Genko Tests

[mehmet123]

Genko vs Stockfish:

Program Elo + - Games Score Av.Op. Draws

1 Genko 1.0 bmi2 : 2401 11 11 300 50.3 % 2399 92.7 %
2 Stockfish 191122 x64 bmi2 : 2399 11 11 300 49.7 % 2401 92.7 %


Individual statistics:

1 Genko 1.0 bmi2 : 2401 300 (+ 12,=278,- 10), 50.3 %

Stockfish 191122 x64 bmi2 : 300 (+ 12,=278,- 10), 50.3 %

2 Stockfish 191122 x64 bmi2 : 2399 300 (+ 10,=278,- 12), 49.7 %

Genko 1.0 bmi2 : 300 (+ 10,=278,- 12), 49.7 %


Game Conditions: Cutechess Gui, 1 Core , Core-i7 12700h ( 12 cores/ 20 threads), Concurrency:18, 1 min + 0.5 sec TC, Balsa 5 Moves Opening Book, 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/gmji8rn1 ... 3.pgn/file

[mehmet123]

Genko vs Kookaburra:

Program Elo + - Games Score Av.Op. Draws

1 Genko 1.0 bmi2 : 2402 9 9 500 50.5 % 2398 91.4 %
2 Kookaburra v1.01-bmi2 : 2398 9 9 500 49.5 % 2402 91.4 %


Individual statistics:

1 Genko 1.0 bmi2 : 2402 500 (+ 24,=457,- 19), 50.5 %

Kookaburra v1.01-bmi2 : 500 (+ 24,=457,- 19), 50.5 %

2 Kookaburra v1.01-bmi2 : 2398 500 (+ 19,=457,- 24), 49.5 %

Genko 1.0 bmi2 : 500 (+ 19,=457,- 24), 49.5 %


Game Conditions: Cutechess Gui, 1 Core , Core-i7 12700h ( 12 cores/ 20 threads), Concurrency:16, 2 min + 0.5 sec TC, Balsa 5 Moves Opening Book, 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/oucoopgn ... 4.pgn/file


Kookaburra is a Stockfish derivative chess engine that looks very promising. In my tests, I observed that this chess engine is one of the most powerful Stockfish derivative chess engine. That's why I choosed this chess engine as a competitor to Genko in this test.

[mehmet123]

Genko vs Stockfish: (Fischer Random Match)

Program Elo + - Games Score Av.Op. Draws

1 Genko 1.0 bmi2 : 2402 8 8 1800 50.5 % 2398 73.5 %
2 Stockfish 191122 x64 bmi2 : 2398 8 8 1800 49.5 % 2402 73.5 %

Individual statistics:

1 Genko 1.0 bmi2 : 2402 1800 (+247,=1323,-230), 50.5 %

Stockfish 191122 x64 bmi2 : 1800 (+247,=1323,-230), 50.5 %

2 Stockfish 191122 x64 bmi2 : 2398 1800 (+230,=1323,-247), 49.5 %

Genko 1.0 bmi2 : 1800 (+230,=1323,-247), 49.5 %


Game Conditions: Cutechess Gui, Core-i7 12700h (14 cores/ 20 threads), Concurrency:18, 1 Core , 1 min + 0.5 sec , Chess960 Book 3 Moves Opening Book, 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/s642y57o ... 5.pgn/file

[mehmet123]

Genko- Stockfish:

Program Elo + - Games Score Av.Op. Draws

1 Genko 1.0a bmi2 : 2402 14 14 1020 50.4 % 2398 55.0 %
2 Stockfish 231122 x64 bmi2 : 2398 14 14 1020 49.6 % 2402 55.0 %

Individual statistics:

1 Genko 1.0a bmi2 : 2402 1020 (+234,=561,-225), 50.4 %

Stockfish 231122 x64 bmi2 : 1020 (+234,=561,-225), 50.4 %

2 Stockfish 231122 x64 bmi2 : 2398 1020 (+225,=561,-234), 49.6 %

Genko 1.0a bmi2 : 1020 (+225,=561,-234), 49.6 %


Game Conditions: Cutechess Gui, 1 Core , Core-i7 12700h ( 14 cores/ 20 threads), Concurrency:18, 2 min + 0.5 sec TC, Dynamic Openings Test Suite 6 (from Dorsz), 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/jz10ic0e ... 8.pgn/file


The result is in error bars. Genko was undefeated in this test against Stockfish Dev, despite having a NNUE net other than the standard NNUE net and containing many failed patches (some minor code and parameter changes had been made in these patches)

[RubiChess]

Game Conditions: Cutechess Gui, 1 Core , Core-i7 12700h ( 14 cores/ 20 threads), Concurrency:18

I don't want to discuss about the engine but about something else: Core-i7 12700h has 14 cores,
- 6 "performance" cores with 4.7Ghz max and
- 8 "efficient" cores with 3.5Ghz max
Specs here

Doesn't that hardware kill every serious engineA vs. engineB test? The engine with more processes on the performance cores wins the test, no?
The OS will probably try to avoid moving a process from one core to another to keep the cache valid. So there may be no equal distribution of engine processes to core type.

Or do all the different cores run at the same speed under heavy load?

Regards, Andreas

PS. Curious how many people will read and hopefully answer my question even in this thread

[mehmet123]

Game Conditions: Cutechess Gui, 1 Core , Core-i7 12700h ( 14 cores/ 20 threads), Concurrency:18

I don't want to discuss about the engine but about something else: Core-i7 12700h has 14 cores,
- 6 "performance" cores with 4.7Ghz max and
- 8 "efficient" cores with 3.5Ghz max
Specs here

Doesn't that hardware kill every serious engineA vs. engineB test? The engine with more processes on the performance cores wins the test, no?
The OS will probably try to avoid moving a process from one core to another to keep the cache valid. So there may be no equal distribution of engine processes to core type.

Or do all the different cores run at the same speed under heavy load?

Regards, Andreas

PS. Curious how many people will read and hopefully answer my question even in this thread

At this test the concurrency is 18
System uses 6 performance cores / 10 hyperthread cores + 8 efficient cores .

For concurrency 14 the speed of Genko and Stockfish at P -Cores are more faster than the speed of Genko and Stockfish at E - Cores. (6P + 8E)
For concurrency 20 the speed of Genko and Stockfish at P -Cores are more slower than the speed of Genko and Stockfish at E -Cores. (6+6 P + 8E). Due to the activation of hyperthreading. Because only P- Cores have hyperthreading property.

All both chess engines use the same cpu (ponder off) and same chess engines has same speed. For example at a X engine (1,2,3...20 cores) vs X engine (1,2,3,... 20 cores) match the speed are same for both chess engines if the two engines have the same configuration.
Therefore, the processor structure doesn' t provide any advantage to any chess engine at during the match.

[RubiChess]

Sorry, it was the wrong place to ask.

[mehmet123]

Genko vs Stockfish:

Program Elo + - Games Score Av.Op. Draws

1 Genko 1.1 bmi2 : 2403 8 7 600 50.7 % 2397 92.8 %
2 Stockfish 211222 x64 bmi2 : 2397 7 8 600 49.2 % 2403 92.8 %

Individual statistics:

1 Genko 1.1 bmi2 : 2403 600 (+ 26,=557,- 17), 50.8 %

Stockfish 211222 x64 bmi2 : 600 (+ 26,=557,- 17), 50.7 %

2 Stockfish 211222 x64 bmi2 : 2397 600 (+ 17,=557,- 26), 49.2 %

Genko 1.1 bmi2 : 600 (+ 17,=557,- 26), 49.2 %


Game Conditions: Cutechess Gui, 1 Core, Core-i7 12700h ( 14 cores/ 20 threads), Concurrency:18, 1 min + 0.5 sec TC,SuperGM 4 Moves Opening Book, 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/n9rhh52b ... 5.pgn/file