My Komodo -- Houdini Test

[WuShock]

Close match all the way . Didn't notice any drop off in later stages.

i7 940 @ 2.93 GHz
2 threads each
F12 gui
ponder off
no EGTB
klo_50 openings
512 MB hash
Time Control : 4 mins + 2 secs


[/img]

Only 100 games , but first time I've ever had Houdini not win.....

D = 0 , Drawscore set to - 0 -

Tom G.

[lkaufman]

Close match all the way . Didn't notice any drop off in later stages.

i7 940 @ 2.93 GHz
2 threads each
F12 gui
ponder off
no EGTB
klo_50 openings
512 MB hash
Time Control : 4 mins + 2 secs


[/img]

Only 100 games , but first time I've ever had Houdini not win.....

D = 0 , Drawscore set to - 0 -

Tom G.

My final result after 200 game on 4 cores at 3' + 2" was 96 to 104 for exactly 48%. Not as good as yours but better than most. It's reassuring to see that my result is not an outlier. But I note that your test is just 2 cores. This is the only test I've seen on 2 cores. If others get good results on 2, it might mean that our MP is especially good for 2 cores, par for 4 cores, and gets good again with higher numbers (relative to standard MP). This last point was confirmed by my test against Critter 1.6a, where I got + 39 elo for 80 games on 4 cores, but a whopping +85 elo after a hundred games on 12 cores. Maybe I'll run some 2 core tests.

[geots]

Close match all the way . Didn't notice any drop off in later stages.

i7 940 @ 2.93 GHz
2 threads each
F12 gui
ponder off
no EGTB
klo_50 openings
512 MB hash
Time Control : 4 mins + 2 secs


[/img]

Only 100 games , but first time I've ever had Houdini not win.....

D = 0 , Drawscore set to - 0 -

Tom G.

You likely can't notice any dropoff. Hash is diff., time control couldn't be any more diff., your number of cores is too small. Needs to be at minimum 4 cores. Not trying to be cute- but 512MB hash at 4+2 reminds me of a pissing contest between 2 males. Good work on your part- I just don't agree with the hash.

gts.

[lkaufman]

Close match all the way . Didn't notice any drop off in later stages.

i7 940 @ 2.93 GHz
2 threads each
F12 gui
ponder off
no EGTB
klo_50 openings
512 MB hash
Time Control : 4 mins + 2 secs


[/img]

Only 100 games , but first time I've ever had Houdini not win.....

D = 0 , Drawscore set to - 0 -

Tom G.

You likely can't notice any dropoff. Hash is diff., time control couldn't be any more diff., your number of cores is too small. Needs to be at minimum 4 cores. Not trying to be cute- but 512MB hash at 4+2 reminds me of a pissing contest between 2 males. Good work on your part- I just don't agree with the hash.

gts.

512M hash is double what I and most testers use for blitz, but I can't see what your complaint is. Perhaps it might slow down the test by a percent or two, but that's hardly a serious matter. I also have no idea why you use only 128 hash yourself at 40/10, I'm pretty sure no one would recommend this low a figure at this level. This isn't a complaint or criticism, I'd just like to know what your reasons are. I would think it would just needlessly lower the quality of play, rather like using a faster time limit but without the benefit.

[mwyoung]

Close match all the way . Didn't notice any drop off in later stages.

i7 940 @ 2.93 GHz
2 threads each
F12 gui
ponder off
no EGTB
klo_50 openings
512 MB hash
Time Control : 4 mins + 2 secs


[/img]

Only 100 games , but first time I've ever had Houdini not win.....

D = 0 , Drawscore set to - 0 -

Tom G.

You likely can't notice any dropoff. Hash is diff., time control couldn't be any more diff., your number of cores is too small. Needs to be at minimum 4 cores. Not trying to be cute- but 512MB hash at 4+2 reminds me of a pissing contest between 2 males. Good work on your part- I just don't agree with the hash.

gts.

I have seen no difference that I could accurately measure using too much hash. I have tested using 1 MB vs 256 MB at game in 5 min. I got around a 20 Elo drop off. In self play with Houdini. I was expecting more.

[geots]

Close match all the way . Didn't notice any drop off in later stages.

i7 940 @ 2.93 GHz
2 threads each
F12 gui
ponder off
no EGTB
klo_50 openings
512 MB hash
Time Control : 4 mins + 2 secs


[/img]

Only 100 games , but first time I've ever had Houdini not win.....

D = 0 , Drawscore set to - 0 -

Tom G.

You likely can't notice any dropoff. Hash is diff., time control couldn't be any more diff., your number of cores is too small. Needs to be at minimum 4 cores. Not trying to be cute- but 512MB hash at 4+2 reminds me of a pissing contest between 2 males. Good work on your part- I just don't agree with the hash.

gts.

I have seen no difference that I could accurately measure using too much hash. I have tested using 1 MB vs 256 MB at game in 5 min. I got around a 20 Elo drop off. In self play with Houdini. I was expecting more.

Wonderful.