Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / engine strength-processor (RYBKA 3.2)
- - By nalleg Date 2008-01-14 15:29
I´ve noticed that  3.2 on  a quad core is about 10 times faster than my old P4 3 GHZ ( measured in KN/second). Does that mean that the odds would be eqaulised  if the older configuration would get 10 times more time ?. I suppose  that is not the case  but can someone explain why?

Regards,
Björn G
Parent - - By Vempele (Silver) Date 2008-01-14 15:39
If neither is allowed to ponder, it's true.
Parent - - By Carl Bicknell (*****) Date 2008-01-14 17:11
yes make sure ponder is off!

I'm REALLY surprised it's as much as 10x speed increase though - really?
Parent - - By Vempele (Silver) Date 2008-01-14 18:27
A single core of a Core 2 processor in 32-bit mode is more than twice as fast as a Pentium 4 running at the same clock speed. I suppose it's a bit hard to grasp just how badly the P4 sucks.
Parent - - By Sesse (****) Date 2008-01-14 18:40
- 3.2GHz vs. 3.0GHz: factor 1.067
- Core 2 vs. Pentium 4: factor 2.0
- Four cores vs. one: factor 2.5 (?)
- 64-bit vs. 32-bit: factor 1.7

End result: about 9.0

So, it isn't all that unrealistic.

/* Steinar */
Parent - - By Vempele (Silver) Date 2008-01-14 20:11

>- Core 2 vs. Pentium 4: factor 2.0


Slightly more than that.

>- Four cores vs. one: factor 2.5 (?)


3.1

>- 64-bit vs. 32-bit: factor 1.7


1.6 IIRC.
Parent - - By Vinvin (***) Date 2008-01-16 09:18
>- Core 2 vs. Pentium 4: factor 2.0

Are you really sure ???

I never thought  it's so big ...
Parent - By Vempele (Silver) Date 2008-01-16 10:24
I measured Pentium M 1.5 GHz vs. Pentium 4 3.0 GHz a while ago. The latter came out a tiny bit slower. It's even worse for some engines (from an exchange on the Winboard forum):

bob (Hyatt):
Nah. Joker runs like sh** on a PIV because the PIV itself is a piece of sh**. :-)My core2 at 2.0ghz is twice as fast (single cpu) as my PIV xeon 2.8ghz. Actually even more factoring in the 64 bit operations. PIV was just a dog from the get-go, until they had time to get the core-2 right...

H.G.Muller:
Well, I won't argue with your classification of the Pentium 4.

But it must be more than that, as I am talking of an equal-hardware situation, where the opponents are also having the handicap of a Pentium 4. And on that machine Joker could not compete with engines that perform ~300 ELO below it on my own Core 2 Duo.
Parent - - By ernest (****) Date 2008-01-18 14:30

>> - Four cores vs. one: factor 2.5 (?)
> 3.1
>


Maybe Zappa, certainly not the current Rybka!
Parent - By Vempele (Silver) Date 2008-01-18 14:42 Edited 2008-01-18 14:51

> Maybe Zappa, certainly not the current Rybka!


For 2.0 beta 8, it was 2.8. Zappa seems to be about 3.5 on the CCRL list, which is consistent with the 3.7 achieved by Cray Blitz.
Parent - - By Carl Bicknell (*****) Date 2008-01-14 20:20
I think it would be great if a list of processors could have their efficentcy noted along side them. For example we could have core 2 with a rating of 1 and, therefore pentium 4 with a rating of 0.5 (ish)

Multiply that by the processor speed (MHz) and the number of cores and we'd have a definative speed rating. Vempele?
Parent - By Vempele (Silver) Date 2008-01-14 20:30

> Multiply that by the processor speed (MHz) and the number of cores and we'd have a definative speed rating.


Remember to factor in scaling and 64-bit. Probably (modest) overclocking, too. And a bunch of other factors that don't matter as much.
Parent - By BB (****) Date 2008-01-14 21:11

> I suppose it's a bit hard to grasp just how badly the P4 sucks.


Indeed, with one sitting on my desk (a 2-processor machine, but the bridge sucks most of the doubling away in any case), I feel this regularly. Incidentally, do you have an opinion as to whether the POPCNT instruction with the Phenom [or Nehalem/SSE4.2] should show a measurable gain for bitboard engines?
Parent - By albitex (***) Date 2008-01-17 18:51
I had a Celeron 2,6 Ghz. Then I are passed to a Pentium IV 3Ghz. Today I have a
dual core 2,2 Ghz.  I know therefore well the performances of Rybka on these CPU.
Here they are:
               Celeron 2,2 Ghz:  Rybka 32 bit 17/30 Kn/s
               Pentium 4 3Ghz HT: Rybka 32 bit 40/60 Kn/s, Rybka 64 bit 80/110 Kn/s
               Core2 Duo E4500 2,2 Ghz: Rybka MP 64 bit 220/260 Kn/s

notes: - the Pentium 4 is the model 631 with 64 bit support
          - speed is that measured after the opening and in the middle game. Obviously in the ends more elevated values will be had.
          - values with CPU not overcloking
Parent - - By FirebrandX (**) Date 2008-01-16 12:47
10x faster sounds about right. My old pentium 4 gets about 50 kN/s whereas the quad cores are getting somewhere around 500 or 600 kN/s from what I've heard.

I like to think of it in time terms, meaning my crappy P4 would need ten hours to calculate what a Quadcore does in just one hour. That's why I can't wait to build my new system next month ($2k budget). I've used my P4 for 4 years now, so this is going to be a huge upgrade for me in all areas of computing.
Parent - - By Carl Bicknell (*****) Date 2008-01-16 16:39
no, this is not what a co-efficent of efficentcy is.

It needs to be based on ONE core at the SAME clock speed. After that you can factor in cores and clock speed differences and 64bit etc. The highest it could possibly be is 3 although 2 sounds about right, and even that surprises me
Parent - - By FirebrandX (**) Date 2008-01-16 17:09
Forget the number of cores, I'm talking about end-result kN/s. If you build a PC with the latest Quad-core cpu, it will be 10x faster than my hunk of junk pentium 4 from years ago, not 2x faster.
Parent - By Nelson Hernandez (Gold) Date 2008-01-17 00:54
This is so pitiful, I still have a P4 running.  In fact I am writing this message on my five-year-old P4.  :(
Parent - - By Vempele (Silver) Date 2008-01-16 17:22 Edited 2008-01-16 17:25

> It needs to be based on ONE core at the SAME clock speed.


Why? The OP was clearly comparing a quad against a single processor. He specified the clock speeds: 3.2 vs 3. Definitely 64-bit - otherwise it'd be implausible.

You seem to have mixed the subthreads.
Parent - - By Carl Bicknell (*****) Date 2008-01-16 23:13
no, the reply was to me which was (in effect) about a coefficent.
Obviously the end result is higher than 2 but to get a definative list of processor speeds it would be far more helpful to do it the way I suggested so ALL core 2 architecture (for example) could be compaired quickly: quads, duals, and single threaded apps.
Parent - By Vempele (Silver) Date 2008-01-17 09:00

>yes make sure ponder is off!


>I'm REALLY surprised it's as much as 10x speed increase though - really?


No mention of a coefficient there as far as I can see. Nor in the 2 posts above it.
Parent - - By Silvian (***) Date 2008-01-16 12:04
WOW !
What can I do with these northmen ?
You have Rybka 3.2 and me (poor Silvian :)) only 2.3.2 a !
Nice ??????

Hello Vas, are you there ! I want to buy this new version !
Give me a link- please !

Thank you !
:)Silvian
Parent - By Silvian (***) Date 2008-01-17 07:35
WOW !
................. masses !

In any case: THANK YOU for your effort !

Regards,
Up Topic Rybka Support & Discussion / Rybka Discussion / engine strength-processor (RYBKA 3.2)

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill