Rybka 4 just around the corner it seems

[Graham Banks]

1. half of 2010 = may be 7 months still!?

Jouni

Seems that the release date has been put off due to what happened with Ippo/Robbo. That should make some people happy.

Graham,
The guy is going to offer an online rent for Rybka until it's released,and that's in term that you have already purchased Aquarium 2010 for 124 $....this is an unheard phenomena in our community,an unbelievable mechanism for gathering money,it's authentic I have to admit....
Can't you understand yet that Vasik is trying to squeeze every penny he can while he's still the king of the hill
Dr.D

Put the blame on the hackers/crackers/cloners.

[Spock]

Graham,
The guy is going to offer an online rent for Rybka until it's released,and that's in term that you have already purchased Aquarium 2010 for 124 $....this is an unheard phenomena in our community,an unbelievable mechanism for gathering money,it's authentic I have to admit....
Can't you understand yet that Vasik is trying to squeeze every penny he can while he's still the king of the hill
Dr.D

I don't know why you are so negative about Rybka. This is a really innovative idea in the chess world, another first as was Aquarium in it's radical re-invention of the GUI. All credit to ChessOK and Vas for doing this. If you don't like it, just don't use it !! But it is there for people who perhaps have low spec machines and just want to have some time at low cost using some better hardware and engine.

[Dr.Wael Deeb]

Graham,
The guy is going to offer an online rent for Rybka until it's released,and that's in term that you have already purchased Aquarium 2010 for 124 $....this is an unheard phenomena in our community,an unbelievable mechanism for gathering money,it's authentic I have to admit....
Can't you understand yet that Vasik is trying to squeeze every penny he can while he's still the king of the hill
Dr.D

I don't know why you are so negative about Rybka. This is a really innovative idea in the chess world, another first as was Aquarium in it's radical re-invention of the GUI. All credit to ChessOK and Vas for doing this. If you don't like it, just don't use it !! But it is there for people who perhaps have low spec machines and just want to have some time at low cost using some better hardware and engine.

I am not so negative toward Rybka and as you stated that if I don't like it I just can skip the offered feature....

My point is that while delaying the release date of Rybka 4,the Rybka team can still benifit from the engine itself yet unreleased,amazing to say the least....

Another point is that people with low spec machines will most probably not take this online offer because if they had enough money and/or enough interest in the computer chess in general,they will find a way to upgrade their hardware,makes sense,right

So this offer in question is targeted toward a certain top proffesional ppersonalities,organazations,elite chess clubs,etc....

Dr.D

[Dr.Wael Deeb]

1. half of 2010 = may be 7 months still!?

Jouni

Seems that the release date has been put off due to what happened with Ippo/Robbo. That should make some people happy.

Graham,
The guy is going to offer an online rent for Rybka until it's released,and that's in term that you have already purchased Aquarium 2010 for 124 $....this is an unheard phenomena in our community,an unbelievable mechanism for gathering money,it's authentic I have to admit....
Can't you understand yet that Vasik is trying to squeeze every penny he can while he's still the king of the hill
Dr.D

Put the blame on the hackers/crackers/cloners.

I don't think it's the one and only reason Graham,all we can do is guessing,speculating & praying....
Dr.D

[Gian-Carlo Pascutto]

Not sure I understand this concept. How is it like getting an extra core for free?

Analyzing with many engines. The original IDeA could only use one engine for the analysis, but the new version can split the analysis between many engines. You can, for instance, use four instances of single core Rybka for the analysis on a quad computer, essentially reducing the multiprocessor overhead to zero. It's like getting one extra core for free compared to running Rybka in 4-CPU mode.

http://www.chesscafe.com/chessok/chessok.htm

If running Rybka on 4 cores gives you a speedup of 3, then they are claiming the new IDeA will give you a speedup of 4 instead. So that's one core more.

Of course the claim is totally bogus. If it would be true, then Rybka should switch it SMP algorithm to IDeA by default, and the claim would be false again.

[Spock]

If running Rybka on 4 cores gives you a speedup of 3, then they are claiming the new IDeA will give you a speedup of 4 instead. So that's one core more.

Of course the claim is totally bogus. If it would be true, then Rybka should switch it SMP algorithm to IDeA by default, and the claim would be false again.

I don't see why it is bogus. Running on 4 cores gives you an effective speed-up of 3 due to SMP "losses", but 4 x 1 = 4, i.e. running 4 instances of single core, means there is no SMP loss. So I see the logic in that statement. Whether it gives a better result or not is another story.

[Rolf]

all we can do is guessing,speculating & praying....

...to "the guy".

[Uri Blass]

Not sure I understand this concept. How is it like getting an extra core for free?

Analyzing with many engines. The original IDeA could only use one engine for the analysis, but the new version can split the analysis between many engines. You can, for instance, use four instances of single core Rybka for the analysis on a quad computer, essentially reducing the multiprocessor overhead to zero. It's like getting one extra core for free compared to running Rybka in 4-CPU mode.

http://www.chesscafe.com/chessok/chessok.htm

If running Rybka on 4 cores gives you a speedup of 3, then they are claiming the new IDeA will give you a speedup of 4 instead. So that's one core more.

Of course the claim is totally bogus. If it would be true, then Rybka should switch it SMP algorithm to IDeA by default, and the claim would be false again.

I will explain the point

Suppose that you have a quad and you are interested in analyzing 4 positions from different games.

You have 2 options to use the quad:

option 1:analyze every position for 6 hours with a quad.
option 2:analyze every position for 24 hours by a single core.

I understand that the new IDeA give you the option to use option 2 and option 2 gives you better analysis because 24 hours on a single core is better than 6 hours on a quad.

Edit:Maybe you can use option 2 by running 4 processes of rybka at the same time but this seems to be less user friendly then having interface that allows it.

Uri

[dadij]

Uri, you seem to understand the advantages.

If I'm going to analyze a position with IDeA on a quad, I have several options, including these:

1) A single instance of Rybka using all 4 cores

2) Four instances of Rybka, each using 1 core.

In 2), IDeA will distribute the analysis tasks to the Rybka instances so all 4 can analyze simultaneously.

Let's say I use a fixed time to analyze each position in 1), e.g. 30 seconds. Then I will have 120 analyzed positions after 1 hour.

Since the speed up of Rybka running on 4 cores is approximately 3 (compared to a single core version), I need to let the single core versions in 2) run for 90 seconds to get the same analysis quality.

Therefore, in case 2), I let each of the 4 engines analyze for 90 seconds. After 1 hour I will have 4*3600/90 = 160 positions analyzed.

In other words, by using 4 single core engines I get 33% more positions analyzed compared to using Rybka running on 4 cores.

I hope this explains what I was talking about in the article.

Btw. currently I'm running IDeA on more than 30 cores.

[Gian-Carlo Pascutto]

Uri, you seem to understand the advantages.

He's talking about analyzing multiple unrelated positions. Which is something totally different from analyzing related positons from the same tree (=what you are talking about).

Since the speed up of Rybka running on 4 cores is approximately 3 (compared to a single core version), I need to let the single core versions in 2) run for 90 seconds to get the same analysis quality.

This math doesn't work out. It's making the mistaken assumption that IDeA analysis is 100% parallelizable, but that's just not the case.

IDeA makes different decisions about which position to analyze next depending on the result of previous analysis. Parallel analysis will hence end up analyzing positions that serial analysis does not, because it launches several searches at the same time. Hence: inefficiency.

I repeat the point I made: If IDeA would work as claimed, then "Deep Rybka" is a scam because it should use the IDeA algorithm instead. But of course IDeA also has parallel losses (maybe worse so than Rybka SMP), so the claim that it gives a free core is just marketing, because IDeA does not scale perfectly either.