Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Dutch Open: Rybka wins with perfect 9/9 score
1 2 Previous Next  
- - By Venator (Silver) [nl] Date 2008-11-16 17:16 Edited 2008-11-16 17:27
Details: CSVN
Parent - - By Abu CT (*) [nl] Date 2008-11-16 17:20
On 40 cores?? Next time with equal hardware!!
Parent - - By Felix Kling (Gold) [de] Date 2008-11-16 17:29
why? those tournaments are not to show the engine's strength (there are rating lists for that) but to have fun and to try to have the best engine+opening book+hardware combination.
Parent - - By Venator (Silver) [nl] Date 2008-11-16 20:16
The discussion about equal hardware is a bit odd. If the organisers have no specific rules about hardware, you can bring anything you like.

Besides, we never had these discussions in the days of Cray Blitz, Mephisto, Deep Blue,  or even Hydra.

Note that a hardware limitation might imply that a new Hydra or a program running on a non-MS machine (i.e. a mainframe) most likely cannot participate.
Parent - By Lukas Cimiotti (Bronze) [de] Date 2008-11-16 21:43
Thanks, Jeroen.
In fact I'd love to play against Hydra or any other supercomputer - I hope we'll get this chance.
Your book preparations were superb - as usual. In the last game we were out of book early with a slightly negative eval - then suddenly my CPUs showed 0% usage - i was worried and asked Hans. He told me we were back in book again - and when we came out we were clearly better :)

The game I liked most was the game against Zunami - I loved that trap for Tsunami's queen. I'm no OTB player, but this was clearly foreseeable :)

Regards,
Lukas
Parent - By Ernst (***) [nl] Date 2008-11-16 22:43
Furthermore, running a cluster with off the shelf hardware started with cluster toga, and there ain't no stopping it now.
Parent - By Carl Bicknell (*****) [gb] Date 2008-11-16 23:38

> The discussion about equal hardware is a bit odd. If the organisers have no specific rules about hardware, you can bring anything you like.
>
> Besides, we never had these discussions in the days of Cray Blitz, Mephisto, Deep Blue,  or even Hydra.
>
> Note that a hardware limitation might imply that a new Hydra or a program running on a non-MS machine (i.e. a mainframe) most likely cannot participate.


100% agree with you. Why do so many people have a hangup over hardware?? You want even-Stevens? Go check out CEGT...
Parent - - By Quapsel (****) [de] Date 2008-11-17 16:08
That's right.
And you did right, and I'm interested in such events.

But the thread should have had the topic:
Rybka on 40 cores and with openingbook XYZ wins with perfect 9/9 score.

To use a very strong combination, and then to declare the engine as the winner, is  rather amazing.

Quap
Parent - By Felix Kling (Gold) [de] Date 2008-11-17 16:29
yes, sure. but people like to mix those things as it's better for marketing :)
Parent - - By George Tsavdaris (****) Date 2008-11-17 16:48
Actually this would be misleading also, since it will not specified exactly what these 40 cores were.
Was it with 40  PIII 400 MHz or maybe it was with 40 AMD K5 processors,...., etc?

But even then it would still be misleading since it will not specify what hash tables memory has been used, and what tablebases have been used and what exact hard disc drive has been used.
Parent - - By Uly (Gold) [mx] Date 2008-11-17 20:40
Don't forget it was private Rybka 3i8 ;)
Parent - - By George Tsavdaris (****) Date 2008-11-17 22:15
Yes indeed, so the title should be something like:

Dutch Open: Rybka 3i8 + 40 cores(Skulltrail 3.8 GHz+Skulltrail 4.2 GHz+Xeon 2x X5460 @ 3.4 GHz+Xeon 2x X5365 3 GHz+
Xeon 2x E5430 @ 3 GHz) + 40 GB DDR2 RAM hash tables + Seagate 500 GB Serial ATA Barracuda 7200 with all 5,4 tablebases + Rybka 3beta book, wins with perfect 9/9 score.

Nothing less than that for a title is misleading as it is the whole combination that won and not only Rybka.
Perhaps we could even add the hamburgers and beefsteaks and all the food the operator of Rybka ate since if he didn't eat them, he wouldn't be able open the PC for Rybka to run.
Parent - By Harvey Williamson (*****) Date 2008-11-17 22:26
I doubt it was the R3 book and as you mention the op. the Rubik's cube should be mentioned as Hans is known to use this all over the world to raise money. here is a picture I took in China:



and here is Hans talking to FM Vincent Diepeveen about his cube during our game
Parent - By Lukas Cimiotti (Bronze) [de] Date 2008-11-17 23:25
Not quite - the hardware changed a bit: the big Skulltrail had only 4 GHz, the X5460 now had 3.8 GHz, the Xeon 5430 now 3.5 GHz, the X5365 replaced by a computer running one Xeon E5430 and one E5420, both running at 3 GHz, only 5x2 GB hash and of course no slow Seagate Barracuda for tablebases but solid state disks in all computers for that purpose.

Regards,
Lukas
Parent - By Uly (Gold) [mx] Date 2008-11-17 23:53

> Dutch Open: Rybka 3i8 + 40 cores(Skulltrail 3.8 GHz+Skulltrail 4.2 GHz+Xeon 2x X5460 @ 3.4 GHz+Xeon 2x X5365 3 GHz+
> Xeon 2x E5430 @ 3 GHz) + 40 GB DDR2 RAM hash tables + Seagate 500 GB Serial ATA Barracuda 7200 with all 5,4 tablebases + Rybka 3beta book, wins with perfect 9/9 score.


I tried editing the title but it's cut at "Dutch Open: Rybka 3i8 + 40 cores(Skulltrail 3.8 GHz+Skulltra" :(

:-D
Parent - By Quapsel (****) [de] Date 2008-11-19 10:00

> Yes indeed, so the title should be something like:
> Dutch Open: Rybka 3i8 + 40 cores(Skulltrail 3.8 GHz+Skulltrail 4.2 GHz+Xeon 2x X5460 @ 3.4 GHz+Xeon 2x X5365 3 GHz+
> Xeon 2x E5430 @ 3 GHz) + 40 GB DDR2 RAM hash tables + Seagate 500 GB Serial ATA Barracuda 7200 with all 5,4 tablebases + Rybka 3beta book, wins with perfect 9/9 score.


I would have written:
"Rybka on 40 cores wins with perfect 9/9 score."
I wouldn't have mentioned the book because it was the Rybka3-book
(But I would have mentioned each book, which is not associated with Rybka)

How do you think:
If Rybka had won playing exclusivly on one of the machines on
http://de.wikipedia.org/wiki/Supercomputer, (IBM-Roadrunner looks nice, perhaps VAS can make a Compile for that machine)
should the hardware have been told of as an important information?
Or should the user be satisfied by hearing "Wow, Rybka had won again!"
;-)

Quap
Parent - By Quapsel (****) [de] Date 2008-11-19 09:49

> Actually this would be misleading also, since it will not specified exactly what these 40 cores were.
> Was it with 40  PIII 400 MHz or maybe it was with 40 AMD K5 processors,...., etc?


:-D
But it would show you dirctly, how that hardware is to be compared to the 'normal' machines.
It can be assumed that those important factors, which aren't mentioned (say: Processor Speed), are similar.

Quap
Parent - - By Vasik Rajlich (Silver) [pl] Date 2008-11-18 08:49

> To use a very strong combination, and then to declare the engine as the winner, is  rather amazing.


Rybka refers to everything - the engine, the book and the cluster. Rule of thumb: if it's good, it's Rybka :)

Vas
Parent - - By George Tsavdaris (****) Date 2008-11-18 10:48 Edited 2008-11-18 10:51

>Rybka refers to everything - the engine, the book and the cluster. Rule of thumb: if it's good, it's Rybka :-)


Actually it's the other way around. The direction of the implication is not It's good => It's Rybka, since then you imply that there is not other good program except Rybka(for example: Suppose Hiarcs is good, then since it's good we imply that it is Rybka, so you say Hiarcs is Rybka. A contradiction so Hiarcs is not good).

So you should have said if it's Rybka, it's good.
Parent - By Vasik Rajlich (Silver) [pl] Date 2008-11-22 07:33
Nah - logic is overrated!

Vas
Parent - - By Quapsel (****) [de] Date 2008-11-19 09:43

> Rybka refers to everything - the engine, the book and the cluster.


OK, but this would result in:
"Rybka ist too expensive, I'm not able to afford this price.
I can only buy the engine. (What's her name?)"


Quap
Parent - By Vasik Rajlich (Silver) [pl] Date 2008-11-22 07:34
Rybka is a name which for now refers to three different things: Rybka the engine (commercial), Rybka the opening book (commercial), Rybka the cluster (private).

Without names everything would be really tedious.

Vas
Parent - - By Roland Rösler (****) [de] Date 2008-11-16 17:30
And without opening book and without EGTB and a drunken operator! :-)
Parent - - By Abu CT (*) [nl] Date 2008-11-16 17:33
Lol. Just like F1 championship: 9 cars, 1 with a Ferrari engine; 8 with a Lada engine. Champagne for the winner!
Parent - - By Felix Kling (Gold) [de] Date 2008-11-16 17:52
we have the strongest driver, the strongest car and the best knowledge. That's what makes a team streong. If other teams are weaker, should we start driving worse to give them a chance? :)
Parent - - By keoki010 (Silver) [us] Date 2008-11-16 21:33
Nah! They probably want you to tie your hands behind your back, kneel down and then if they can't beat you give them the game regardless.  Nothing but a bunch of cry babies.
Parent - - By zwegner (***) Date 2008-11-17 02:18
Where has a competitor actually said anything negative about Rybka using such hardware?
Parent - By Graham Banks (*****) [nz] Date 2008-11-17 03:20
Try another forum.   ;-)
Parent - - By Vempele (Silver) [fi] Date 2008-11-16 18:13
Yeah. The results are just meaningless when they aren't using equal engines.

Next year's list of participants:

rybka
Rybka
rYbka
ryBka
...
Parent - - By turbojuice1122 (Gold) [us] Date 2008-11-16 18:18
...
Fruit
:-) :-P :-D
Parent - By Vempele (Silver) [fi] Date 2008-11-16 18:19
No, Chess64.

The '64' referring to the number of nodes in the cluster they're going to use. :-p
Parent - By Ray (****) Date 2008-11-16 19:14

> Lol. Just like F1 championship: 9 cars, 1 with a Ferrari engine; 8 with a Lada engine. Champagne for the winner!


Very bad analogy, actually 6 cars in F1 had a Ferrari engine - the 2 Ferraris, the 2 Toro Rossos and the 2 Force Indias. The Ferrari team comfortably won the team title with 172 points, Force India scored ZERO points
So the Ferrari engine - which you are likening to 40 core Rybka - didn't help Force India now did it...
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2008-11-16 21:51
HIARCS used a Skulltrail - is that a Lada? What would be Sjengs's quad then? A Trabant ?
In case you want to find out about the hardware used, here's a link: http://www.csvn.nl/index.php?option=com_content&task=view&id=322&Itemid=1
Should we reduce hardware to the level of the weakest ?
In WCCC in China that would have been a mobile phone.
Parent - - By Roland Rösler (****) [de] Date 2008-11-16 22:06
No bad words about Lada. The Russians hear it! :-)
Parent - By Lukas Cimiotti (Bronze) [de] Date 2008-11-16 23:05
Hehe, my words say a Lada is better than a Trabant. So my words were good ;)
Parent - - By Banned for Life (Gold) Date 2008-11-16 19:43
OK, but we already had the drunken operator.
Parent - By Roland Rösler (****) [de] Date 2008-11-16 20:14
Ohh, I missed something!? Please, help me!
Parent - - By Venator (Silver) [nl] Date 2008-11-16 19:45
There is no way you are going to force Hans to drink any kind of liquor/alcohol, so in that case we need another operator ;-)
Parent - By Roland Rösler (****) [de] Date 2008-11-16 20:10
Okay, I see. Miscast! :-)
Parent - - By turbojuice1122 (Gold) [us] Date 2008-11-16 17:59
If you're interested in equal hardware, equal books, equal this, equal that, there are CEGT, CCRL, and many other testing sites with rating lists.
Parent - - By M ANSARI (*****) [kw] Date 2008-11-16 18:21
Rybka's play was perfect in all the games I saw.  It was really very enjoyable.  I will say that I think that Rybka 3 without 40 cores would have had the same result.  On my Skulltrail almost every move that was chosen by Rybka was also the best move on my computer.  There was one move that I thought was very strong in the last game against Joker which was bishop retreat protecting a possible bishop sac by white which could have made things a little messy.   I am not sure if that was a book move or if Rybka found it alone.  I will have to check again to see if that would be a best move on my system again, but if I remember correctly on my system was going to play c5 and allow the Bxp+ shot.

Overall the play was extremely clean and precise with fanatastic endgame play ... including the handling of a potential difficult bishop endgame which was handled perfectly.  Seems like the little bugs in Rybka 3 have been ironed out and look forward to getting it soon.  Congrats to all the Rybka team for a nice clinic of power chess.
Parent - - By Roland Rösler (****) [de] Date 2008-11-16 20:50
I will say that I think that Rybka 3 without 40 cores would have had the same result.  On my Skulltrail almost every move that was chosen by Rybka was also the best move on my computer.

The real problem with your proclamation is, that I also could produce most of the move on my C2D w32. On 40 cores you are more sure to make the right choice than on 8 cores and further on. Be aware: If you have a real competition, one blunder (not best move) is nearly loss! In Leiden, Rybka is so far above, that she can win three times in a game.

PS: I also noticed Rybkas good endgame play in Leiden. We have two explanations for this. First: Depth on 40 cores is good enough for good endgame play (with some Monte Carlo removals). Or second: Vas did something for better endgame play in source (mp EGTB bug, bad bishop, maybe something in rook endgames). My guess: He made both! :-)
Parent - - By Banned for Life (Gold) Date 2008-11-16 21:14
I will say that I think that Rybka 3 without 40 cores would have had the same result.  On my Skulltrail almost every move that was chosen by Rybka was also the best move on my computer.

The other explanation for this is that Dr. Hyatt is right and the method Vas chose for implementing the cluster doesn't offer much of a speedup.

PS: I also noticed Rybkas good endgame play in Leiden. We have two explanations for this. First: Depth on 40 cores is good enough for good endgame play (with some Monte Carlo removals). Or second: Vas did something for better endgame play in source (mp EGTB bug, bad bishop, maybe something in rook endgames).

More likely is that the improvement in the endgame was due to greater tree width rather than depth.
Parent - By Roland Rösler (****) [de] Date 2008-11-16 21:26
More likely is that the improvement in the endgame was due to greater tree width rather than depth.

For me it´s the same like this: Depth on 40 cores is good enough for good endgame play (with some Monte Carlo removals).

PS: I see, you don´t believe, that Vas has made any bug update. I´m not sure. But you can be right. Vas has his own preferences. And history tells us, he wasn´t always wrong! :-)
Parent - - By Roland Rösler (****) [de] Date 2008-11-16 18:16
In the future we will play ecological tournaments; one Watt per game for each engine. I´m sure CC will become famous.
And in human vs. engine matches, we compare Joule vs Watt; energy consumption must be equal! :-)
Parent - - By Abu CT (*) [nl] Date 2008-11-16 19:46
In the future we will play ecological tournaments; one Watt per game for each engine. I´m sure CC will become famous.
And in human vs. engine matches, we compare Joule vs Watt; energy consumption must be equal! 

Sure Roland. In human chess match: one player gets 2 hours, the other 8 hours! Energy consumption must be equal!!
Have a nice/day/week
Parent - By Roland Rösler (****) [de] Date 2008-11-16 20:24
I´m not sure about relation 2 to 8. But I´m sure, the older (fatter; less basal metabolism) will become more competitive.
Parent - - By Banned for Life (Gold) Date 2008-11-16 19:47
Wrong! Each entry will have to bring its own power source. The other teams will show up with small solar panels. The Rybka team will build a solar farm in the parking lot...
Parent - - By Roland Rösler (****) [de] Date 2008-11-16 20:08
Wrong! It´s survival of the fittest! They have to move the dynamos of their bicycles.

PS: Maybe the smartest will survive (in chess). But I don´t like this idea.
Parent - - By Banned for Life (Gold) Date 2008-11-16 20:26
If they go to powering the engines with human powered dynamos, Hans will be replaced by Lance and a large supply of EPO (unless there will be drug testing for the operators). :-)
Up Topic Rybka Support & Discussion / Rybka Discussion / Dutch Open: Rybka wins with perfect 9/9 score
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill