Komodo 12.1 x64 (MCTS) @ CEGT 40/4

[ThatsIt]

Hi !

Just started:

http://cegt.forumieren.com/t1001-testin ... 1-x64-mcts

Best wishes,
G.S.
(CEGT team)

[ThatsIt]

Hi !

Test cancelled. New start with version 12.1.1.

http://cegt.forumieren.com/t1001-testin ... -mcts#2097

Best wishes,
G.S.
(CEGT team)

[ThatsIt]

Update, 600 games are played:

http://cegt.forumieren.com/t1001-testin ... -mcts#2097

This time a little lesser as Larry predicted,
but a very small sample until now!

Best wishes,
G.S.
(CEGT team)

[lkaufman]

This version has much better scaling than the original (k12 MCTS), so I think you will get significantly better results at 40/20. Based on my results at fairly similar time limits to what you use in 40/20, I would expect about 3010 on 1 cpu and about 3125 on four cpus at that level.

[ThatsIt]

Hi !

Now with 2 CPU too (same link).

The standard Komodo scales as follows @ CEGT 40/4:

Code: Select all

Komodo 11.3.1 x64 2CPU  3417 out of 1750 games + 75
Komodo 11.3.1 x64 1CPU  3342 out of 2300 games
 
Komodo 11.2.2 x64 2CPU  3392 out of 1900 games + 60
Komodo 11.2.2 x64 1CPU  3332 out of 3300 games
 
Komodo 11.01 x64 2CPU   3369 out of 1160 games + 60
Komodo 11.01 x64 1CPU   3309 out of 2650 games

                                             ø + 65

Best wishes,
G.S.
(CEGT team)

[lkaufman]

Your early results are quite good on two threads, nearly 3050 after just 59 games, which would be more than plus 100 elo from 1 thread. That's really too much; probably the two thread result will drop and the one thread result will rise.

Best,
Larry

[ThatsIt]

Update, 1100 games 1CPU and 100 games 2CPU are played.

http://cegt.forumieren.com/t1001-testin ... -mcts#2097

Best wishes,
G.S.
(CEGT team)

[ThatsIt]

Update, 1300 games 1CPU and 300 games 2CPU are played.
So far + 88 gain for 2CPU, but still a very small sample.

http://cegt.forumieren.com/t1001-testin ... 1-x64-mcts

Best wishes,
G.S.
(CEGT team)

[lkaufman]

Update, 1300 games 1CPU and 300 games 2CPU are played.
So far + 88 gain for 2CPU, but still a very small sample.

http://cegt.forumieren.com/t1001-testin ... 1-x64-mcts

Best wishes,
G.S.
(CEGT team)

The typical gain from 1 to 2 cpus for the strong engines on that list seems to be in the mid 60s, so anything above 70 would indicate better than normal scaling behavior. We'll see if it holds above 70.


Best regards,
Larry

[ThatsIt]

Update, 1500 games 1CPU and 700 games 2CPU are played.
Still + 88 gain for 2CPU.

http://cegt.forumieren.com/t1001-testin ... 1-x64-mcts

Best wishes,
G.S.
(CEGT team)