TalkChess.com

[Timo Klaustermeyer]

Hi Don,

I guess you got a lot of feedback, but anyways (for what it's good for) here are my results for Komodo 4+5 and Houdini 2.0c Pro:

512MB Hash, Depth 20, Houdini uses 1 thread

Platform: Intel Xeon 5660 @ 3.06GHz
-----------------------------------------------
Houdini 2:
info depth 20 time 46803 nodes 91415743

Komodo 4:
info depth 20 time 33859 nodes 31373049

Komodo 5:
info depth 20 time 79390 nodes 92104642



Platform: AMD Phenom II x6 @ 3.00GHz
--------------------------------------------------
Houdini 2:
info depth 20 time 58296 nodes 91415743

Komodo 4:
info depth 20 time 38672 nodes 31373049

Komodo 5:
info depth 20 time 95737 nodes 92104642


Comparison
---------------
Houdini-Ratio:
46803 / 58296 = 0,80285

Komodo4-Ratio:
33859 / 38672 = 0,87554

Komodo5-Ratio:
79390 / 95737 = 0,82925


We can see that Houdini suffers most from changing platform from Intel to AMD. The Komodo4 compile was a bit better on AMD than Komodo5, but still Komodo5 loses less than Houdini on AMD platform.

So there is no need for a special AMD compile I guess. Maybe Robert should think about an AMD compile for his next Houdini version to score even better on IPON rating list (which uses only AMD platforms for tests)...

Best regards
Timo