p.s.i support Houdini,and i dislike rybka,for the ocasional reader
just a thought
Houdini1.5 or houdini2 is better than Rybka4 but it means nothing about the best houdini and the best rybka considering the fact that Vas released no new rybka in the last year.
I am not sure if people are going to consider Houdini as favourite to win the tournament in case of having both rybka and houdini in the tournament.
Note that when Rybka was the best based on rating lists and a clear favourite to win tournaments it did not stop Vas to participate in tournaments so it seems that the reason that you give for houdini not to participate is not convincing(otherwise you could expect rybka not to participate when she was leading the rating lists).
so for rybka,was easy to win every tournament,and easier if it was run on the cluster
now think about Houdini,the second engine is only about 25-30 point below?so putting that 2 engines in the same tournament,can be Houdini beaten,or not?easy answer:yes...so Robert claims:my engine is the strongest engine on the worl!and is true...but...if you put to play Houdini with the other engines that are near of him by only 25-30 points,and Houdini start lossing games and tournaments (that will occur,since Houdini isnt 150 elo points or 200 like rybka was),how Robert will claim that?people will say,yes,you tell your engine is the strongest in the world,but losses the tournaments
a very simply form to avoid it,and continue telling his engine is the stronger in the world without any single lost tournament?
dont participating in any tournament
(Or did someone remember that Vas stopped tournament participation when Rybka became No.1?)
Isn't it the poor truth: Houdini is not allowd to participate?
Or: Robert doesn't really feel, that Houdini is his child enough to present Houdini in that public!
No. I even recommended the CSVN to invite all the programs of the top 10 of the rating list and call their tournament 'The strongest computer chess event ever'.
I suspect other programmers may be boycotting, but that is their loss and their stupidity if so.
It's good to see Rybka back.
the ''decent'' engine is 22 elo points stronger than rybka based just in 1 core,so it doesnt need a strong hardware like seem you claim,you can see the ipon list
> if rybka plays with cluster then what >>>>>>>>>>>>>>
It shouldn't lose a single game or draw probably!
> we want to see the engine we are idolizing to use perform VS other top engines.
Hi mocha, I'd like to see it too, but 300 cores against 16 or so is overkill!
and i recall Houdini running on a 64 core machine,and for the moment Houdini only support 32 core machines
or running on a 32 core machine BUT with a gaining elo of about 40 extra points,like seems that Houdini 3 will have,and that 40 elo points is the same than doubling the cpu's (in terms of elo gaining)
But we know it looses to Highendman, but maybe it cant lose in tournament time control with its depth and tree size
that is why i told about Houdini wont play the tournament,to avoid something like that yout told
> the ''decent'' engine is 22 elo points stronger than rybka based just in 1 core,so it doesnt need a strong hardware like seem you claim,you can see the ipon list
LOL. If Rybka plays with 300 cores then Critter on a quad won't stand a chance.
but in equal hardware,critter will have chances against rybka
rybka playing with cluster only wins for the powerful of the hardware...is the same that in a race,me witha ford,and you wilt a ferrari...even me being a top pilot,you will surely win
Houdini on 12 cores at 4 gig in near lane. Total 48 gig.
the cluster (at 2.5 Gh) isnt more than about 80 elo points stronger than Houdini running on a 16 core system at 4 Gh
including twenty-nine motherboards,
six hundred fifty-six gigabytes of RAM,
five terabytes of solid-state disks,
fifty-eight CPU sockets,
and two hundred ninety-six physical Intel Nehalem cores.
vs 12 cores.
maybe we ought to update the elo rating system.
296 cores vs 12 cores = 80 elo.
Unless you are saying Houdini is such a badass that the gap has closed to 80 elo.
I dont believe it, Vas is smarter than that.
I think your maths are a bit wrong...
2680GM vs 2600GM is not crushing to me, pretty close.
296 cores vs 12 cores seems crushing.
Unless we say the problem is the weak software, but I believe the commercial version is very strong.
Robert uses a 16 core machine,not a 12 core machine
do your maths right
16 cores by 2 by 2 by 2 by 2 = 256 cores so if you double 4 times the 'power' of the machine uses Rober,you get near the cores the cluster has
and Houdini can use 32 cores,and that means ONLY DOUBLING 3 TIMES THE PROCESSORS
so even more for me;if Robert wants to put the engine runing in a 32 core machine at 4 Gh,the cluster is only 3 times doubling processors...well,3 times and a few more,i dont want to use the calculator,if you want,do it yourself
4 times doubling processors scaling at 1,7 level programming like vas claimed is a gaining elo about 150 elo stronger than Houdini,and lets say another 25 elo for the other cpu's to 296
and now you must think the computer running Houdini runs at 4 Gh (from your post cluster at 2.5 Gh) ,and that means is another plus for Houdini of about,lest say,35-40 elo points
now use your ''maths'':
Houdini ahead of Rybka by about 60 elo points,minus 175 elo points that is stronger the cluster,plus 35 elo point from the more speed that have the 32 core system:
60-175+35= -80 elo points,pro to rybka,like i claimed
now lets use a 32 core system,like Houdini is able to use...since using a 32 core system is reducing one double core,is reducing about 35 elo points (i remember you im using the EFECTIVE DOUBLING,NOT REAL DOUBLING,that is why im using only a value of 35 elo points instead of 50 elo points),so if you substract 35 elo points to the total of elo point pro to rybka:
80-35=45 elo points
so in general terms,a 32 core computer runing Houdini at 4 Gh is only about 50 elo points below 'rybka cluster' runing rybka at 2.5 Gh...
do the maths yourself
I thought your system was 2 socket AMD 8 core or 4 socket x 4 core.
I thought when we tested Houdini kns in the beginning your system was slower than my SR-2.
and another thing, why would anyone pay big money to rent the cluster if Houdini for $100 is as strong as you say Jerry.
Someone should tell Anand to just buy Houdini for $100 and SR-2 for $3500, way cheaper than cluster.
do your maths right.
you get near the cores the cluster has
what, you think Robert has how many cores?
I said 12 x 4gig because that is what mine is and I knew my kns was more than Roberts and the cluster should draw or beat mine 100 times out of 100 games, and I assume Vas has best book.
anyway it seems I am a bigger fan of the monster 300 core cluster running the commercial version of Rybka than you are.
Houdini on a 12 core system cannot beat the Rybka cluster if both books are best and no bugs, disconnects, no operator errors, etc.
do the maths Jerry. you know you kind of got me excited thinking about the Cluster, maybe it is not near as strong as I thought it was.
I put SuperGM at 2850, Houdini at 3100 and cluster at 3300.
SuperGM on a good nights sleep, Houdini on 12 and Cluster on 300
do the maths.
Dont take elo so serious, it is just a tool to keep a group of humans in some type of order for a short period of time.
is as symply as you need improve your maths
take a 32 (NO AMD) CORE SYSTEM RUNNING AT 4 GH,and do the maths yourself,i made it for you,but seems that you dont understand a symply maths:
a 32 core system runing at 4 gh,is only about 50 elo points weaker than the cluster rybka
if you dont understand that from 32 cores to 256 you only need double 3 times the cores,and the real doubling gain elo (with a scaling of 1,7) is about 150 elo points,is your problem,not mine
and as i claimed,a 32 core system is only 50 elo points weaker than the cluster,so if you substract 16 cores,you must add 35-50 elo point,so a 16 core system (NO AMD) runing at 4 Gh IS ABOUT 80-100 elo points weaker than the cluster only!
again,you are wrong with your maths,not me
>and another thing, why would anyone pay big money to rent the cluster if Houdini for $100 is as strong as you say Jerry.
people is *****...yes,that
why 32 core, I thought we decided HT should be off for chess programs.
Rybka is #1 in my book with 300 cores and unknown commercial Release above 4.1 strength.
With 300 cores maybe Houdini cannot even draw in tournament time control against Cluster Rybka.
It is everyone else playing for second place.
2)you are confusing 'cores' and 'threads'
2)and again,you seem unable to understand very simply maths,so my wasted time ends here
I do not understand your post.
If Robert has a 16 core pc then he has 32 threads available with HT on.
He does not have 32 cores available.
I tested HT and it didnt do much for my system so I just run 12 threads with 12 cores and HT off.
I dont like silicone breasts, I dont like HT, I dont like bad breath, I dont like big buts on women, and I dont like Monty Python.
I cannot care too much about an elo rating, there are many variables with that word, in computer programming and in human play.
It is a much needed system and I respect it, but it is misleading info sometimes.
I respect the energy that you were putting in your posts, but at this point in my understanding no one is going to downplay the Strength of "The Unreleased Cluster Version" of Rybka on 300 cores. I have gathered all of the cluster games I could find and ran them through my 12 core system and I am convinced it plays on another level that any program on standard hardware is just playing for second "in tournament time control" "unattended".
Now if someone wants to play one move in three days using multiple computers, with Idea or Aquarium and esp if he has a respectable rating then this is a lethal combination such as in WBCCC and the Cluster Rybka can lose to this combo.
> If Robert has a 16 core pc then he has 32 threads available with HT on.
> He does not have 32 cores available.
correct. (he might have a 32 core machine though)
> I tested HT and it didnt do much for my system so I just run 12 threads with 12 cores and HT off.
That's exactly right. On my 12 core I've turned HT off as has every sensible person.
and perceive the correlation about the strength of Houdini 2 running on 12 cores and cluster rybka running 40 cores,and extrapolate the results to what i said about Houdini running on 32 cores and Rybka cluster running on 296 cores,and tell me who is wrong,you or me
the same correlation,ratio 1/3
who is stronger,Houdini at 12 cores or Rybka playing at 40 cores?
even is worst than i imagined,and im really happy to prove that i were RIGHT and you were WRONG
guy at 11 seconds remembered a lot a video that i made of myself when i was military with my comrades,before we were arrested
and that made me feel nostalgic...what times...
so thanks,im not being sarcastic
> i think everyone is missing the point here,cluster is not running Rybka 4.1.
Yes I was thinking that too. Even on equal hardware private Rybka is much stronger than Critter 1.4.
My guess is that Critter is weaker than the newest Rybka that only Vas has on equal hardware(even when you do not consider the fact that Rybka is probably going to use better hardware.
I am not sure if latest rybka is better or worse than latest houdini on equal hardware but I am almost sure that latest rybka is better than latest critter on equal hardware.
pick Houdini 2.0c against rybka,run a lets say 200 games tournament (4m +2s),and you will see the result
p.s. if you want run a faster tournament,run a 2m+2s
and that means that will be about 100-110 elo poins stronger than rybka,that at the moment,still havent a release date (Houdini is released for september,but from his word,may be even sooner)
You can con most of the people most of the time but not all people all the time.
In the meantime Robert will have made a lot of money out of blitz players.
In a game i played against Ruben,Harvey asked why Ruben why he did not play Hiarcs book move,Rubens reply was simple and to the point.Because it loses.
Moral of story be careful of some book lines.
>Harvey asked why Ruben why he did not play Hiarcs book move,Rubens reply was simple and to the point.Because it loses.
i lost one of my correspondence chess game,because i was stupid,as easy as that...having the new Hiarcs opening book,i started using it...when were remaining a lot of games (lets say,300 games in the book,or maybe more),i played the move marked in the book like the best,without double checking it with the engine...and the moving marked in the opening book like the best,losses the game...
when i played it,thinking was a good move,and i saw the reply of my opponent,and i saw the scores going higher and higher,my face was something like
edit:about the elo gaining between Houdini 1.5 and 2.0,was really very little...but well,lets see if Robert keeps his promisses...if he keeps it,i will continue trusting in him,if he doesnt keep it,i will stop trusting in his words
and knowing that,im pretty sure that the move i found at the book marked like good,when it was a lossing move,it was collected from playchess blitz games,because with an analisis of only 30 hour,Houdini found it was a lossing move
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill