Author: Robert Hyatt
Date: 15:42:10 04/12/00
Go up one level in this thread
On April 12, 2000 at 10:45:14, Graham Laight wrote: >On April 12, 2000 at 10:24:40, Robert Hyatt wrote: > >>On April 12, 2000 at 06:13:40, Graham Laight wrote: >> >>>On April 11, 2000 at 22:13:32, Dann Corbit wrote: >>> >>>>On April 11, 2000 at 17:36:10, Jeremiah Penery wrote: >>>> >>>>>On April 11, 2000 at 06:04:48, Graham Laight wrote: >>>>> >>>>>>On SSDF, Crafty is rated as 2624 on a 450 Mhz PC. >>>>>> >>>>>>Generally, a doubling of processor speed results in an improvement of 60 elo. >>>>>> >>>>>>So, if Crafty were put on a 900 Mhz PC with 4 processors, it could achieve 3 >>>>>>doublings, or 180 elo improvement, to get a rating of 2804. >>>>> >>>>>Or you could just use a 16-CPU Alpha 21264 767(?) MHz machine for some real >>>>>speed. Just one of these processors ran faster than Bob's 4x400 Intel machine. >>>>>:) (Each is about the speed of an Athlon 1GHz.) >>>>> >>>>>I think they're making 32-processor machines even... >>>> >>>>Something even bigger on the burner. From >>>>http://www.digital.com/hpc/systems/sys_hpc320.html: >>>> >>>>"The HPC160 and HPC320 are part of Compaq's strategy to deliver HPTC systems >>>>that are based on collections of standard, volume SMP systems, linked together >>>>via System Area Network technology. While the HPC160 and HPC320 are targeted at >>>>ISV applications requiring up to 32 CPUs, there is a need for much larger >>>>systems, with hundreds or thousands of CPUs, for in-house applications >>>>previously supported by systems such as the Cray T3D and T3E. The ASCI >>>>(Accelerated Strategic Computing Initiative) program, for example, is demanding >>>>parallel systems into the teraflops and beyond. For those customers, Compaq is >>>>creating a family of packaged systems, based on a System Area Network licensed >>>>from Quadrics Supercomputers World, which will provide systems starting at 64 >>>>CPUs up to hundreds, or even thousands, of CPUs." >>>> >>>>A few thousand 21646's ought to run chess rather quickly. >>> >>>A couple of months ago, I saw Bruce Moreland complaining at CCC that IBM had >>>taken the credit for beating the world champion (the "Crown Jewells" of computer >>>chess), then disappeared into the night without allowing any further analysis of >>>their machine. >>> >>>The essence of IBM's claim to fame is that they built a supercomputer and proved >>>how good it was by using it to beat the world champion at chess. >>> >>>How ordinary this achievement would look if, just 3 or 4 years later, a similar >>>level of achievement was made by a computer which Joe Public could buy off the >>>shelf! >>> >>>-g >> >> >>It will eventually happen. I am 52 years old. I am not sure it will happen in >>my lifetime, although it might. I am sure it will happen within the lifetime of >>my kids... > >When are you planning to die? :-) > >From your message, you certainly give the impression that you're going to die at >a young age for a man of your demographic criteria! > >Please make sure that Crafty's "finished" before you go... > >-g I hope to live at least another 30 years. I don't think that by then programs will be able to dismiss the kasparov's of the world as weaklings. IE by my measure, since the first serious program in 1965 rated at about 1500, after 35 years we are getting near to 2500, with much of that based on hardware. How long to get to 3000? It (to me anyway) gets harder and harder, the better my program gets. It is easy to go from (now) 1500 to 2200 in a month. 2200-2400 in a year. 2400-2500 in 3-4 more years... etc... the curve flattens...
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.