Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF There is something wrong about doubling the speed ?

Author: Uri Blass

Date: 03:29:55 12/06/02

Go up one level in this thread


On December 06, 2002 at 06:10:01, Jorge Pichard wrote:

>On December 06, 2002 at 04:57:01, Uri Blass wrote:
>
>>On December 06, 2002 at 04:44:05, Jorge Pichard wrote:
>>
>>>On December 06, 2002 at 04:18:38, Uri Blass wrote:
>>>
>>>>On December 06, 2002 at 03:47:31, Jorge Pichard wrote:
>>>>
>>>>>On December 05, 2002 at 16:52:06, Roy Eassa wrote:
>>>>>
>>>>>>1200 MHz Athlon results only:
>>>>>
>>>>>I read several times that doubling the speed account for at least 50 Rating
>>>>>points, but if you consider Chess Tiger 15 performance on a mere K6 450 Mhz
>>>>>which is less than half the speed of an AMD Athlon 1.2 Ghz, the difference is
>>>>>less than that.
>>>>
>>>>I think that you miss the statistical error and it is also not clear that
>>>>you get always at least 50 elo from doubling the speed when the hardware get
>>>>faster.
>>>>
>>>>Diminishing returns is possible.
>>>>
>>>>Uri
>>>
>>>That is true, but it would be interesting to find out what is the breaking point
>>>where the diminishing returns take effect, there has to be a breaking point in
>>>hardware's speed where the difference in speed no longer is important.
>>>
>>>Pichard.
>>
>>I do not believe in one breaking point.
>>
>>I guess that this is a process and I also believe that it is a slow process.
>>The difference may get down from 50 to 45 and from 45 to 40 but you are not
>>going to see close to 0 returns in the near future.
>>
>
>The difference has already decrease from 50 to 45 and even lower than 40 if you
>take into consideration Chess Tiger 15 from K6-450 Mhz to 1.2 Ghz.

I think that there is a bigger statical error and I expect the difference to be
more than 50 even for tiger in the next list

The difference for Deep Fritz7 is 110 elo and the right gain is probably in the
middle.

(110+52)/2=81 is an estimate for the gain from the new hardware(52 is for tiger)

It is possible to calculate better estimate by using more programs but I am too
lazy to do it at this moment.

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.