shiv wrote:Hmm, if the goal is for chess analysis and not for playing games, why cant we use a map reduce solution?
Essentially this can mean only one machine (can be multicore) can analyze a unique bunch of positions and machines all output analysis to the shared opening book. This is essentially the idea behind rybka aquarium.
Because the parallel efficiency will suck compared to a real parallelized tree search (which I presume Rybka Cluster is using, or it wouldn't be so strong).
Of course, some people here are willing to believe it when the Aquarium guys claim perfect speedups with IDeA, so if you truly believe that then what you propose is better. (And you'll get your ass kicked if you try to play a game with this "perfect speedup" solution)
shiv wrote:Hmm, if the goal is for chess analysis and not for playing games, why cant we use a map reduce solution?
Essentially this can mean only one machine (can be multicore) can analyze a unique bunch of positions and machines all output analysis to the shared opening book. This is essentially the idea behind rybka aquarium.
Because the parallel efficiency will suck compared to a real parallelized tree search (which I presume Rybka Cluster is using, or it wouldn't be so strong).
Of course, some people here are willing to believe it when the Aquarium guys claim perfect speedups with IDeA, so if you truly believe that then what you propose is better. (And you'll get your ass kicked if you try to play a game with this "perfect speedup" solution)
GCP can you explain? I thought running the engine in single core mode several times over was more efficient....
tano-urayoan wrote:Extract from the Rybka forum:
The Rybka Cluster is a high-performance supercomputer located in Minden/Westfalen, Germany. Combining software and hardware engineering, the Rybka Cluster features exclusive Rybka versions not available to the public running on customized hardware built with the specific goal of providing top Rybka performance at an affordable price.
Starting on February 1, 2011, the Rybka Cluster will be available to the public for the first time under the Rybka Cluster Rental Program.
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.
Rybka's author has certinly developed a mental disease....before I read this article I had my doubts,now I am sure....
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
Werewolf wrote:
GCP can you explain? I thought running the engine in single core mode several times over was more efficient....
The search function here should turn up the relevant discussion.
But basically, the parallel search algorithm in the engine is much more efficient than the one in the GUI, despite the GUI getting "perfect" speedups and the engine only getting partial efficiency from the extra CPUs.
Put another way, it's much easier to parallelize an algorithm that sucks.
Werewolf wrote:
GCP can you explain? I thought running the engine in single core mode several times over was more efficient....
The search function here should turn up the relevant discussion.
But basically, the parallel search algorithm in the engine is much more efficient than the one in the GUI, despite the GUI getting "perfect" speedups and the engine only getting partial efficiency from the extra CPUs.
Put another way, it's much easier to parallelize an algorithm that sucks.
Thanks for answering. So if I wanted the 'truth' of a position and I had, say, a few days to find it, I'd be better off leaving an engine on infinite analysis than leaving IDeA to chug away by itself?
I'd suggest that some should read this fine excerpt :
"There are some things which cannot be learned quickly and time, which is all we have, must be paid heavily for their acquiring. They are the very simplest things and because it takes a man's life to know them, the little new that each man gets from life is very costly and the only heritage he has to leave."
Peter Skinner wrote:If Vas can get people to pay that amount, then all the power to him.
Look at how many people whine over the cost of Windows, yet still purchase it...
In all reality, I hope he gets it 100% rented for the next 5 years. I would wish the same for anyone striking out on a venture and trying something that no one has either done or thought about
Peter
Hello Peter,
Oddly enough, I find myself in agreement with you on the bold part. Audacity should be rewarded.
seems doomed to failure...i can't see it succeeding.
what's to stop any individual or company with sufficient funds (IBM/Microsoft) from throwing an open-source engine on a 1000 core IBM server... and renting it out?
this would be a chess 'entity' that would crush Rybka cluster as it exists today.