Testing Quocvuong82 Stockfish 2.3 SSE 4.2 - 6 cores.

[Tomcass]

TESTING Stockfish2.3QUO SSE 4.2 120 GAMES

I7 980 3.33 Ghz.
6 real cores
GUI: Fritz 12
Book: Perfect 2012
Time control: 2 minutes +2 sec/ game.
Ponder: Off
No tablebases

201209Stockfish2.3iQUO 2012

Stockfish 2.3Qi 64bit SSE4.2x6- Critter 1.6 64-bitx6_nob 19.0 - 21.0 +9/=20/-11 47.50%
Stockfish 2.3Qi 64bit SSE4.2x6- Deep Rybka 4.1 SSE42 x64 (x6) 21.0 - 19.0 +10/=22/-8 52.50%
Stockfish 2.3Qi 64bit SSE4.2x6- Houdini 2.0c Pro x64(x6) 18.0 - 22.0 +10/=16/-14 45.00%

58.0 – 62.0 Average score after 120 games = 48.33% (!!) Excellent!.

Games: http://www.mediafire.com/?1q6if15xzfp2b56

Regards from Barcelona.

Tom.

[S.Taylor]

TESTING Stockfish2.3QUO SSE 4.2 120 GAMES

I7 980 3.33 Ghz.
6 real cores
GUI: Fritz 12
Book: Perfect 2012
Time control: 2 minutes +2 sec/ game.
Ponder: Off
No tablebases

201209Stockfish2.3iQUO 2012

Stockfish 2.3Qi 64bit SSE4.2x6- Critter 1.6 64-bitx6_nob 19.0 - 21.0 +9/=20/-11 47.50%
Stockfish 2.3Qi 64bit SSE4.2x6- Deep Rybka 4.1 SSE42 x64 (x6) 21.0 - 19.0 +10/=22/-8 52.50%
Stockfish 2.3Qi 64bit SSE4.2x6- Houdini 2.0c Pro x64(x6) 18.0 - 22.0 +10/=16/-14 45.00%

58.0 – 62.0 Average score after 120 games = 48.33% (!!) Excellent!.

Games: http://www.mediafire.com/?1q6if15xzfp2b56

Regards from Barcelona.

Tom.

Why not make it 2 seconds per game?
No problem, just use 100 real cores!

[Tomcass]

I will ask a computer with 96 real cores to The Three Kings . Who knows?.

Tom.

[Tomcass]

TESTING Stockfish2.3.1 QUO SSE 4.2 120 GAMES

I7 980 3.33 Ghz.
6 real cores
GUI: Fritz 12
Book: Perfect 2012
Time control: 2 minutes +2 sec/ game.
Ponder: Off
No tablebases

201209Stockfish2.3.1 QUO-1

Stockfish 2.3.1Qi x64 SSE4.2x6 - Critter 1.6 64-bitx6_nob 22.0 - 18.0 +14/=16/-10 55.00%
Stockfish 2.3.1Qi x64 SSE4.2x6 -Deep Rybka 4.1 SSE42 x64 (x6) 20.5 - 19.5 +9/=23/-8 51.25%
Stockfish 2.3.1Qi x64 SSE4.2x6 - Houdini 2.0c Pro x64(x6) 18.0 - 22.0 +12/=12/-16 45.00%

60.5 – 59.5 Average score after 120 games = 50.42%

Games: http://www.mediafire.com/?r2alt828l6rvkc2

Before suggesting that we can have a possible new World Number 2, I will repeat this test. Congratulations to Marco Costalba and team for this new Stockfish and to Quocvuong for his very good compile.

Regards from Barcelona.

Tom

[Vinvin]

What's the difference with the regular Stockfish ?

[Tomcass]

What's the difference with the regular Stockfish ?

Hi, Vincent, I can not answer your question. I am simply a tester. From a source made by Marco Costalba and team, there are some great compilers such as Jim Ablet, PeterPan -Izak-, Quocvuong 82 or Velasco Marín and some others that make the files ready to be used in a GUI such as Chessbase - Fritz. Any of them can explain better than me what a compiler does.

In my role of simple tester I try to discover which compile is better. In this case, Jim's compile is very fast but it seems that Quocvuong's is stronger. Obviously I can be wrong. It depends of a sufficient number of games, time control, book used and number of cores of the computer.

Tonight I hope I will be able to post the second part of my test of the latest compile made by Quocvuong. The games are simply exciting. Let's wait for the final result. I will post it later together with the games.

Regards from Barcelona.

Tom

[Vinvin]

If it's only a compiler difference, no need to play hundreds of games, simply take 4-5 positions and look at the Nodes/sec in mono-thread (or number of node after 1 min).

Or may be simply type a bench command if there's one (I think there's one in SF).

Vincent

[Tomcass]

The clue is that the best compile is not always the fastest one, Vincent. This is the reason of the need for testing intensively at different time controls. Usually I try to concentrate my 120 games tests at 2 min + 2 sec/move against three specific engines -using 4 or 6 real cores- and when I am sure that the analyzed compile is strong, then I test it at longer time controls against the same engines. Currently Critter 1.6a , Deep Rybka 4.1 and Houdini 2.0c Pro.

Other testers work in a different way. I do it only for my own enjoyment, but if I find an exceptional source + compile such this Stockfish 2.3.1 compiled by Quocvuong I share my results with the computer chess lovers.

Regards,

Tom.

[Carlos Ylich]

Thanks Tom,
Great Job !!

[Vinvin]

The clue is that the best compile is not always the fastest one, Vincent.
...
Regards,

Tom.

If only the compile changed and not the code, fastest executable is _ALWAYS_ the strongest one.