Author: Vincent Diepeveen
Date: 09:22:43 06/29/04
Go up one level in this thread
On June 29, 2004 at 11:36:52, Joshua Shriver wrote: >I sadly agree. While software is important, hardware does play a big role. >No offense to anyone, but I think a more fair test of "engine strength" would be >decided by everyone having the same hardware. Nonsense, When you limit it to a single cpu then you ignore those who worked hard to get it parallel to work, also the most interesting for users is how strong it *can* be. Just giving a bit extra time at their PC will then give the same strength. I have played a few years at micro world champs and no one cares a shit for micro world title, under the condition that you run on a 1Mhz limited cpu, that's the hard fact. I prefer crafty at a 64 processor opteron cluster and you at a 3000 processor opteron system. The stronger the competition the better! Big hardware means more publicity and that's what computerchess needs a lot now! >This makes it rather unfair for the amature developers who roll in with a P3/P4 >or whatever is offered there, when compared to these quad opteron beasts. Even Uri should be able to roll in an AMD64 as we all read in CCC. Next year all cpu's are parallel capable if you buy a single cpu machine as it will have 2 or more cores. So even for the 'poor amateurs' who are not parallel capable yet there is not a single excuse like you write here. >I've read some ppl say that even if everyone had the same machine, then results >would be bias on those engines written for that arch. For example if everyone >was using a single Opteron, then those using bitboards would have an advantage. >In that case I say more power to the developer. In my eyes, it's the code that >matters, not necessarily who can create the best "computer chess machine >(soft+hardware). >For me I'm utilizing this to my advantage. I hope to be in the 2005 WCCC >competition, and already have time allocated on a supercomputer in Sweden. We welcome you of course. The more processors the better. I didn't know by the way that Sweden had many supercomputers. It only looks like a big flame to me what you write here. >If I can get even in the top 5, even with my lowly code, it really goes to show >what value WCCC has on real coding ability. I'll never be as good of a chess >coder than Dr. Hyatt or many other engine writers. Things nowadays are different than they were in start of 80s. In start of 80s you showed up with big hardware and knowing the bad tactical level of others and even having a bad branching factor of 20.0 which was pretty normal in the 80s and start of 90s, you could still get champion with big hardware. That changed however. a) supercomputers tend to have slow processors (exception is the many opteron supercomputers we will see come now) b) search depths are very big c) testing time is limited on supercomputers So you need a very good speedup and some brilliant algorithm to get a good speedup out of a lot of processors. If the machine is cc-NUMA you will first need a year or 3 to develop an algorithm that can get you faster than SMP gives at a quad opteron. It will be hard to succeed for you. It's summer 2004 now. You won't make it before end of 2005 to get something to work that works parallel great, not to mention that you will not have time to make a program that can have a chance to win from the strong programs. >I dont mean this as a flame, just a personal feeling. Your feeling is 100% wrong. It is much harder to score with a big supercomputer than it is to score a point with a single cpu A64. Just to parallellize your program bugfree you will already lose a year. When did you start parallellizing, in 2010????? >If WCCC is for the best "computer chess machine" then wonderful. >Are there any Comp Chess competitions that really test the accurate strength of >engines on a equal hardware level? Only the WCCC does give a notion who is best. Let's discuss openingsbooks to start with. The openingsbooks used at WCCC are much better than what you use in your backyard. If you want that also amateurs can get a quad opteron, feel free to give everyone a quad opteron for the wcc2005. If you fail to do so, then just shut up about inequality of hardware. It never can be an excuse. If Chrilly shows up with 120 chips and each chip getting 2 million nps in hardware, i will even cheer harder when i win from it, and encourage him to show up with such a machine in 2005 wcc instead of the 16 processors he uses now. If you show up with a lot of processors, i will cheer loud from joy too. If your software doesn't work parallel by 2005, you just are busy flaming here. >Sincerely, >Joshua Shriver >On June 29, 2004 at 07:28:11, Peter Berger wrote: > >>Was there any reason to believe that Crafty would have a big hardware advantage? >> >>It doesn't surprise me that ChessBase has similar abilities to organize fast >>hardware as Bob Hyatt. >> >>The only ones who will truely suffer are the ones who have to use the default >>hardware, playing against opponents with 10 times faster systems ... >> >>Peter
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.