Rybka Chess Community Forum
I truly hope your idea gets off the ground, it certainly gives you a great deal more security and therefore peace of mind that all your future development efforts are secure from others wishing to steal them, however, I for one cannot afford the rates you have set.
Best of luck.
I'm asking myself why Deep Blue was disassembled after winning against Kasparov...
We're launching with just the bigger stuff but as soon as we automate everything, we'll offer smaller, cheaper options.
i am still unclear of the market you are aiming at.i have decent hardware and could afford to rent time,but i wont.anyway best of luck maybe we are all missing something.
there is no market for this:
1. everybody has 4-8 cores already
2. there are stronger (maybe illegal) engines who needs exactly half that cores
>2. there are stronger (maybe illegal) engines who needs exactly half that cores
Perhaps Rybka cluster is much better from Rybka 4 and from any of the "illegal" as you call them, engines.
That is an interesting question. How Deep Rybka 4 on 1/2/4/8 etc cores would fare against Rybka cluster on 1/2/4/8 etc cores?
That's another big problem here. Testers will refuse to test it, even for free, if they cannot control the hardware, which in this case is not constant. Not to mention that the engine will be periodically updated, with potentially adverse effects.
Yes, I had the same problem with Rybka 1 and Rybka 2. One tester refused to test it because there were too many updates.
We'll probably provide periodic performance measurements and let users extrapolate for the week-to-week updates.
ok, thanks for making that clear, i'm at least slightly relieved. could you provide any details?
I think it's better to wait until final decisions have been made. There is still a long road ahead ..
true -- so why is this even public?
You can think of it like Intel's roadmap.
So you are trying to make a more powerful cluster?
Edit: or your trying to make a more manageable cluster.
> You can think of it like Intel's roadmap
Better to call it
"The Rybka Road map for peace"!
To keep us apprised of his breaking groundwork with the new remote project.
While it is nice to have the strongest possible chess entity available for hire, I really doubt this will be a good or at least profitable business model. I always believe that it is better to sell a lot of cheap items rather than hope for one big sale. The price mentioned is very reasonable when you consider how much it would cost to setup and build your own cluster, but really it is not for 99.99999999% of people interested in chess. Personally I will just wait for a PCI card or FPGA card Rybka ... there is something about having the hardware where you can touch it. Having said that, if I was to advice on a successful business model, it would be to develop an Iphone and Android application. Of course you could always put a hook in the software to rent the cluster, but really most would be happy to have Rybka on their Iphone. I bet the money made on that would far exceed profits made by going high end, but again there is no reason why you can't have both in one application.
Yes but the reality is that in between major tournaments the cluster is probably doing development work or sometimes nothing at all. The expenditure has already been made, that is sunk costs. So any revenue is pretty much money straight into the bank apart from obvious running costs like electricity
You can connect to any UCI engine on your iPhone now so I guess you can connect to the Cluster in the same way. It is possible in the Glaurung iPhone GUI.
Ports to mobile devices are another box to check, yes.
>You can connect to any UCI engine on your iPhone now so I guess you can connect to the Cluster in the same way
>Ports to mobile devices are another box to check, yes.
There is enough paranoia about when your opponent wanders off to the toilet already knowing everyone has a mobile phone lmao
M ANSARI, I think Vas and his clan understands the basic lemonade stand model of business after 5 years. We don't have all the facts, Vas might have a queue of TOP 10 GMS and sultans interested in remote rybka at the Cluster-40 and 100 price points. Otherwise why would they go through the trouble of setting them with no demand? If Topalov or some other big sponsor prebought time on the server and is offering a rather large sum to hold on to it this might sway Vas to go one way rather than the other.
> Vas might have a queue of TOP 10 GMS and sultans
That is what is making Lukas so damned cavalier toward the local membership here on the forum- he sees his setup paying for itself without even breaking a sweat- Probably, with a few bucks going into add on hardware to boot.
I personally would buy a RYBKA FPGA card if it were around $600. But chances are it would be closer to 2k-3k :)
5-10 of them would make the cluster very interesting to play.
Yes. Another point (besides the pricing model) is security. Certainly I do believe that the cluster guys won't sell or otherwise use the analysis I'm doing on their hardware, but we've seen enough piracy movies going on in real life during the last months and years, so that would be something I'd really be concerned about...
> I always believe that it is better to sell a lot of cheap items rather than hope for one big sale
The problem is that Vas cannot sell alot of Rybka's anymore. The moment its up for sale, you'll find it in the internet for free
But thats not the case for the Cluster. So the Cluster should be much cheaper. Instead of it setting idle alot of time, it should be making money at least 8 hours every day, and that cannot happen with these astronomical prices
> I for one cannot afford the rates you have set.
Well, its not targeted to hobby-players but GM's & enthusiasts... and be sure the queue is full for this program.
238 Euro (incl. VAT) a day nah think i may leave it hehe
Very few GMs have enough money to pay that.
GMs are not ritch for most of them.
again, true, hence my comment about the market being 'psychos and sheiks'. hopefully this is just a beta prematurely announced & something worthwhile will be coming down the pipe.
I don't know about psycho's ... but a sheikh would want the damn cluster in his home basement
But the time almost coincides with the candidate matches tournament so maybe there are 8 potential costumers.
unless the cluster has seen some kind of revolutionary opening knowledge present w/in it (possible, i guess?) there is simply NO WAY it wins cost vs benefit as far as tournament matches go. the idea that a computer can outsmart 150 years of opening theory from the greatest chess minds of all time, complete with hypercompetitive SGMs armed with a team of seconds & powerful hardware of their own -- well, if vas has done that with the cluster alone, he's done a helluva lot more than any of us seem to realize at this point. and i would bet that most SGMs think the same way ( remember, they're even more beholden to their own strengths than us patzers). topalov'll probably take a shot at it, maybe to see how it compares to his blue gene. other than that maybe an experiment or two from a couple others. but i predict zero book wins as a result of the cluster being available.
no one will buy into this except for ppl who can afford it as a tinker toy & unfortunately i know vas is realistic enough to realize that. idk where his head is.
prestige project as the engine furs are flowing away
also recall that jiri (R4 book author) + topalov + blue gene faced anand in the WC. we all know how that turned out.
Doesn't seem that he invented the opening-wheel new if you browse through Iweta's recent games at the WWCC
> Very few GMs have enough money to pay that.
And what do they get? A Rybka which is simply faster than that Rybka4 working on his own 4core-machine 24 hours a day?
Or, if he wants and spends a bit money, working about 3 problems on 3 such machines?
Or brings the cluserRybka something very new in a directly chess-relevant manner?
Positional, or tactical, or ... improvements? Innovations?
For months (years?) only reachable for those who pay for rentable Rybka?
IMHO this is just to expensive for the value added for a GM.
For a rich correspondance player that may do the job.
Thats why I specified "top 10 GM" and not just any GM. Most of these guys have sponsors who will buy anything they need.
> Thats why I specified "top 10 GM" and not just any GM. Most of these guys have sponsors who will buy anything they need.
This is false.
"sponsored players: carlsen, anand, maybe the Armenian government gives some support to Aronian but that's it."
> This is false.
> "sponsored players: carlsen, anand, maybe the Armenian government gives some support to Aronian but that's it."
I think Huitzilopochtli is on the money! (pun equally intended)
Otherwise the Cluster would be doomed from the start and Lukas wouldn't be so damned cocky!
> ... and be sure the queue is full for this program.
I think you overestimate the need for such a chess monster by top chess players. Ask them, many do not like engines for their preparation at all. If they analyse a position and the engine says the score is 0.21 so what? They want to understand a position and no engine can do this for them.
The score is not the important part.
The important part is the move that the chess engine suggest and the chess engine may suggest
an interesting move that a top player did not think about or considered it as bad but can discover with more time that it is good.
Well this is exactly as expected. The package is launched at the elite players for whom it might prove good value just before major events, other GMs with rich sponsors, and just plain rich people who want a powerful toy.
For people who want to spend hours playing engine matches or analysing correspondence games the package is hopelessly expensive.
For me it may or may not make sense. I will continue using my laptop with R3 for most (casual) analysis.
I probably average 2.5 hours a week serious analysis for which I would like a faster machine and the best possible software. If the 0.25 Euro per core hour holds dwn to the smaller pacakges then it would cost me 195 Euro per year to rent 6 cores for my serious analysis. That would represent good value for me - cheaper than buying or building a 6 core machine.
However, I guess that if you rent a few hours here and there of only 6 cores it might be cost more than 0.25 per core hour. At say 0.75 per core hour it would look like a poor value for me and I would build myself a 6+ core machine.
So I await with great interest the pricing of the "cheaper" packages. (In the meantime I won't buy R4 and wouldn't anyway because of the unfixed bugs).
This makes me laughing. No bugfixes for old customers, but try to make money with new rich customers!
This idea is far away from real business. Im in a leading position in one of the biggest OnlineShops in Germany, and i can tell you: this businessidea is bullshit and will never work .
may be the Rybkateam has lost the view for the market and the old customers. Beside, everybody in here know, that there are some engines which are stronger than Rybka, and that for free.
With a fast six core even Rybka cluster will have only a very small chance to win against this Copperfields of the world.
Time will tell on the business model. I am far from sure that it will work but it may do. (I too have plenty of experience with large-scale on-line retailing in the UK and to an extent Germany too, but this is something completely different).
My goal is not to win but is to analyse.
And personally I would not use something of dubious pedigree. It's a moral issue. Absolute proof is not needed for me to take a moral position, any more than if a stranger in a pub offered me a very cheap TV. I may know for sure it was stolen but I would not buy it from him.
If you are happy with the morals around "Copperfield" et al, fine use it. I won't.
The one thing I do agree with you on is the issue of no bug fixes for serious problems. That really has been shabby.
How does the buyer know that it is getting 100 cores and not 34? Because the screen on his computer says so?
How does the buyer know that any given time period has not been oversubscribed?
We have a credibility gap, here.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill