Author: Vincent Diepeveen
Date: 15:16:18 05/27/05
Go up one level in this thread
On May 27, 2005 at 13:50:53, Uri Blass wrote: >On May 27, 2005 at 11:54:03, Torstein Hall wrote: > >>On May 27, 2005 at 08:45:58, Vincent Diepeveen wrote: >> >>>On May 26, 2005 at 16:29:46, Joshua Shriver wrote: >>> >>>>Interesting... so how many of the 32 cpu's are CPU vs FPGA custom chips? >>> >>>It has 32 custom FPGA cards. >>> >>>>Have there been any stats as to nps on each fpga, and what kind of bus does it >>> >>>You are interested in objectivity, the hydra team is a very bad person to ask in >>>that case. They are interested in blowing it up to unheard dimensions. >>> >>>>rest on? I thought the FPGA's where attached on a PCI backplane, and if they're >>>>working with that many, have they come across any latency problems over the PCI >>>>bus. >>> >>>That's why they do not have hashtables inside the fpga cards. >>> >>>>Hydra intrigues me, since parallel programming is what inspired me to want to do >>>>an engine. (Besides also loving the game) >>> >>>Their parallel algorithm isn't worth much in the real world i guess. I don't >>>give a penny for their parallel efforts. If they would have a good parallel >>>algorithm, hydra would already run at 1024 cpu's by now for sure. >>> >>>Also the way it gets shown to the world is kind of the usual way. >>> >>>First dumb down 1 cpu a lot by not using hashtable last 6 ply, then show a good >>>speedup of a lobotomized 1 cpu versus 32. >>> >>>We know that drill from the past already. >>> >>>Like cilkchess (MiT), a fast bitboard engine, getting just like 2500-5000 nodes >>>a second at a single cpu at 512 cpu's 500Mhz, and diep a program 20 times >>>faster, gets at the same hardware at 512 cpu's 10000-20000 nodes a second a cpu. >>> >>>I mean, cilkchess team can claim whatever speedup they have, but without cilk >>>their program at 1 such cpu gets 200000 nodes a second, as it's a simplistic >>>bitboards engine with near to no eval (crafty has 4 times more knowledge in eval >>>than cilkchess in fact). So from my viewpoint they are missing a factor 40+ >>>somewhere in the compare to the single cpu. >>> >>>See the problem of the parallel speedup of Hydra? >>> >>>They can brag whatever about speedup, as long as they aren't comparing an >>>optimized single cpu version with the 32 processor version, it's not a very fair >>>compare. >>> >>>You can get a million nodes, but if all what you do with it is search 6 ply in a >>>highly selective manner, that still is 6 thin plies. >>> >>>It isn't 7 ply which i get with perhaps at most 100000 nodes, without any >>>forward pruning (just nullmove). Then we didn't discuss the huge evaluation >>>difference even. >>> >>>Rudolf Huber (SOS) calls that: "they first slow it down, or make the branching >>>factor horrible, in order to be able to claim a better speedup". >>> >>>>-Josh >> >>What kind of interest has the Hydra team of doing that? Anyway, the fact >>remains, it blows all PC programs, including Diep, of the board! >>Simple or complicated paralell search, null move etc. etc. who cares as long as >>it is the strongest chess machine on the planet! >> >>Torstein > >No proof for it. >It is only an opinion. > >I believe that Shredder is stronger than Hydra and Shredder was unlucky to >suffer from a bug so the version that played hydra was weaker than the >commercial version. > >Uri First of all Diep is a chessprogram, not a chessmachine. So no matter what happens hydra will probably always remain the strongest chessmachine. Any claim in that direction the answer is always: "yes it is". I feel first half of 2005, Hydra really was strong. Yet let's be honest, i didn't prepare anything for diep for paderborn2005. I didn't prepare anything for ict5 either. It will be disaster perhaps. But i'm changing it. Changes are needed to bigtime improve a product. After the changes are fixed it's strong once again. Same thing for shredder and all other software products. Hydra however will play 100% as strong as it plays now the next year, but it hardly will improve. Hardware is just too difficult to improve in that respect. Probably by august 2005, many are again better than Hydra. The only way to prove that is by playing world champs each year. I understand why a country in war with Israel doesn't show up at world champs in that country. But if it doesn't show up at Reykjavik 2005, that will be only because the team fears to lose there. It needs a sheik to order them to play a world champs. At the same time such a sheik gets real real utmost furious when not finishing first there, demanding heads roll (if not litterary). So i clearly realize the consequences for the team if they WOULD join and not win the title. However if they would have a reasonable good chance to win there, they sure join. I feel in 2005 they do make a chance to win. Therefore if that order doesn't come from sheik, all computerchess experts should realize that whatever their odds to win in 2005, for sure by end of 2005, Hydra will be surpassed by software. Chrilly showed a succesful trick (agressive kingsafety tuning). Shredder and Diep didn't have that yet. For sure Shredder will recover from that. Fritz and Junior already had that trick, but especially fritz lacks massive amounts of chessknowledge. Junior is a bit underestimated in that respect, despite having a bunch of world titles. The path of hydra, putting a program in hardware already means it's extreme difficult to improve more than with 1 byte a week. It's already hard enough to improve software... ...and they need days to just compile a copy of hydra and put it into action. Apart from software being ready for a world champ, software has another advantage that most here might not realize at such a world champ. You always can get good hardware for it. Those development FPGA cards that are inside now, i mean what are they clocked. 60Mhz or so? State of the art. But in 2010 it'll still run at 60Mhz. How about software, with processors like cell processor and multicore coming? Perhaps not cell itself is ideal for chess, but the technology idea (with branch prediction) sure takes care that normal processors will get released in multicore too. They'll have to compete with it, or all 'cheap' clusters in highend will be replaced by cells. Vincent
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.