Some Questions to Chess Experts

By Sedat Canbaz Date 2012-08-29 10:39

Dear Chess Experts,

I wonder a lot about:
-Why Deep Shredder 12's performance suffer in SCCT competitions ?
-Is there is something wrong in my used test engine conditions ?
*Or maybe,the reasons are my used hardwares or my opening books or ... ?!
*Or DS12's low performance can be due to Elo error bar margins ?

Or Who Knows ?!
Perhaps SCCT does not serve reliable data and i need to:
- switch to 1 core engine testings ?

Some notes about DS12's conditions in SCCT:
-Played with Opening Advise OFF
-Played mainly under Fritz GUIs
-Used Perfect 2012 book series
-Played with 6 cores on my i7 6-core machines

*For full details about SCCT conditions:
http://www.sedatcanbaz.com/chess/2838-2/

*Its interesting to note also,
In some other test conditions (NON-SCCT),DS12 engine performs much better than SCCT !

For example,here are available some Elo comparisons between Deep Shredder 12 and Zappa Mexico II
*Note also that Spike 1.4 Leiden,Spark 1.0,Deep Junior 13.3... performed better than DS12 (in SCCT)
Where exactly these engines are rated lower than DS12 in some other testings (Non-SCCT results)

Code:


SCCT 3m+2s = Elo difference: 30,6

37 Deep Shredder 12 x64 6c       3041.7     
39 Zappa Mexico II x64 6c        3011.1   

----------------------------------------------

Clemens 5m+3s = Elo difference: 47

  22 Deep Shredder 12 UCI x1      2800   
  30 Zappa Mexico II x1           2753

----------------------------------------------

IPON 5m+3s = Elo difference: 97

9 Deep Shredder 12           2800 
19 Zappa Mexico II           2703

----------------------------------------------

CEGT 40/4 = Elo difference: 61

12   Deep Shredder 12 x64 4CPU     2894 
19   Zappa Mexico II x64 4CPU      2833
 
----------------------------------------------

CEGT 40/20 PB ON = Elo difference: 56

10   Deep Shredder 12 x64      2800 
19   Zappa Mexico II x64       2744

----------------------------------------------

CCRL 40/40 = Elo difference: 58

13  Deep Shredder 12 64-bit OA On 4CPU   3136
16  Zappa Mexico II 64-bit 4CPU          3078

----------------------------------------------

CCRL 40/4 = Elo difference: 91

10  Deep Shredder 12 64-bit OA On 4CPU   3093
17  Zappa Mexico II 64-bit 4CPU          3002

----------------------------------------------

SWCR 40/10 = Elo difference: 74

60 Shredder 12 w32                2800 
21 87 Zappa Mexico II x64         2726

----------------------------------------------

SSDF 40/120 = Elo difference: 57

5  Deep Shredder 12 x64 2GB Q6600 2,4 GHz   3115
13  Zappa Mexico II x64 2GB Q6600 2,4 GHz   3058

For More Details:
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=25552

Thanks in advance,
Sedat

By Kurt Utzinger Date 2012-08-31 18:11

Hi Sedat

Is it not well known that MP-Shredder does only perform slightly better than single core Shredder?

Kurt

By Simon Gros Date 2012-08-31 18:30

Also wenn ich mir dies hier ansehe:
http://computerschach.forumfrei.net/t107-cegt-blitz-vergleich-1cpu-4cpu?highlight=Vergleich,
scheint Shredder 12 von SMP durchschnittlich zu profitieren. Der Betreiber dieser Seite oder dieses Forums scheint einer der wenigen zu sein, welcher sich mit diesem Thema schon häufiger beschäftigt hat.
Wie kommen Sie zu Ihrer Meinung/Einschätzung? Gibt es andere (weitere) Vergleiche zu diesem Thema?
Simon_G

By Ernest Bonnem Date 2012-08-31 18:47 Edited 2012-08-31 18:50

[quote="Kurt Utzinger"]Is it not well known... [/quote]
If you look at the rating lists (CCRL, CEGT...) you will find that it is not so, especially taking in account the error bars.

And anyway, why would SMK be less SMP-competent than the other programmers?

P.S. I just see Simon Gros also answered in the same way...

By Sedat Canbaz Date 2012-09-01 19:03 Edited 2012-09-01 19:11

A few calculations more...

For more reliable data,this time i preferred to use Hiarcs 13.2 chess engine

And as we see...again DS12 6 cores has the lowest performance in SCCT (as before)
It seems, (especially on AMD platforms) DS12 on single core testings is performing much better

Code:


SCCT 3m+2s = Elo difference: 13.1

  37 Deep Shredder 12 x64 6c     3039.6   
  38 Hiarcs 13.2 w32 6c          3026.5

*********************************************

Clemens 5m+3s = Elo difference: 24

  23 Deep Shredder 12 x64 x1      2796 
  26 HIARCS 13.2 MP T1            2772

*********************************************

IPON 5m+3s = Elo difference: 52

32 Deep Shredder 12         2800
50 HIARCS 13.2 MP 32b       2748 

*********************************************

CCRL 40/4 = Elo difference: 31

Deep Shredder 12 64-bit OA On 1CPU  2977
Hiarcs 13.2 1CPU                    2946

---------------------------------------------

CCRL 40/40 = Elo difference: 38

Deep Shredder 12 64-bit OA On 4CPU  3061
Hiarcs 13.2 4CPU                    3023

*********************************************

CEGT 40/4= Elo difference: 26

Deep Shredder 12 x64 1CPU   2800
Hiarcs 13.2 1CPU            2774

---------------------------------------------

CEGT 40/20 = Elo difference: 19

Deep Shredder 12 x64 4CPU   2873
Hiarcs 13.2 MP 4CPU         2854

---------------------------------------------

CEGT 40/20 PB ON = Elo difference: 44

Deep Shredder 12 x64   2800 
Hiarcs 13.2            2756

*********************************************

SWCR 40/10 = Elo difference: 39

Shredder 12 w32              2800
Hiarcs 13.2 w32              2761

Greetings,
Sedat