Author: Christophe Theron
Date: 21:44:43 11/06/00
Go up one level in this thread
On November 07, 2000 at 00:22:43, Uri Blass wrote:
>On November 06, 2000 at 21:41:03, Christophe Theron wrote:
>
>>On November 06, 2000 at 16:49:29, Uri Blass wrote:
>>
>>>On November 06, 2000 at 16:43:45, Uri Blass wrote:
>>>
>>>>On November 06, 2000 at 16:21:54, Christophe Theron wrote:
>>>>
>>>>>On November 06, 2000 at 16:13:43, Uri Blass wrote:
>>>>>
>>>>>>On November 06, 2000 at 15:17:32, Jonathan Lee wrote:
>>>>>>
>>>>>>>On November 06, 2000 at 01:43:48, Christophe Theron wrote:
>>>>>>>
>>>>>>>>On November 05, 2000 at 15:33:57, Jonathan Lee wrote:
>>>>>>>>
>>>>>>>>>I understand Walter Irvin's point of view of contemporary greats versus
>>>>>>>>>the others point of view who are talking about influential greats.
>>>>>>>>>It would be safe to say that the 20th century influential greats (such as
>>>>>>>>>Richard Lang) are a SUBSET of the year 2000 contemporary greats listed _some_ by
>>>>>>>>>Walter Irvin.
>>>>>>>>>Grandmasters in gigahertz: Just tell me how many gigahertz will it take to
>>>>>>>>>equal Kramnik.
>>>>>>>>>Jonathan (83rd message)
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>500MHz will be enough. Not with current programs, but in several years software
>>>>>>>>improvements will compensate.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> Christophe
>>>>>>>500 GHZ
>>>>>>>Jonathan (86th message)
>>>>>>
>>>>>>I am not christophe but I am almost sure that he means 500 mhz and not 500GHZ.
>>>>>>I suspect that 500GHZ may be enough with the best programs of today.
>>>>>>
>>>>>>Uri
>>>>>
>>>>>
>>>>>I meant 500MHz.
>>>>>
>>>>>500GHz is enough with programs of ten years ago.
>>>>
>>>>I disgree here.
>>>>
>>>>I do not think that 500GHz with programs of 1990 is enough.
>>>>
>>>>I believe that 500GHz with programs of 1990 are going to lose at tournament time
>>>>control against the top programs of today on 1 GHz.
>>>>
>>>>Uri
>>>
>>>I can add that it is possible to check it by a correspondence match when a
>>>program of 1990 will have 25 hours per move and a program of 2000 will have 3
>>>minutes per move when both programs use 1 GHz processors.
>>>
>>>Uri
>>
>>
>>
>>I'm ready to bet.
>>
>>
>>
>> Christophe
>
>Remember that we talk about the best program of 1990.
>I believe that these programs do not do extensions like the programs of today.
Programs of 1990 include the Mephisto line on 680xx. In 1990, it was the
Mephisto Lyon. I guess it was running on a 68030.
>I know that even the best program of 1994(Genius3) does not do more than 12
>plies of extensions when the time control is not relevant.
>
>I expect that the top programs of today(3 minutes per move) may sometimes
>outsearch the top programs of 1990(25 hours per move) because of this reason.
>Here is an example:
>
>[D]r4rk1/2p3pp/p7/1p1pq3/8/2P2N2/PPQ2KPP/R1B5 w - - 0 1
>
>Avoid Kg1
>Part of the top programs but not all of them may avoid Kg1 in less than 3
>minutes if they have 1gh machine.
>
>I remember that I gave Ganius3 48 hours on p100 and it could not avoid Kg1
>inspite of predicting the correct line.
>25 hours on p1000 is more than 48 hours on p100 but Genius3 is also better than
>the top program of 1990 so it is possible that the top program of 1990 is going
>to blunder here by Kg1 even after 25 hours.
It's just an exception. Generally, 12 plies of extensions is more than enough.
>I do not say that the top programs of today will always outsearch the top
>programs of 1990 with more time but if you remember the fact that the top
>program of today has better positional understanding then I expect the top
>programs of today to win a match at this condition.
In my opinion, the improvements in search and positional understanding will not
compensate for the overwhelming difference in speed.
500GHz is 1000 times 500MHz. If doubling the speed accounts for 70 additional
elo points (as it is with current computers), that makes a 698 elo points
difference.
If doubling speed accounts for only 30 elo (to take into account an hypothetical
"dimishing returns"), that still makes 299 elo points.
I don't believe there is 300 elo points of difference between current best
programs and the best program of 1990 if they are run on equal hardware.
Branching factor at very slow time controls can also be an issue, but I don't
believe 1990 programs had such a terrible branching factor.
Christophe
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.