Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Dutch Open 2009: Rybka wins!
1 2 Previous Next  
- - By Venator (Silver) Date 2009-10-18 16:06
Rybka has just won the Dutch Open 2009 tournament in Leiden, for a record 4th consecutive title. Rybka scored 7.5/8, the only half point being lost to The Baron that played an excellent game.
Parent - - By staylor (***) Date 2009-10-18 16:28
did this tournament include the top programs?
Parent - - By Lukas Cimiotti (Bronze) Date 2009-10-18 17:11
It did include Deep Sjeng on 64 cores - second in last WCCC and very good. They got 2. place which was well deserved.
Parent - By Roland Rösler (****) Date 2009-10-18 21:33

>They got 2. place which was well deserved.


They made 7/8! So, if there would be no game Deep Sjeng vs Cluster Rybka, they would have won the tournament! :-)
Parent - By FWCC (***) Date 2009-10-19 23:05
Where are the games posted or do you have a pgn file?
Parent - - By Felix Kling (Gold) Date 2009-10-18 17:30
Rybka and Sjeng have the strongest chess playing units at the moment, so the win against Sjeng was basically the decision, even if the Baron managed to draw.
Parent - - By Banned for Life (Gold) Date 2009-10-18 19:08
"strongest chess playing units" :-) Well worded. The fact that the winner was playing on the strongest hardware and the runner up was playing on the second strongest hardware was just a coincidence in this case! :-D
Parent - - By Geomusic (*****) Date 2009-10-18 20:05
they will power pocket calculators in 5 yrs :) I just noticed I called them "Pocket Calculators" ugh showing my age.
Parent - - By Roland Rösler (****) Date 2009-10-18 21:39

>they will power pocket calculators in 5 yrs :-)


I don´t believe that! They will not have enough power to play a 75' game with ponder=on! :-)
Parent - - By Eelco de Groot (***) Date 2009-10-19 22:50

>> they will power pocket calculators in 5 yrs :-)
> I don´t believe that! They will not have enough power to play a 75' game with ponder=on! :-)


Some of those early pocket calculators need incredibly little power. I have a programmable Casio FX-501P calculator which I bought in 1980, it was really used a lot when I was a student and I still have it within reach, but in all those 29 years, I only recall once having had the batteries replaced! Two little LR44 is all it needs, the kind of batteries that go into watches. According to the manual, the calculator only consumes 0.0008W!



128 program steps is not enough for a chess program though, a 'tablebase' for playing Tic Tac Toe perfectly was about the limit :)

Then again I also have a much later American model Texas Instruments that is not so energy efficient, 1.2 V rechargeable batteries don't give it enough power, it uses 4 AAA 1.5 V batteries which are much larger than the batteries for the Casio, plus a button cell and if I don't replace them in time or the separate battery for memory contents, it does not remember any of the programs I stored in its memory bank. It does use a CMOS version of the Z80 which uses much more current than the chip for the Japanese Casio, but not as much as the original Z80 of course. The Z80 microprocessor powered many of the early 8 bit home computers like the Tandy TRS 80. In principle one should be able to translate the Sargon chessprogram assembly book and with some hacking run Sargon I on the TI85! But I have never tried that...

Parent - - By Roland Rösler (****) Date 2009-10-19 23:38
Okay I see, we only need programming skills! :-)
Here in this forum we discussed frequently Elo performance of God in chess. I´m not so much interested in his chess skills, but in his programming skills!
Is God able to write a programm for Casio FX-501P (or current iPhone (I´m agnostic! :-))) which is equal to (or better than) Cluster Rybka 3w91 on 9 Nehalem EP?

PS: If God can do it with the Casio, I will convert! :-)
PS2: There isn´t enough place in the Casio to store his 32-men!
Parent - By Eelco de Groot (***) Date 2009-10-20 01:22

> Okay I see, we only need programming skills! :-)


Yes I think you are right about that Roland! Hardware is still moving along so rapidly that state of the art programming I think, in general, is lagging behind, it just can't evolve fast enough to make full use of everything that is possible with hardware. I imagine if you wanted to write something that is optimal for today's technology, the hardware is already much further once you are ready and the consumer still has the computers of a few years ago, they are more familiar with that and could not run your program optimally anyway.

Maybe the Japanese will make a supercomputer that can fit in your pocket. There were some pictures discovered recently of Fischer visiting the Sony plant in Tokio in 1973, I wonder if there was any chess interest there from Bobby, maybe Sony had some plans for making chess computers at the time? Were there already chesscomputers in 1973? Maybe not yet... Also there are not many chess players in Japan itself, Shogi is more popular and Go too. So without a market for that they would have had to sell everthing in Europe and the USA, without a market you don't begin such a thing. That's why it is interesting that chess is becoming more popular in China and in India too with their own world champion Anand, much opportunity and I'm sure also new chess programming talent will come from there! Not from Russia I think, at least after making Kaissa I don't think they lived up to their potential. But there it was more that they could not compete against Western hardware, only in their home market. Now they all have iPods and iPhones, at least the rich Russians do. So maybe Russian chessprogramming will finally take up again and continue where they left off after making Kaissa. I'm not counting Pioneer by Botwinnik, that it seems was never a real program.

> Here in this forum we discussed frequently Elo performance of God in chess. I´m not so much interested in his chess skills, but in his programming skills!
> Is God able to write a programm for Casio FX-501P (or current iPhone (I´m agnostic! :-))) which is equal to (or better than) Cluster Rybka 3w91 on 9 Nehalem EP?
>
> PS: If God can do it with the Casio, I will convert! :-)
> PS2: There isn´t enough place in the Casio to store his 32-men!

Parent - By Bill Forster (**) Date 2009-10-22 22:08
Wonderful! Thanks for posting this. I also bought one of these calculators as a student, in the same year. That was my first year at engineering school. I remember there were two versions, identical except that one had twice as much memory. I can't remember whether the 501 was the big dog or the little dog. Maybe the 502 was the big dog? I also can't remember whether I decided to spring for the big version myself, although I remember it seemed a huge decision for me at the time :-)

I feel particularly nostalgic about this calculator because I wrote my very first program on it. In fact it was only by playing with the calculator I got to understand what the concept of a program really was. I remember my first program well. At the time the students in my residence hall were all playing dominoes, it was something of a fad. My program took as input the tiles it was dealt, in the form of a list of two digit numbers, with each digit constrained to the range 1 through 6. Plus the end points of the current snake on the table, again a two digit number of the same format. It would then attempt to calculate an optimal move. My algorithm was to generate a list of legal moves, to score each one, then to pick the best scoring move. The scoring function calculated the standard deviation of the set of remaining single digit tile halves assuming the move under consideration was played. A large standard deviation was a good thing because that was associated with a wider variety of remaining single digits, which in turn meant a reduced likelihood of being blocked later on. I must admit I chose the algorithm solely because the calculator had a built in standard deviation function.

None of the other players had the slightest objection to me playing games using my "engine" as an assistant, how times have changed!

Thanks again for posting. I wish I had been as wise as you and kept this great little calculator.
Parent - - By Roland Rösler (****) Date 2009-10-18 21:48 Edited 2009-10-18 21:50
It would be interesting to see a match over 100 games (50 opening positions) Rybka 3t88 (or w91) on notebook (x64, C2D with ~2.5 GHz) vs. Deep Sjeng on tournament hardware.
Your estimation?
Parent - By Banned for Life (Gold) Date 2009-10-18 23:45
I think DS would win this, but I don't think it would be a blowout.
Parent - - By GCP (***) Date 2009-10-19 06:11
How strong is Rybka 3t88 / 3w91?
Parent - - By Roland Rösler (****) Date 2009-10-19 07:29
Nobody knows! My estimation: ~ +300 Elo better than Deep Sjeng 3.0 and ~ +70 Elo better than Rybka 3.
Parent - - By Lukas Cimiotti (Bronze) Date 2009-10-19 07:56 Edited 2009-10-19 07:59
:lol: Nobody knows? :lol: Nobody will tell you ;)
Parent - By Roland Rösler (****) Date 2009-10-19 07:59
Okay, there is at least one! :-)
But this one don´t tell us anything! :-(
Parent - By Highendman (****) Date 2009-10-19 08:22
lol - good one to start the day with.
Parent - - By Permanent Brain (*****) Date 2009-10-19 04:24
Is it safe to say that Cluster-Sjeng is stronger than e.g. Deep Shredder 12 or Naum 4 on an i7 quad? I really don't know but I wouldn't bet on it.

Anyway, except Rybka and Deep Sjeng all top engines were missing (unfortunately, The King is not a top engine anymore nowadays), including some which did participate in previous years, in the Dutch Open CCC. I think that is what staylor's question points to, and it can also explain why this tournament doesn't raise much interest on other computer chess boards.

Nevertheless, congratulations! :grin: Another trophy for the growing collection...
Parent - - By GCP (***) Date 2009-10-19 06:15

>I really don't know but I wouldn't bet on it.


Ah, but would you bet on the reverse?

>Anyway, except Rybka and Deep Sjeng all top engines were missing (unfortunately, The King is not a top engine anymore nowadays), including some
>which did participate in previous years, in the Dutch Open CCC


FYI, Hiarcs withdrew when Cluster Rybka and Cluster Sjeng entered.

If the other "top engines" don't play, we can't beat them. Tried and tested strategy from the Fritz team :)
Parent - By Ray (****) Date 2009-10-19 06:18

> FYI, Hiarcs withdrew when Cluster Rybka and Cluster Sjeng entered.


A pity, but with a new release rumoured to be imminent, they wouldn't want a bad result, so perfectly understandable. Hopefully they will be back next time, maybe with a cluster too
Parent - - By Permanent Brain (*****) Date 2009-10-19 07:34
I also wouldn't bet on the reverse, it is unclear. It would require tests like CCRL or CEGT do.

> FYI, Hiarcs withdrew when Cluster Rybka and Cluster Sjeng entered.


That is exactly the public relations problem which you, and Rybka has. Some time ago I tried to explain here that cluster hardware achievements are meaningless. But I have to accept that certain people don't get this, or chose to ignore it as long as possible. I have to sit on the shore of the river and just wait... :lol:
Parent - - By George Tsavdaris (****) Date 2009-10-19 08:16

>That is exactly the public relations problem which you, and Rybka has. Some time ago I tried to explain here that cluster hardware >achievements are meaningless.


Where did you explain that? And what do you mean?
Parent - By Permanent Brain (*****) Date 2009-10-20 04:41
Just that.

Get it or not.
Parent - - By GCP (***) Date 2009-10-19 08:21

>That is exactly the public relations problem which you, and Rybka has. Some time ago I tried to explain here that cluster hardware achievements
>are meaningless.


I'm sorry but there is nothing meaningless in wiping out every tournament you play in, nor is it a public relations problem that some of the opponents are becoming too afraid to show up.

Clustering support is for a chess program now what SMP support was in 2002. A feature for high end users that will become inevitable in order to be taken seriously soon.
Parent - By Permanent Brain (*****) Date 2009-10-20 04:40
Dream on.
Parent - By Uly (Gold) Date 2009-10-19 09:01
By that logic only the rating lists are meaningful.
Parent - - By Uri Blass (*****) Date 2009-10-19 11:58
I agree with GCP and there is nothing meaningless in the results.

If top programs like Shredder and Naum do not compete then my guess is that they are probably weaker.
Parent - By Permanent Brain (*****) Date 2009-10-20 04:42

> I agree


That's the perfect proof that I am right.
Parent - By Permanent Brain (*****) Date 2009-10-20 04:44
Send in the clowns... NO! :lol: They are already here.
Parent - - By George Tsavdaris (****) Date 2009-10-19 08:14

>Cluster Sjeng


First on how many cores it run? 72 if i recall correctly?

And can you make any guesses about the speedups it has on different number of cores?
For example from an i7 quad what is the speedup and what is from an octal?
Also i guess you've used different number of cores on the cluster. What was the speedup for them on example 72 cores Vs 40 cores or 72 cores Vs 240 cores?

And the same question to Lukas also.

I remember a post of someone suggesting with numbers, that a fairly good approximation for Rybka's efficiency for up to 16 cores was given by:
Sp_Ef(n) = (20-3·log2(n)) · n / 20
Where Sp_Ef(n) is the speed efficiency for n cores.
That gives us the results:
     n       Sp_Ef(n)
     1         1
     2         1.7
     4         2.8
     6         3.7
     8         4.4
    16         6.4


Any such formulas for Sjeng and Rybka cluster for up to 240(or 72) cores?
Parent - - By GCP (***) Date 2009-10-19 08:51 Edited 2009-10-19 08:55
We're comparing local version vs clustered version and measuring the improvement in ELO, and then playing with whatever setup maximizes the latter. This is not always the max amount of cores.

At least the following influences the performance (in rough order of importance):

- Master CPU speed
- Amount of CPUs on the master
- Used settings on master
- Timecontrol of the tournament
- Number of clients
- Interconnect of the PCs
- Speed of the client CPUs
- Amount of CPUs per client

Putting that in a formula is hopeless. Which is why we only look at the test result as described above: WYSIWYG.

Even for SMP your formula breaks down rapidly. A 16 core Core 2 machine will scale vastly different from a 16 core Nehalem.
Parent - - By Roland Rösler (****) Date 2009-10-19 09:03

>We're comparing local version vs clustered version and measuring the improvement in ELO


And what improvement you measured here? More than 150 Elo, more than 200 Elo?
Btw: What was you master? Nehalem EP (W5590)?
Parent - - By GCP (***) Date 2009-10-19 09:19

>And what improvement you measured here? More than 150 Elo, more than 200 Elo?


"Enough to be happy and not enough to having nothing more to do" :)

>Btw: What was you master? Nehalem EP (W5590)?


4x6 core 2.4Ghz AMD

Seems faster than the Nehalems.
Parent - - By Roland Rösler (****) Date 2009-10-19 09:56
Wow, a 24-core system! :-)
What´s the real speed up for Deep Sjeng on this system (from 1 core to 24 cores)? When current Deep Sjeng 1 core plays with tc 40/40, what tc would you choose for DS on this system to be rather equal? 40/5 or 40/4?
Parent - By GCP (***) Date 2009-10-19 10:59
I don't know, haven't run such a test yet.
Parent - - By Watchman (***) Date 2009-10-20 01:37 Edited 2009-10-21 19:51
[DELETED] (offensive post)
Parent - - By GCP (***) Date 2009-10-20 08:40

>Aha.... just like Sjeng 6 months ago due to the 8 core limit.


Yeah, that has kept us away as we clearly can't compete with Hiarcs in those conditions. Oh wait, nevermind reality:

http://www.grappa.univ-lille3.fr/icga/tournament.php?id=192
http://www.taccl.org/ACCAWCRCC/2009ACCAWCRCC/WCRCCResults.html (was 8 core Nehalem, not cluster)

>u nut.


Woo, we are getting to that level.

>Harvey was busy in Manchester with his apartment.


The tournament organisers got another version of the story.
Parent - By Vempele (Silver) Date 2009-10-20 09:00

> http://www.taccl.org/ACCAWCRCC/2009ACCAWCRCC/WCRCCResults.html (was 8 core Nehalem, not cluster)


Heh, the way they present it makes it look like almost everybody used a cluster, especially Equinox. :)
Parent - - By Watchman (***) Date 2009-10-20 12:13 Edited 2009-10-21 19:53

>>>nor is it a public relations problem that some of the opponents are becoming too afraid to show up
>>Aha.... just like Sjeng 6 months ago due to the 8 core limit.
>Yeah, that has kept us away as we clearly can't compete with Hiarcs in those conditions. Oh wait, nevermind reality:


Well... it's just as stupid as implying Hiarcs Team was afraid to compete.

>>u nut.
>Woo, we are getting to that level.


A term I use in an affectionate sense (that I regularly use with friends).  If it offends you... I withdraw it and sincerely apologize.

>>Harvey was busy in Manchester with his apartment.
>The tournament organisers got another version of the story.


From whom?  You?  Seems so, as I have "another version," but this one from the Chairman of The CSVN, Cock de Gorter:

"Harvey,

Your message from a couple of months ago was regretted very much by our organisers. We always appreciate Hiarcs as a competiror.

I read a message on the net that Hiarcs subscribed and later withdrew.
That is not true. Hiarcs was never on our list of competitors for the October 2009 tournament.

Moving from London to Manchester looks a reasonable excuse for not being in Leiden. You had a whole lot of other things on your mind then computerchess, did'nt you?

kind regards
Cock"
Parent - - By GCP (***) Date 2009-10-20 12:40

>From whom?  You?  Seems so, as I have "another version," but this one from the Chairman of The CSVN, Cock de Gorter:


Thanks, this is interesting. And no, the other version didn't came from me.

>Please Gian... let's grow up and at least act mature... enough with this petty sniping.


I have no advice to take from you about what acting mature means.
Parent - - By Felix Kling (Gold) Date 2009-10-21 19:54
I sent him a warning. Of course it's good to get a clarification if there is a rumor, but I strongly dislike the way "watchman" did it.
Parent - - By Watchman (***) Date 2009-10-22 08:51

>Of course it's good to get a clarification if there is a rumor, but I strongly dislike the way "watchman" did it.


When I was in the Marines, we had a saying for something like that Felix...

"Sounds like a personal problem"

i.e. "too bad..." "who cares..." "don't want to hear it..."  "so what, get lost"

I have absolutely no desire to please you Felix.  Gian posts what is essentially a lie... and that is ok by you.  I post something as evidence to the contrary, and "you don't like that."  Well too bad.  Now I suppose I should submit all my posts in advance to you and each night bring you some milk and cookies, read you a bedtime story, and hope you approve them all.
Parent - - By Banned for Life (Gold) Date 2009-10-22 15:10
"Sounds like a personal problem"

Thanks Rob!!! I've heard this from Marines from time to time, and never really understood it. :-)

Now that Harvey is settled in, I hope to see Hiarcs competing in future tournaments, maybe even in a cluster format.
Parent - By Watchman (***) Date 2009-10-22 18:12
Sir, you are welcome sir... :-P

It doesn't take long for someone who hears that several times to realize how petty or stupid they are being... when used in the "serious" sense.  Although with friends it was also used to tease. 

(btw, I think pretty much ixnay on the cluster-ay)
Parent - By Felix Kling (Gold) Date 2009-10-22 17:00
I said I don't like the way you did that. There was no reason to get offensive, a short note that this is not true would have been enough. Let's be nice to each other and assume that anyone just wants to be nice, too :)
Parent - - By Watchman (***) Date 2009-10-22 08:50
To the so-called "Moderator" who deleted my post and issued me this warning:

"A moderator edited a post. Reason: It's not ok to offend people, espacially  Gian-Carlo or Lukas in our forum. This is a final warning, next would be serius consequences (ban). You have the right to write anything, but only as long as it's expressed in a civilized way."

Hypocrite!

How many nasty things have been said to Harvey & myself (by you yourself Felix!), derogatory comments... provocative things meant to incite... innuendo... petty, immature crap... all because we are somehow responsible for a rule Lukas or you don't like???  And because Felix you're a Moderator, that allows you this abuse of your position???  If you were a member of Hiarcs Team and said on our Forum the things you have said here, someone by now would have turned you over on their knee!

I have never made "personal insults" here at Rybka Forum.  Never once warned about any post until this date... never once said to me, "hey... what you said was wrong."  No sir! Nothing! No... not even a "This is your First Warning."!

If I have said something that was "wrong" or violated Rybka Forum policy... then testify to that wrongdoing.  But it rubs you the wrong way and you censor my words?!  But if I say something that is TRUE! and it bruises someone's sensitive feelings... that is worthy of having a post deleted and told next time I will be banned???

Ok then... if you feel froggy... then leap.

What now I have to walk on egg-shells for the rest of my natural life, worried I  may say something that "offends" someone (what the heck is that supposed to mean anyway?) and will result in a RF ban?  You got to be kidding me.  Again... talk about  pettiness...

Someone has an axe to grind... thinks they have  something on me and is acting in a wildishly abusive and immature way.  "Someone"  needs to grow up.

I would appeal to the Senior Moderators to review what was deleted and determine for themselves if it was worthy of a "Final Warning... next would be... ban."  And I would appreciate not being censored (i.e. having my post replaced) because it rubs an over-zealous moderator the wrong way.  But if you feel that post is worthy of me being banned... I will accept your judgment.  No need to "wait until next time... if it's worthy of a ban then that is what I should receive now.  Thank you.
Parent - - By Felix Kling (Gold) Date 2009-10-22 17:13
You don't understand our concept, it's:
-Warning
-Short ban
-Permanent ban.

You got level 1 by offensive comments, maybe you get level 2 by continuing and not understanding that this is a place where we don't want abusive posts.
It doesn't matter what the reasons for your offensive posts are, if you can't control your feelings, this is not the right place for you.
Up Topic Rybka Support & Discussion / Rybka Discussion / Dutch Open 2009: Rybka wins!
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill