Yamashita! Congratulations! And also for winning the KGS tournament
Thanks, Remi!
I am singing the Yamashita song now.
Please stop it ^^;
The wife of Hideki? That's a surprise!
Yes, I was impressed her entry. And two women programmers in one
tournament was maybe first time. (Kasumi's author Manabe is also lady.)
I think my result is thanks for new machine. i7 980X 3.3GHz
(6 cores, with Hyper Threading) is good for MC Go program.
And it can run on 4.0GHz overclock easily. FudoGo run 4.0GHz.
threads playout/s in 19x19
1 2789
2 5457 x1.95 (multiply against 1 thread)
3 7833 x2.80
4 8788 x3.15
5 12304 x4.41
6 13594 x4.87
7 15439 x5.53
8 15970 x5.72
9 16308 x5.84
10 17374 x6.22
11 17937 x6.43
12 18936 x6.78
13 16924 x6.06 ---> over cores, as a test
14 15950 x5.71
Another Go program HiraBot author reported similar result.
Q9650 4threads 10sec 49752 Playout 100%
i7 980X 1 thread 10sec 17645 Playout 35%
2 thread 10sec 34062 Playout 68%
3 thread 10sec 50482 Playout 101%
4 thread 10sec 66560 Playout 134%
5 thread 10sec 82059 Playout 165%
6 thread 10sec 97493 Playout 196%
7 thread 10sec 105062 Playout 211%
8 thread 10sec 112469 Playout 226%
9 thread 10sec 119889 Playout 241%
10 thread 10sec 126885 Playout 255%
11 thread 10sec 133618 Playout 269%
12 thread 10sec 140533 Playout 282%
But in my shogi(Japanese chess) program, Hyper Threading is useless.
threads nps
1 461984/s
2 812457/s x1.75
3 1109094/s x2.40
4 1325327/s x2.86
5 1531969/s x3.31
6 1656106/s x3.58
7 1851148/s x4.00
8 1735507/s x3.75
9 1865190/s x4.03
10 1844316/s x3.99
11 1583967/s x3.42
12 1699983/s x3.67
Hiroshi Yamashita
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
