comparision of time control

[John Grant]

Geschrieben von: / Posted by: John Grant at 16 June 2000 20:58:12:
Would moves reached at the same time control be better on a much much faster machine (with more memory for calculations) than on a 10-times-slower machine in terms of cpu speed (there is no correlation between search depth and cpu speed IMO)?
For example if I wish to study a certain variation of the Benko Gambit I would save time on a faster machine, and would know more about the performance of WinBoard programs. I think allocation of extra memory when the game begins between 2 WinBoard programs is not necessarily greater when longer time control is used on the same cpu speed. For example 40 MB is sufficient at 2 different time controls.
The game I was interested in was from Toronto, 1971.
White: Z.Vranesic ; Black: P.Benko
1.P-Q4;N-KB3 2.P-QB4;P-B4 3.P-Q5;P-QN4 4.P*P;P-QR3 5.P*P;B*P 6.N-QB3;P-Q3
7.N-B3;P-N3 8.N-Q2;B-KN2 9.P-K4;O-O 10.B*B;N*B 11.O-O;N-Q2 12.N-B4;N-N3
13.N-K3!;Q-Q2...0-1.
I do not know whether you thought that this is a good way to compare programs with the same opening.

[pete]

Geschrieben von: / Posted by: pete at 16 June 2000 22:08:58:
Als Antwort auf: / As an answer to: comparision of time control geschrieben von: / posted by: John Grant at 16 June 2000 20:58:12:

Would moves reached at the same time control be better on a much much faster machine (with more memory for calculations) than on a 10-times-slower machine in terms of cpu speed (there is no correlation between search depth and cpu speed IMO)?
For example if I wish to study a certain variation of the Benko Gambit I would save time on a faster machine, and would know more about the performance of WinBoard programs. I think allocation of extra memory when the game begins between 2 WinBoard programs is not necessarily greater when longer time control is used on the same cpu speed. For example 40 MB is sufficient at 2 different time controls.
The game I was interested in was from Toronto, 1971.
White: Z.Vranesic ; Black: P.Benko
1.P-Q4;N-KB3 2.P-QB4;P-B4 3.P-Q5;P-QN4 4.P*P;P-QR3 5.P*P;B*P 6.N-QB3;P-Q3
7.N-B3;P-N3 8.N-Q2;B-KN2 9.P-K4;O-O 10.B*B;N*B 11.O-O;N-Q2 12.N-B4;N-N3
13.N-K3!;Q-Q2...0-1.
I do not know whether you thought that this is a good way to compare programs with the same opening.

I rarely engage in those technical discussions to avoid saying too silly things , but this time I am just too certain
Forget about memory .
CPU speed is what counts ( "and that's all" to say it in the Tina Turner way ) .
pete

[Alexandre Côme]

Geschrieben von: / Posted by: Alexandre Côme at 17 June 2000 14:12:46:
Als Antwort auf: / As an answer to: comparision of time control geschrieben von: / posted by: John Grant at 16 June 2000 20:58:12:

Would moves reached at the same time control be better on a much much faster machine (with more memory for calculations) than on a 10-times-slower machine in terms of cpu speed (there is no correlation between search depth and cpu speed IMO)?
For example if I wish to study a certain variation of the Benko Gambit I would save time on a faster machine, and would know more about the performance of WinBoard programs. I think allocation of extra memory when the game begins between 2 WinBoard programs is not necessarily greater when longer time control is used on the same cpu speed. For example 40 MB is sufficient at 2 different time controls.
The game I was interested in was from Toronto, 1971.
White: Z.Vranesic ; Black: P.Benko
1.P-Q4;N-KB3 2.P-QB4;P-B4 3.P-Q5;P-QN4 4.P*P;P-QR3 5.P*P;B*P 6.N-QB3;P-Q3
7.N-B3;P-N3 8.N-Q2;B-KN2 9.P-K4;O-O 10.B*B;N*B 11.O-O;N-Q2 12.N-B4;N-N3
13.N-K3!;Q-Q2...0-1.
I do not know whether you thought that this is a good way to compare programs with the same opening.

What you say is very stupid !!!!
Of course there is direct correlation between depth and cpu speed !!!
Best Regards
Alexandre Côme

[John Grant]

Geschrieben von: / Posted by: John Grant at 17 June 2000 16:28:41:
Als Antwort auf: / As an answer to: Re: comparision of time control geschrieben von: / posted by: pete at 16 June 2000 22:08:58:
The following text is from e-mail discussion between myself and Volker Pittlik:
Would 900MHz P-III half the(average thinking time per move)time control? from 40 moves in 90 minutes on 450 MHz P-III to 40 moves in 45 minutes??
I dont believe this. There is no linear correlation of search depth and cpu speed IMO. To reduce the time to the half you have to use a computer running at 4.5 GHz I guess (compared to a 450 MHz system). But your question is interesting.

[pete]

Geschrieben von: / Posted by: pete at 17 June 2000 17:39:41:
Als Antwort auf: / As an answer to: Re: comparision of time control geschrieben von: / posted by: John Grant at 17 June 2000 16:28:41:

The following text is from e-mail discussion between myself and Volker Pittlik:
Would 900MHz P-III half the(average thinking time per move)time control? from 40 moves in 90 minutes on 450 MHz P-III to 40 moves in 45 minutes??
I dont believe this. There is no linear correlation of search depth and cpu speed IMO. To reduce the time to the half you have to use a computer running at 4.5 GHz I guess (compared to a 450 MHz system). But your question is interesting.

Aha , now I understand you much better . In your post you stated "there is no correlation ... " which is something different .
In fact I think the statement of Volker didn't answer the question you initially asked ( if I understand you correct ) .
Volker pointed out that searching twice as long doesn't mean getting twice as deep .
But if you have a machine exactly two times faster than another ; it should mean : yes , you can use 40 moves in 45 minutes instead of 40 moves in 90 minutes to reach exactly the same strength/depth than on the slow one ( pondering or special settings in the timing code aside ) .
So :
a.) is this what you wanted to discuss ?
b.) is there a logical mistake in my reasoning ( in fact I am not sure I understood the above statements 100% when it is about language ) ?
greetings.
pete

[CDC]

Geschrieben von: / Posted by: CDC at 17 June 2000 23:04:01:
Als Antwort auf: / As an answer to: Re: comparision of time control geschrieben von: / posted by: pete at 17 June 2000 17:39:41:

The following text is from e-mail discussion between myself and Volker Pittlik:
Would 900MHz P-III half the(average thinking time per move)time control? from 40 moves in 90 minutes on 450 MHz P-III to 40 moves in 45 minutes??
I dont believe this. There is no linear correlation of search depth and cpu speed IMO. To reduce the time to the half you have to use a computer running at 4.5 GHz I guess (compared to a 450 MHz system). But your question is interesting.

Aha , now I understand you much better . In your post you stated "there is no correlation ... " which is something different .
In fact I think the statement of Volker didn't answer the question you initially asked ( if I understand you correct ) .
Volker pointed out that searching twice as long doesn't mean getting twice as deep .
But if you have a machine exactly two times faster than another ; it should mean : yes , you can use 40 moves in 45 minutes instead of 40 moves in 90 minutes to reach exactly the same strength/depth than on the slow one ( pondering or special settings in the timing code aside ) .
So :
a.) is this what you wanted to discuss ?
b.) is there a logical mistake in my reasoning ( in fact I am not sure I understood the above statements 100% when it is about language ) ?
greetings.
pete

Yes, I agree. A machine twice faster than another (same cpu with a doubled frequency) computes almost exactly two times faster. If there is also an upgrade in the cpu technology, moreover, the rate is even higher. For example a pentium II 400MHz is more than twice faster than a pentium 200MHz MMX, i.e. is 2*a times faster, with a>1. I guess there exist some tables with those coefficients for many existing cpus.
There's one thing that's not doubled though: the time of access to the hard disk in the opening phase, being driven by the bus, is not likely double, but that thing influences only the blitz games, when the time spent for the opening is comparable with the game time.
Greetings
CDC