10 minutes blitz

After 26 games.
Blitz:10'  0     

1   Deep Rybka 4 w32         +7/-6/=13 51.92%   13.5/26
2   Houdini 1.03a w32 2_CPU  +6/-7/=13 48.08%   12.5/26

Rybka is 13 Elo stronger.
It seems that Houdini 1.03a tuned for 5 minutes or less
and for longer games it is weaker. running 20 or 40
minutes games should be interesting decision.
Note: it is possible that 26 games is not sufficient.

Houdini 1.03

GUI: fritz
5 minutes games
Book: Rybka3Gen.ctg (optimize)
large pages disabled.

After 180 Games

    Program                                     Elo      +     -    Games  Score    Av.Op.  Draws

  1 Houdini 1.03 w32 2_CPU       : 3178   36  34   111     55.9 %   3137    70.3 %
  2 IvanHoe-Beta-999955m-Beta : 3146   32  33   126     49.2 %   3151    71.4 %
  3 Deep Rybka 4 w32                 : 3129   33  34   123     45.5 %   3160    69.9 %


It seems Houdini 1.03 is strongest among these engines and it is very fast.

New tests

I couldn't find some evidence about cloning claim or rejection of it; but it is possible that ippolit engines are clones of rybka.
so this is my new test:

GUI: fritz
5 minutes games
Book: Rybka3Gen.ctg (optimize)
large pages disabled.
                              
1   IvanHoe-Beta-999955m-Beta  +12/-8/=49 52.90%   36.5/69
2   Deep Rybka 4 w32            +8/-12/=49 47.10%   32.5/69



or +20 Elo.
this is my build with SSE 2 support (I compile it only; without robbobases).
note: these are 69 games.