Author: Albert Silver
Date: 07:48:23 07/29/00
Go up one level in this thread
On July 28, 2000 at 20:40:00, Chris Carson wrote:
>On July 28, 2000 at 18:08:36, Dann Corbit wrote:
>
>>On July 28, 2000 at 16:06:40, blass uri wrote:
>>
>>>On July 28, 2000 at 15:26:43, Enrique Irazoqui wrote:
>>>[snip]
>>>>And this way is not feasible, so it's not effective either. Besides, it would
>>>>give at best a new global calibration, but wouldn't tell us in the future if
>>>>ratings based on comp-comp are also valid for human-comp. Are they? The
>>>>anti-computer games of Frankfurt and Dortmund wouldn't have been helped by a
>>>>faster search
>>>
>>>I disagree.
>>>
>>>Junior could find Bh6 against piket if it has more time and in this case piket
>>>could not get the position that junior does not understand(in this case less
>>>time could also help Junior but I think that seeing deeper is more productive
>>>than counter productive).
>>>
>>>More time could probably also help Junior against kramnik because a few plies
>>>after Kh8 it could see a drop in the evaluation and these few plies include
>>>captures that are not quiet moves.
>>
>>But according to Amir, longer times sometimes hurts also, since his laptop got
>>the right answer while the 8 CPU behemoth got the wrong one on a few occasions.
>>
>>On the other hand, I rather suspect that the extra horsepower helps a lot more
>>often than it hurts.
>>;-)
>>
>>On the other, other hand, since chess is an exponential process, it would take a
>>lot more CPU power to make a significant difference in playing strength, and (in
>>fact) such a machine does not exist. For the architecture of the Deep Junior
>>program, there is no faster machine on the planet than the one that was used.
>
>Some good points Dann. :)
>
>However, I think that the speed will double very quickly.
Perhaps, but not as fast as we may hope. Interestingly enough, Intel announced
they expected there to be 10GHz chips by 2011
(http://www.zdnet.com/zdnn/stories/news/0,4586,2601717,00.html), which is a
significant slowdown from Moore's Law. Looks like it will be taking us more time
than expected to reach DB's speeds.
Albert Silver
>Faster chips already exist and soon both the faster chips and more
>processors on a server will be here. Competition and demand
>will drive that in my opinion. So the speed issue will solve itself
>(or already has) very soon and speed doubling is an exponential function.
>
>The bigger problem I see (I have switched to this belief over the
>last few months) is that of improving the program itself, however,
>I think the top programmers have a handle on it and if not, then the
>door is wide open for new ideas. The programs are very adaptable and
>will only get better. I think we are seeing GM level results now on P-200
>and faster with wide variability on the slower machines. My opinion is that
>as the programs get better and the hw gets faster, we will see higher results
>and less variability (with the exception of operator error as in the
>ISR league, or hw failure). Just my two cents. You do not have to believe me.
>:)
>
>Sorry, that sounds like a speach and I did not mean it to be. :)
>
>Best Regards,
>Chris Carson
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.