It has been a barren 6 months. I have been waiting patiently for all these programs and it is taking too long:
HIARCS 14
Houdini 3
Rybka 5
Komodo MP
Deep Fritz 13
Shredder 13
Critter 1.5
In the lull I decided to upgrade my hardware to a dual Westmere Xeon @4.2 GHz (12 cores) and I re-ran some of the tests. It had to happen one day: one program has finally surpassed the Rybka Cluster I rented last year. Also, I got my hands on the latest Vitruvius which looks quite interesting. At the moment I'm not totally convinced it's as aggressive as they claim, but it's early days yet.
HIARCS 14
Houdini 3
Rybka 5
Komodo MP
Deep Fritz 13
Shredder 13
Critter 1.5
In the lull I decided to upgrade my hardware to a dual Westmere Xeon @4.2 GHz (12 cores) and I re-ran some of the tests. It had to happen one day: one program has finally surpassed the Rybka Cluster I rented last year. Also, I got my hands on the latest Vitruvius which looks quite interesting. At the moment I'm not totally convinced it's as aggressive as they claim, but it's early days yet.
For those of you who want to see the comparison with my old quad core Nehalem @ 3.6 Ghz:
You forgot houdini 1.5a.
it's number 7 on the old list
> It had to happen one day: one program has finally surpassed the Rybka Cluster I rented last year.
Yes, I still think it happened too fast

Software advances so fast, the Rybka Cluster looks less and less impressive by the second.
> Yes, I still think it happened too fast
>
> Software advances so fast, the Rybka Cluster looks less and less impressive by the second.
I was a bit surprised too, although I imagine the Rybka Cluster is stronger now than it was a year ago.
edit: I also believe the 40 core Rybka Cluster from 1 year ago is about 50 elo stronger than Houdini 2 Pro on my 12 core.
As always interesting results.
For chess engines software progress is significantly faster than hardware progress.
My current Houdini dev version is about 115 Elo stronger than Houdini 1.03a released in July 2010 (it typically beats it by 66%-34% in 2 minute games). This means that Houdini "dev" on a single core is now about as strong as Houdini 1.03a on 6 to 8 cores.
In other words: in the span of 2 years the software improvement of Houdini has made computers 6 to 8 times more powerful for chess.
For chess engines software progress is significantly faster than hardware progress.
My current Houdini dev version is about 115 Elo stronger than Houdini 1.03a released in July 2010 (it typically beats it by 66%-34% in 2 minute games). This means that Houdini "dev" on a single core is now about as strong as Houdini 1.03a on 6 to 8 cores.
In other words: in the span of 2 years the software improvement of Houdini has made computers 6 to 8 times more powerful for chess.
How long do you think this rate of improvement can be sustained? At some point it has to start slowing down!
Historically speaking there's been about 1,000 Elo software progress since 1980, or about 30 to 35 Elo on the average per year.
115 Elo in 2 years is significantly higher than this historical trend.
I guess that at some point in time we will run out of useful ideas, it could be after Houdini 3 or after Houdini 10...
115 Elo in 2 years is significantly higher than this historical trend.
I guess that at some point in time we will run out of useful ideas, it could be after Houdini 3 or after Houdini 10...
>115 Elo in 2 years is significantly higher than this historical trend
tendentious reasoning,as always the improvement of all the things is faster at the beggining,and slower later,so is perfectly fine that at the begining of your development of Houdini you will improve faster the engine,than when you will be in the lets say Houdini 5 version (if none breaktrough appears,of course)
p.s.''tendentious'',i love that word...im gonna send it a marriage proposal
you like "tendentious" more than "obfuscation"?
Me gusta la confidencia del Sr Houdart. El se esta felicitando por una ganacia de 115 puntos y ese producto todavia ni esta en el mercado!
Me gusta la confidencia del Sr Houdart. El se esta felicitando por una ganacia de 115 puntos y ese producto todavia ni esta en el mercado!
creo que ha entendido mal lo que dijo...el dijo que Houdini ha ganado 115 puntos elo desde la version 1.03 hasta la version 2.0
well,i like both words
and i also like the quote ''like a hawk around the midnight in the sky''
well,i like both words

and i also like the quote ''like a hawk around the midnight in the sky''
entonces si entendi mal, porque yo pense que el estaba hablando de la mas reciente version que todavia no ha salido al publico
no,el hablaba que Houdini desde la version 1.03 hasta la version 2 subio 115 puntos elo
lo que si que ha dicho es que el incremento esperado para Houdini 3 es de 40 puntos elo sobre Houdini 2
lo que si que ha dicho es que el incremento esperado para Houdini 3 es de 40 puntos elo sobre Houdini 2
> no,el hablaba que Houdini desde la version 1.03 hasta la version 2 subio 115 puntos elo
No. Robert habla de Houdini ''dev'' que es la versión actual, no de la versión 2.
Gaмßito.
gracias.en cualquier caso,mi razonamiento sigue siendo correcto de que al principio se acumula mas aumento elo que despues de unos años
Glad to see Zappa on the list. What about Stockfish on 12 cores ?
I'll include a number of others in my next update, including Stockfish.
ok.
Hi can you try deep fritz 13 when it's out on single core as well? this would give a good indication of the difference between "fritz" and "deep fritz" under the same conditions.
> can you try deep fritz 13 when it's out
I doubt I'll buy it. The tactical strength of Fritz 13 is so weak that it's virtually useless for analysis. DF13 will cost a fortune and I can't see the point of spending all that money just to test it. If DF13 is a big elo jump I will get it, otherwise probably not.
Sorry.
We need Komodo MP!! :#
> We need Komodo MP!! :#
yes, but it could be ages yet.
i think,but of course i can be wrong,that they are trying to make the MP version since at least Komodo 3...and what i think is intolerable,is that they developed the ''paid'' version,WITHOUT being able to develop the MP version...one of the latest posts of Larry,he told about the MP version something like that:''now we have a version (about MP)that at least compiles...''
I am also surprised that Don didn't manage to release the MP version up to now bearing in mind his long experience in the field of muliprocessing according to his own words
Dr.D

Dr.D
well,i dont know Don and his long experience,i only know that they are trying to release the MP version,and they are failing from a lot of time ago
True....
so if he has that long experience with MP,i dont know where is he working...maybe the MP version is abandoned and they wouldnt give that detail to us
We will see a nice contest between rybka & houdini in the next years.
from the hard work he is putting developing rybka 5,the maximum that rybka 5 can imagine to afford is wash a car or cook some good food,so forget about the nice contest my friend
dont you see how hard vas is working in the project,and how much notices about the project he is posting on the forum to inform all the members about the hard work he is doing on rybka?
dont you see how hard vas is working in the project,and how much notices about the project he is posting on the forum to inform all the members about the hard work he is doing on rybka?
What do you know about Vas and the work he is doing re: R5?
Nothing,the same than you.and why?because he doesnt post anything since more than 1 year ago,and when he posted,he told he was "working" on rybka 5,just 1 time in more than 1 year,and studying poker
Im not wrong if i asure you that a lot of people is deceived about it and waiting rybka 5,but i can asure you that im not,i wouldnt rybka in my entines,even free or like a gift
Im not wrong if i asure you that a lot of people is deceived about it and waiting rybka 5,but i can asure you that im not,i wouldnt rybka in my entines,even free or like a gift
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill