Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka Cluster Rental Program
1 2 3 4 5 6 7 Previous Next  
Parent - By mdraith (**) [de] Date 2010-12-31 03:48
What is unlikely in depth 20 analysis might be a genius move in depth 30 analysis.
Parent - - By nebulus (****) [no] Date 2010-12-31 03:56
The number is just soooooooooo huge :yell:.

Okay, how do you propose you extract good positions from the complete set? Where are you going to store the results even if you could? You are way too optimistic :smile:.
Parent - - By Felix Kling (Gold) [de] Date 2010-12-31 04:05
imagine you have something 1000000 times faster than the current cluster. Would that be close to solving chess? Who says that the results need to be stored? The positions where the computers are wrong could be stored, but usually there is no need for that.
Parent - - By mdraith (**) [de] Date 2010-12-31 04:22
The GHz rally has ended, the chipmakers have reached physical boundaries in shrinking their chips. So they decided to put more cores on a die but this also comes to an end because the chips will be too big and will get too hot. They want to do some nano / quantum / light CPU's but this is not realistic for the near future. So in 30 years we might have 1024 cores in one CPU but it will not run at 1.000.000 GHz, dream on.
Parent - By Felix Kling (Gold) [de] Date 2010-12-31 13:59
I talk about huge clusters. Is it really unrealistic to build clusters of some hundred million cores in 30 years? I don't think so.
Parent - - By patrick delaurentis (**) Date 2012-08-19 16:31
i live in the usa, how much money would it cost me in american dollars for full 48 hour rental of the cluster?
Parent - By Barnard (Bronze) Date 2012-08-19 20:38
A Rybka Cluster 40 consists of 40 physical cores and can be rented for 238 Euro (incl. VAT) per day, or 200 Euro (without VAT) per day for customers outside the EU or businesses outside Germany. The minimum continuous Rybka Cluster 40 rental time is 5 days.

A Rybka Cluster 100 consists of 100 physical cores and can be rented for 595 Euro (incl. VAT) per day, or 500 Euro (without VAT) per day for customers outside the EU or businesses outside Germany. The minimum continuous Rybka Cluster 100 rental time is 2 days.

from the pure money point of view,is the same money rent the cluster 40 cores x 5 days,than 100 cores x 2 days

1000 euros = 1235 dollars

so the cost will be 1235 dollars

p.s.that is just what they have at their page;if they have up to date the page or not,i dont know it

http://cluster.rybkachess.com/rent.html
Parent - By Felix Kling (Gold) [de] Date 2012-08-19 22:25
just contact lukas and ask him what it costs. There's the possibility to make individual offers.
Parent - - By ChessMonster (***) Date 2010-12-31 03:41
I am a computer scientist just like Vas, and yea I can see the end is coming :-) It is not scary, it is good. More brainpower will be spent on productive things. 30 years is a good estimation. Regards,
Parent - By Androcles (**) [fr] Date 2010-12-31 11:34
If quantum computation:cool: actually comes out of the laboratories, it may be.
Parent - - By ChessMonster (***) Date 2010-12-31 03:55 Edited 2010-12-31 04:00
You will be surprised to know that Jiri's book even has some theory vs. the 1.a4.... and rest of the weaker moves you might throw.

Yes the theory on those lines is less, because it eventually leads to a position where Rybka is giving something like -0.50...

Yes, this -0.50 might be not sufficent to win for Black, but it is obvious that white did not play his best. The best lines of white & black will be exhausted faster than these type of sidelines. But then the sidelines will start getting exhausted also.

I was part of a team that wrote a loser chess engine, so I know about the staff you talk about. But with Aquarium, you can cover much wider than you think. I feel Aquarium concept is not fully understood by the chess community yet. It is not a chess engine, it is a tool that covers the imperfections of a chess engine.

You are also overestimating the capabilities of human brain in chess. For most of the positions that happen in a chess match, computers are more trustworthy than SGMs. SGMs have a better understanding of the drawing tendency of a position, which the engines lack. and SGMs also have superb endgame strategy which engines again would be lacking. In those positions, yes, I trust the SGMs more. I also trust SGMs more in opening preparation, but few would be willing to share their hard work for free.

Anyway, I am tired from posting on this thread. I'll be silent now so I can focus on life.

