an average rating of 2541, an increase of 66 points >(5. Fritz 5.32 128MB K6-2 450 MHz 2552 , and 9. Hiarcs 7.32 128MB K6-2 450 >MHz 2530). > > The number 1 and 2 programs had an average rating of 2611 (1. Fritz 6.0 128MB >K6-2 450 MHz 2630, and 2. Junior 6.0 128MB K6-2 450 MHz 2592) > > If we once again go up a generation in microprocessor to Pent III, or AMD >Athlon, double the size of the hash table to 256MB, and double + the speed to >1,000 MHz, should we expect an increase in average rating or around 66 points to >2677 or thereabouts? My guess is that there is a diminishing return and that you are going to see smaller improvement from faster hardware. I do not think that there is a big difference and you can see 60 elo improvement instead of 70 elo improvement. > > My more basic question is have we reached the point in computer chess >programming where the programs can play at the super elite level, 2700+, if they >can run fast enough or have multi processors re Deep Junior with 4 processors >which appears to now play at the super elite level. I think that you cannot learn the answer to this question based on the ssdf results because I do not think that the difference in comp-comp games is the same as the difference in comp-humans games. Another problem is that public programs are at disadvantage against humans because humans can prepare and learn killer lines against them. Can more be done with >programming? Obviously, re Deep Blue we can use brute force and play at or near >world champion level. I am not asking about that level of hardware. Programs always get better and there is no doubt that it is possible to improve programs significantly. My guess is that we are going to see more than one program with ssdf rating of more than 70 elo better than Fritz6a in 2001. Uri
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.