Author: Vincent Diepeveen
Date: 05:46:30 07/27/03
Go up one level in this thread
On July 27, 2003 at 08:28:49, Jonas Bylund wrote: >On July 27, 2003 at 07:37:31, Vincent Diepeveen wrote: > >>On July 27, 2003 at 06:31:58, Jonas Bylund wrote: >> >>>On July 26, 2003 at 17:22:02, Russell Reagan wrote: >>> >>>>On July 26, 2003 at 16:25:37, O. Veli wrote: >>>> >>>>>Since it is hardware, can >>>>>we expect to be stronger than top software? >>>> >>>>I would expect it to be slower than top software, because cpu improvements >>>>happen so quickly, and FPGA programming (from what I've heard) is not a simple >>>>task. If he spends another two years working on it before releasing it (as >>>>Slater said), just imagine how much faster the cpus will be by then. >>>> >>>>If you're talking about something massively parallel like Deep Blue, that is one >>>>thing, but a single PCI card? I doubt that is going to do any better than break >>>>even with top of the line hardware, so why bother? IBM threw so much hardware at >>>>the problem that desktop cpu improvements wouldn't catch up for a LONG time, but >>>>a single PCI card doesn't seem to be worth the trouble of programming the thing, >>>>because desktop/server cpus will probably outperform it before too long. >>> >>>The way i understand it, the whole idea with running FPGA is that no matter how >>>much knowledge you add, you won't lose speed, will that not more than compensate >>>for the PC programs gain through faster hardware? >> >>Quote from Chrilly Donninger Paderborn, februari a few years ago (98 or my >>memory says 99 now): >> "I do not believe in knowledge at all Vincent. You are taking the wrong path. >>Nimzo in fact only grew stronger when i REMOVED knowledge from it". > >You kind of missed th point here, 1) Chrilly said in an interview that adding >more knowledge would not slow down the engine when using FPGA (which probably >also mean that your quote is outdated in the sense that some people have the >ability to change their point of view) >2) The question was not wether Chilly believe/d in more or less knowledge= more >or less strenght, the question was if the assumption that adding knowledge to >the FPGA does not slow down the search is true, then does that not potentially >compensate for hardware gain in conventional PC systems? a) yes it slows down a bit, because if there is knowledge that is dependant upon other knowledge then it takes more than 1 clock. We know that Brutus is doing eval within 1 clock so it can only have simple knowledge. If it is needing 2 clocks for example it would slow down directly 20% nearly in nps. b) yes you can of course independantly add knowledge in parallel of course if you have enough gates. However then they cannot sell brutus very well, because the more gates you need the more expensive it is to press a cpu AFAIK. I was told that if you need half a million gates that the price is then 2000$ for that (then customer price is $3000 minimum) and that a fpga cpu which is needing around 100k gates is going to be affordable (including card it can get sold for $500 then). So there is major practical limits. We know that brutus ran in 50000 gates around februari 2002. But this was very primitive knowledge still, though it had mobility and attacks and such which is very expensive in software. Of course Chrilly can't put all the knowledge that for example DIEP has in fpga hardware, because then the price where it will sell for will be like $5000 each fpga chip. As far as i know Chrilly is simply guessing that in the future it will get cheaper to produce fpga chips with many gates. So that is a big gamble. I am not going to vote against it getting cheaper, but *when* is it going to happen? Also the Mhz at which fpga cards run is very slow. Now how can we compare brutus with at the moment? The best program is Hiarcs8. See it as hiarcs8 in hardware. So if you search a lot more nodes a second. Around 20 MLN in hardware, then the problem is how you are going to do that over the pci bus which can only handle 100000 requests a second at most (and with the old fpga card brutus already was getting near to 100k requests a second). So theoretical discussions is useless when the chip produces for a lot of money. Chessbase will not pay Chrilly for a good program in hardware where they can sell only 20 versions from to collectors who are willing to pay $3000. But i do not believe that FPGA is very interesting. Because just ask Chrilly what time it takes to modify 1 byte to the program and then recompile. 24 hours? If i make a change then i simply email a version to all my testers. Chrilly has a problem here. He also lives on the other side of Austria from his big tester Kure. So i do not know how they develop brutus, but it must go very slow for sure. It means that the limiting factor is the slow development time for each pattern. Suppose 1 week for each pattern which is realistic. They will need 10 years to make up with DIEP, assuming i do nothing coming 10 years. So 2003 is the year where brutus has a real chance for the world title. After that software will outdate it. If Chrilly is lucky he wins in 2003 the world title. Then in 2004 with some luck prices drop. End of 2004 he sells the card. Januari 2005 it is outdated but sold already :) >>Someone who always follows simple solutions i could not possibly believe he >>manages to put a lot of knowledge in hardware. Where 'a lot' is measured by 2300 >>FM standards. > >Well this sounds like speculation to me, besides "someone" never always follow >the same exact patterns, that is too black and white (no pun intended). > >Oh and BTW i did not know that 'a lot' had a rating :) > >Jonas
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.