Regards to all,
Parent - By Felix Kling (Gold) [de] Date 2010-12-31 04:15
I'll go to sleep now. it's always good to exchange thoughts, even if there's no immediate result :smile:
Parent - By Geomusic (*****) Date 2010-12-31 05:25
not really surprised Chessmonster as I've seen a lot of that garbage in (C) tournaments for those who attempt to avoid books completely.
Parent - - By ChessMonster (***) Date 2010-12-31 02:57
To solve chess, you don't need 32-men tablebase. A pretty good solution is enough for the highest human play. Openings will be perfected within several years if you ask me due to cluster-prepared opening books. This would leave fixing engine endgame play. Yes, it won't be easy, but with many authors working on endgame skills of their engines, we will see huge improvements. I just see there is getting less and less room for uncertainty in chess. With so much uncertainty gone, I would think that chess is almost solved.

I don't know I am thinking to teach my children programming instead of chess to be honest :-) Regards,
Parent - By mdraith (**) [de] Date 2010-12-31 03:01
I don't agree on this. If some engine says move A is best, another engine thinks that move B is better. And what do the GM's think of it? They just don't have a clue because they can't think ahead 20 moves. Modern chess programs have selectivity and hash tables build in, so there is enough room for errors.
Parent - - By Felix Kling (Gold) [de] Date 2010-12-31 03:11
Actually wasting time on chess is pretty stupid in general. It doesn't even help you in math or something like that. If someone is good in math and plays chess, he would be even better in math if he would have spent the time he spent on chess on math :)

If I would have done as much for chemistry as I did for chess, ... :)

I always say that you can see in chess if a child is really clever - the intelligent one learns quickly, becomes a strong player. The super intelligent children realise it's a waste of time ;)

I started teaching some simple chemistry, physics or math stuff at my chess club. Sometimes I don't want to play chess anyway :)
Parent - By ChessMonster (***) Date 2010-12-31 03:16
ah :-/ I wished I was smarter when I was younger :-) but hopefully our children will be guided better thanks to our mistakes :-)
Parent - By Geomusic (*****) Date 2010-12-31 05:45
"openings will be perfected"?
"chess almost solved"?
chess is about as close to being solved, and openings being perfect as I am a millionaire.
Parent - By ChessMonster (***) Date 2010-12-31 02:45
True, human memory won't remember it all.

Also, I think endgame knowledge is getting more and more important these days, because almost everybody has pretty good openings thanks to these powerful opening books, and you cannot win as quickly as in the old times. So in human play, I think the superior technique in endgame will the main factor to decide the outcome.
Parent - By tomgdrums (****) Date 2010-12-31 01:44
Didn't Topalov use a cluster type Rybka and Jiri?
Parent - - By mdraith (**) [de] Date 2010-12-31 00:33 Edited 2010-12-31 00:35

> The Rybka 4 book (and R3) is a one off deal with no updates. The Hiarcs book gets updated 3-4 times a year. The Junior book got updated with no extra cost.


You cannot compare the HIARCS book with the Rybka 4 book. The Rybka book has ~18.000.000 positions, whereas the HIARCS book has ~3.000.000 positions. The HIARCS book is optimized for engine tournaments, whereas the Rybka book is an interpretation of the actual opening theory by Jiri Dufek. If you want to play engine games, killerbooks like the HIARCS book is your choice. If you want to study and learn opening theory, the Rybka 4 book is the way to go.

> Rybka has no way to use it's book internally in GUIs other than Aquarium or Chessbase.. Harcs and Junior (junior to a lesser extent) do have a way to use their book internally.


The bad thing about internal engine books is that you cannot access it from the GUI. And every GUI has it's own proprietary format which is a pain in the ass.

> All of that adds up to fairly bad customer service.


It's up to the programmer and the distributors, if and when updates will be available. It's up to the customer to choose the engine which fits best for him and customer service is one point to be considered.
Parent - - By tomgdrums (****) Date 2010-12-31 00:39

>> The Rybka 4 book (and R3) is a one off deal with no updates. The Hiarcs book gets updated 3-4 times a year. The Junior book got updated with no extra cost.
> You cannot compare the HIARCS book with the Rybka 4 book. The Rybka book has ~18.000.000 positions, whereas the HIARCS book has ~3.000.000 positions. The HIARCS book is optimized for engine tournaments, whereas the Rybka book is an interpretation of the actual opening theory by Jiri Dufek. If you want to play engine games, killerbooks like the HIARCS book is your choice. If you want to study and learn openeing theory, the Rybka book is the way to go.


I disagree.  The Hiarcs book has proven to be great for my study.  As has the Shredder book.  (the Junior book a little less so)   Is Jiri's book really all that much better?

>> Rybka has no way to use it's book internally in GUIs other than Aquarium or Chessbase.. Harcs and Junior (junior to a lesser extent) do have a way to use their book internally.
> The bad thing about internal engine books is that you cannot access it from the GUI. And every GUI has it's own proprietary format which is a pain in the ass.


I agree with this!  Which is why I love Hiarcs solution!  Make it available as a ctg book for viewing in certain GUIs and to be used internally in other GUIs that don't read ctg.  See?  Good customer service!  They solved a problem.

> It's up to the programmer and the distributors, if and when updates will be available. It's up to the customer to choose the engine which fits best for him and customer service is one point to be considered.


You are basically correct.  Which is why it is harder and harder to want to buy Rybka products.  A customer does buy a product expecting it to work as advertised.  When it doesn't he or she usually gets their money back or the provider/manufacturer fixes the product.  Programmers are not somehow above that standard.
Parent - - By mdraith (**) [de] Date 2010-12-31 00:49

> Is Jiri's book really all that much better?


It's much wider (don't know if it's deeper) and you can use it for non engine games. For the HIARCS book, they add games played on playchess which is a very bad thing I think. Engines have no understanding of opening theory and adding such games is a very incompetent step when compiling an opening book. It will be good when playing engine games, but it will fail badly against human players.
Parent - - By Harvey Williamson (*****) Date 2010-12-31 00:57
so we add playches games and that is bad. are you going to claim there are none in the very good Rybka book?!
Parent - - By mdraith (**) [de] Date 2010-12-31 01:02
As I said, it's good for engine games but it's a bad idea for games against human players. I don't know if there are any playchess games in the Rybka 4 book, but Jiri said that he thoroughly chose games he thought good enough to add.
Parent - By tomgdrums (****) Date 2010-12-31 01:06

> As I said, it's good for engine games but it's a bad idea for games against human players. I don't know if there are any playchess games in the Rybka 4 book, but Jiri said that he thoroughly chose games he thought good enough to add.


And I would assume the makers of the Hiarcs book also pick games they think are good enough to add.

I like consulting the Hiarcs book and then checking Shredder's book (and the huge book via the web at Shredderchess) and even Junior's book which is compiled by a GM.

I like comparing and then figuring out what I like.  I would love to add Jiri's book to the mix but I can't justify the cost since it does not get updated and can only be used in chessbase or ctg.  I didn't buy the Rybka 3 book for the same reason.  It is not cost effective.
Parent - - By Harvey Williamson (*****) Date 2010-12-31 01:08
Just like the Rybka book it includes the best games they may be Human they may not be. out of interest how do you know the lines are not good v Humans have you run some tests?
Parent - - By mdraith (**) [de] Date 2010-12-31 01:17
On the HIARCS website Mark claims that the HIARCS book is over 100 ELO stronger than the Rybka 4 book because of engine games played. That can only be true if the HIARCS book is a killerbook compiled to beat other engines. The Rybka 4 book is an opening book with the latest opening theory compiled by an expert on this topic. If you optimize an opening book or an engine to be strong against other engines then it will be very good for that special purpose. This is common sense, I did not test the book against human players, nor has the HIARCS team to prove the opposite. But when I compare 18.000.000 positions to 3.000.000 positions then I assume that there is much more information in the Rybka book.
Parent - - By Harvey Williamson (*****) Date 2010-12-31 01:22
Hiarcs book has been updated several times since the R4 book was released with human and engine games so it is no surprise the current version is better,
Parent - By mdraith (**) [de] Date 2010-12-31 01:30
When saying it's better, you should add for engine/engine games. As a chess player I can't believe that an 3.000.000 positions opening book can be considerably stronger than an 18.000.000 positions opening book. Sorry! I like the HIARCS engine very much and also use the internal opening book. But if the Rybka 4 book would be available in abk format or could be imported into Arena, I would use it instead.
Parent - By tomgdrums (****) Date 2010-12-31 01:02

> It's much wider (don't know if it's deeper) and you can use it for non engine games. For the HIARCS book, they add games played on playchess which is a very bad thing I think. Engines have no understanding of opening theory and adding such games is a very incompetent step when compiling an opening book. It will be good when playing engine games, but it will fail badly against human players.


I don't know how much of the book is made up of playchess games, but I do know the theory holds up when stacked against other sources.  I also know it has been great for my study (as I use it as a very nice guide to my own discovery)  Along with Shredder's book and Junior's book.

And the customer service rocks!
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-12-31 07:39

> Instead what I see is that the Rybka team is moving on to different type of customers.


Let me just assure you that public Rybka versions are still extremely important. We have lots of plans for those. There will be some cool new features in Rybka 5. :smile: One small but steady step at a time ..

Vas
Parent - By Ricky (***) [us] Date 2010-12-31 08:18
Sorry Vas, but it does not show that "still extremely important".

Best regards,

Ricky
Parent - By Geomusic (*****) Date 2010-12-31 08:46
We are patient Vas, hell we are still here aren't we? LOL
Rybka 5 will be the real meat and potatoes of all your work. Plus remote rybka will make beta testing a breeze and also safeguard your .exe
Parent - By Goran Trajkoski 02 (*) [mk] Date 2010-12-31 17:09
Most of the useres are expacting Rybka 5.
Parent - By vrboska (**) [de] Date 2010-12-31 11:37
Yes, you are right. This is my oppinion, too. Nothing to add to your words - bad customer service and bad basic attitude to the paying rybka customers.
Parent - - By oudheusa (*****) [nl] Date 2010-12-31 11:52
Jiri's book and the Hiarcs book are two completely different pricing models.

Jiri's book is a one time fee of $25,-. Hiarcs is a subscription model for $47,95 per year, including updates ofcourse.

Imo Jiri's book is a steal at that price. you can't expect updates for that.
Parent - By tomgdrums (****) Date 2010-12-31 14:26

> Jiri's book and the Hiarcs book are two completely different pricing models.
>
> Jiri's book is a one time fee of $25,-. Hiarcs is a subscription model for $47,95 per year, including updates ofcourse.
>
> Imo Jiri's book is a steal at that price. you can't expect updates for that.


Jiri's book is  $33 or more for the Chessbase version.  I feel like the Hiarcs book is a steal since I also get to use it internally with other GUI's of my choosing.  It is just different levels of customer service.
Parent - - By suj (***) Date 2010-12-30 21:43
Is there any option of buying the uci engine for a price?
Parent - - By Vasik Rajlich (Silver) [pl] Date 2010-12-31 07:42

> Is there any option of buying the uci engine for a price?


We don't currently have any plans to do this, sorry Suj!

Vas
Parent - - By Bouddha (****) [ch] Date 2010-12-31 12:59
Will there be an option to have for example a rent of 40 cores for an certain amount of hours non consecutive ?
Parent - By Vasik Rajlich (Silver) [pl] Date 2010-12-31 17:40
That's the plan - see the "Alternative Rental Possibilities" section. But it's for later ..

Vas
Parent - By Geomusic (*****) Date 2010-12-31 08:37
Rybka 5 uci = Rybka 4 Remote improvements which were supposed to go into Rybka 3+ LOL
Parent - - By Fulcrum2000 (****) [nl] Date 2010-12-30 20:02 Edited 2010-12-30 20:07
Ok, that's a pity. I write it down as 2014 in my agenda then.
Parent - By Geomusic (*****) Date 2010-12-31 08:39
well 2014-2015 :)
Parent - By Quicksort (**) [fr] Date 2011-01-07 03:51
Something is rotten in the state of Denmark...
Parent - - By Carl Bicknell (*****) [gb] Date 2011-03-27 12:04
I've been thinking about this. A Rybka FPGA card would be wonderful and more appealing to most people than renting the Cluster. However, there is one (rather large IMO) issue: Would it work with IDeA?

If it doesn't, then there may be limited use for such a card. If it does (maybe by running multiple instances of Rybka on the card or by just allowing the card short thinking time in IDeA if it isn't affected by determinancy - I don't know how determanancy and FPGA cards work) then that would be amazing - a kindof chess solution.
Parent - - By Labyrinth (*****) [us] Date 2011-03-27 13:47
Why wouldn't it? I figure it would serve mainly as an integer processor.
Parent - By Carl Bicknell (*****) [gb] Date 2011-03-27 15:43
Well I don't know if a FPGA card is 'one powerful processor' or 'lots of chips on a card running in parallel'. If it's the latter then surely they'll be issues of determinancy which will make some analysis unrepeatable.

Someone who knows about these cards should be able to say. I'd love to know what sort of speed up they give over an Intel 6 core machine as well.
Parent - - By Lukas Cimiotti (Bronze) [de] Date 2011-03-27 16:45
Of course such a card would work with IDEA - but you wouldn't be able to run more than one Rybka at a time on it. An FPGA card works like a lot of dedicated chess chips runing at the same time. So the output will be non-deterministic - like normal mp. I hope Vas will get this running in the near future (certainly not in 2011). I badly want a new toy :wink: Of course I would build a cluster of it.
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka Cluster Rental Program
1 2 3 4 5 6 7 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill