Komodo 6, Houdiini 3, latest Stockfish dev. all equal.

[lkaufman]

I've run a decent number of games with Komodo 6 vs. Houdini 3 and Komodo 6 vs. Stockfish dev Sep 29, all in single core mode, with both results extremely close. The Houdini games were run on two i7 quads at 5' + 3" (ponder off), the Stockfish games on a 16 core machine at 5' + 0.3" (because longer increments caused time forfeits by Stockfish on this hardware). I used LittleBlitzer gui, with a huge opening book and color reversal (by using gauntlet mode). My results as of now are:

Komodo 6 vs Houdini 3: +102 -104 = 140
Komodo 6 vs. Stockfish dev. Sep29: +305 -304 =489

Both results virtually tied. I noticed that Komodo was slightly ahead of Houdini on my laptop but slightly behind on my desktop. Of course this could just be random, but I checked NPS and Komodo was running about 4% faster relative to Houdini on the laptop, so this is probably the reason. Conclusion: they are so close (at the level tested) that the exact hardware might decide which will win a long match. In general Komodo does better on Intel than on AMD.

[Nelson Hernandez]

In general Komodo does better on Intel than on AMD.

Is there a leading engine where the reverse is true?

[lkaufman]

In general Komodo does better on Intel than on AMD.

Is there a leading engine where the reverse is true?

I meant that in matches with other engines Komodo performs better (both by results and by relative NPS) on Intel than on AMD. I believe Houdini performs better against other engines on AMD. Of course all engines get more NPS on Intel than on AMD at the same clock speed.

[shrapnel]

In general Komodo does better on Intel than on AMD.

Is there a leading engine where the reverse is true?

I meant that in matches with other engines Komodo performs better (both by results and by relative NPS) on Intel than on AMD. I believe Houdini performs better against other engines on AMD. Of course all engines get more NPS on Intel than on AMD at the same clock speed.

False. Houdart has himself stated that Houdini performs better on Intel Processors.

[lkaufman]

In general Komodo does better on Intel than on AMD.

Is there a leading engine where the reverse is true?

I meant that in matches with other engines Komodo performs better (both by results and by relative NPS) on Intel than on AMD. I believe Houdini performs better against other engines on AMD. Of course all engines get more NPS on Intel than on AMD at the same clock speed.

False. Houdart has himself stated that Houdini performs better on Intel Processors.

Well, Houdini performs worse against Komodo on Intel than on AMD, especially if the Intel processors are new (this year's). Perhaps he just meant that Houdini gets more nodes per second on Intel, which is true of all engines, or perhaps it does better against some weaker engines on Intel.

[lkaufman]

Final results: K6 vs h3: +152 -166 =222 (-9 elo)
k6 vs SFsep29: +484 -468 = 828 (+ 3 elo)

In total, K6 scored exactly one point over 50% in 2320 games! Talk about close!

Now I'm moving on to four thread MP matches under similar conditions. Naturally it will take four times as long to get the same data. I'm using Fritz 11 GUI for this test on the two quads to see if I get any hangups or crash.

[Modern Times]

Looks like these 3 engines are too close to call, well within statistical error margins. Great job by Stockfish and Komodo. Enthusiasts have three great engines to choose from. Houdini 3 is a year old, I wonder what progress Robert has made since then.

[Dr.Wael Deeb]

In general Komodo does better on Intel than on AMD.

Is there a leading engine where the reverse is true?

I meant that in matches with other engines Komodo performs better (both by results and by relative NPS) on Intel than on AMD. I believe Houdini performs better against other engines on AMD. Of course all engines get more NPS on Intel than on AMD at the same clock speed.

False. Houdart has himself stated that Houdini performs better on Intel Processors.

Never the less,Houdini's author performs his main test framework on an AMD system....

Maybe this simple fact caused the confusion here....
Dr.D

[Dr.Wael Deeb]

Looks like these 3 engines are too close to call, well within statistical error margins. Great job by Stockfish and Komodo. Enthusiasts have three great engines to choose from. Houdini 3 is a year old, I wonder what progress Robert has made since then.

I expect around 40-50 Elo improvement Ray....
Dr.D

[mwyoung]

I've run a decent number of games with Komodo 6 vs. Houdini 3 and Komodo 6 vs. Stockfish dev Sep 29, all in single core mode, with both results extremely close. The Houdini games were run on two i7 quads at 5' + 3" (ponder off), the Stockfish games on a 16 core machine at 5' + 0.3" (because longer increments caused time forfeits by Stockfish on this hardware). I used LittleBlitzer gui, with a huge opening book and color reversal (by using gauntlet mode). My results as of now are:

Komodo 6 vs Houdini 3: +102 -104 = 140
Komodo 6 vs. Stockfish dev. Sep29: +305 -304 =489

Both results virtually tied. I noticed that Komodo was slightly ahead of Houdini on my laptop but slightly behind on my desktop. Of course this could just be random, but I checked NPS and Komodo was running about 4% faster relative to Houdini on the laptop, so this is probably the reason. Conclusion: they are so close (at the level tested) that the exact hardware might decide which will win a long match. In general Komodo does better on Intel than on AMD.

I am only getting close results Stockfish Vs. Houdini 3, if Houdini 3 is set to contempt = 0 on my system running with a i7 intel cpu. I have been testing Stockfish Vs Houdini 3 at long time controls for several days now with Houdini 3 set at contempt = 0. Here is the results so far....

Code: Select all

Blitz, 120m/40+60m/20+(15m+30s)  0

                                      
1   Stockfish 280913 64 SSE4.2   +11  +5/=22/-4 51.61%   16.0/31
2   Houdini 3 Pro x64            -11  +4/=22/-5 48.39%   15.0/31