Author: Charles Worthington
Date: 07:58:49 02/22/03
Go up one level in this thread
On February 22, 2003 at 01:22:22, Aaron Gordon wrote: >On February 22, 2003 at 01:13:00, Charles Worthington wrote: > >>On February 22, 2003 at 01:05:46, Aaron Gordon wrote: >> >>>On February 22, 2003 at 00:31:38, Charles Worthington wrote: >>> >>>>On February 21, 2003 at 23:52:51, Charles Worthington wrote: >>>> >>>>>On Febronuary 21, 2003 at 23:48:41, Charles Worthington wrote: >>>>> >>>>>>On February 21, 2003 at 19:57:10, Aaron Gordon wrote: >>>>>> >>>>>>>On February 21, 2003 at 18:02:39, Charles Worthington wrote: >>>>>>> >>>>>>>>On February 21, 2003 at 17:37:35, Aaron Gordon wrote: >>>>>>>> >>>>>>>>>On February 21, 2003 at 08:04:18, Mike Byrne wrote: >>>>>>>>> >>>>>>>>>>On February 21, 2003 at 08:02:34, Mike Byrne wrote: >>>>>>>>>> >>>>>>>>>>>On February 21, 2003 at 07:14:47, Charles Worthington wrote: >>>>>>>>>>> >>>>>>>>>>>>On February 21, 2003 at 07:05:22, Charles Worthington wrote: >>>>>>>>>>>> >>>>>>>>>>>>>On February 21, 2003 at 06:47:11, Mike Byrne wrote: >>>>>>>>>>>>> >>>>>>>>>>>>>>On February 21, 2003 at 04:46:53, Charles Worthington wrote: >>>>>>>>>>>>>> >>>>>>>>>>>>>>>Bob what program is required for me to conduct benchmark tests with Crafty? >>>>>>>>>>>>>>>Could you please e-mail it or post a link here to it? Thank you. >>>>>>>>>>>>>> >>>>>>>>>>>>>>Charles, >>>>>>>>>>>>>> >>>>>>>>>>>>>>You have it, the "crafty" program has a built in benchmark ....start crafty in >>>>>>>>>>>>>>dos mode (console) with no crafty.rc file ( a plain taxt file you create with >>>>>>>>>>>>>>engine parameters - but in this case - do not have a crafty.rc file in the same >>>>>>>>>>>>>>directory as crafty). >>>>>>>>>>>>>> >>>>>>>>>>>>>>Type word "bench" at the command prompt. >>>>>>>>>>>>>> >>>>>>>>>>>>>>Mike >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>>Thanks Mike. I haven't set the Crafty you sent me up yet so I didn't know. In >>>>>>>>>>>>>all honesty I have no Idea how to set it up to run on the Chessbase server. The >>>>>>>>>>>>>Crafty that comes with fritz is already set up so I have never had to set one up >>>>>>>>>>>>>yet. The other foreign progs are easy just drop in the eng and dll and you are >>>>>>>>>>>>>done. This does not look so easy. :-) >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>>640 kNs.....Not good :-) >>>>>>>>>>> >>>>>>>>>>>on your new machine?? >>>>>>>>>> >>>>>>>>>>ok I see it in your title ...that is respectable for 1.2Ghz Celeron -- it's in >>>>>>>>>>the ballpark -- I think a dual 3 Ghz will get 3M nps.... >>>>>>>>> >>>>>>>>>Hyatts Dual xeon 2.8GHz only gets 2.1 million in the benchmark.. >>>>>>>>>If you scale it up to 3.06x2 + HT you'll only see about 2.3 million. >>>>>>>>> >>>>>>>>> >>>>>>>>>Crafty v19.4 (1 cpus) >>>>>>>>> >>>>>>>>>White(1): mt=4 >>>>>>>>>max threads set to 4 >>>>>>>>>White(1): bench >>>>>>>>>Running benchmark. . . >>>>>>>>>...... >>>>>>>>>Total nodes: 104415743 >>>>>>>>>Raw nodes per second: 2130933 >>>>>>>>>Total elapsed time: 49 >>>>>>>>>SMP time-to-ply measurement: 13.061224 >>>>>>>>>White(1): end >>>>>>>> >>>>>>>> >>>>>>>>Thats good to know aaron but i never said i would pull 3 million with crafty. I >>>>>>>>run fritz primarily. I may experiment some with crafty though. where are u >>>>>>>>getting a 3000kNs figure? >>>>>>> >>>>>>>Mike mentioned 3000kN/s. By the way, from what I've seen Crafty will get more >>>>>>>kN/s than Fritz will. Crafty has better support for HT, too. The only way you >>>>>>>will come close to 3 million is by using an AMD system. Those don't need HT and >>>>>>>are fast already. Crafty gets 1.3 million nodes/sec on a P4-3.06GHz with HT, >>>>>>>1.8x speedup and you've got just over 2.3 million nodes/sec, just as Hyatt has >>>>>>>shown us here with his 2.8's. Now, you don't get to use HT (well, it's not >>>>>>>useful yet) in Deep Fritz. Take my CPU for example. I get about 1.6 million >>>>>>>nodes/sec in crafty's benchmark. 1.6 million * 1.7 speedup = 2.72 million >>>>>>>nodes/sec. This will also be quick for Deep Fritz, too. If you want speed and >>>>>>>don't mind pushing your chips a little (only 10% overlock, not much at all) >>>>>>>you'll have the fastest box. >>>>>>> >>>>>>>Considering the motherboard will be $150 and each chip is less than $100 thats a >>>>>>>pretty good deal. ~$1000 for a system that gets 2.72 million nodes/sec without >>>>>>>any HT help OR $4000-5000 for a system that gets 2.31 million and has to support >>>>>>>HT. I don't know about you but the decision is pretty clear to me.. >>>>>> >>>>>> >>>>>> >>>>>>well actually aaron i just watched one of the guys from ccc here in playchess >>>>>>with a dual 2600mp and he was pulling 1995 kNs with deep fritz 7 and was >>>>>>overclocked to over 2800...I saved the game and the kNs posted by the >>>>>>server...and it was more than one game as he was getting demolished by a 2.2 ghz >>>>>>p4 single. Oh and BTW i dont need 3.06 to beat that...Bob's 2.8's will pull 2250 >>>>>>themselves. You guys can make all the wild claims here you like but the fritz >>>>>>server tells on you every time. The 3.06 dual will be here next week so If an >>>>>>amd system here can get a higher kNs i want to see it on the server where it >>>>>>counts but u had better do a whole bunch of overclocking and get those cpu's >>>>>>steaming. You would have to be dilusional to believe that AMD 2600MP is superior >>>>>>to a high end Intel Workstation. And I don't need a Degree in computer science >>>>>>to figure that one out. Just a little observation on the server will do. So if >>>>>>you can outpost the xeons dont tell me...come show me..then i will be the first >>>>>>to come back here and congratulate you... :-) Of course it wont happen so... >>>>> >>>>> >>>>>Oh and also i almost forgot There were others from here that witnessed it too >>>>>along with the speed he claimed to get here...the two wern't even close. >>>> >>>> >>>>Oh and i also believe that if AMD had the technology or the research funds to >>>>have invented hyperthreading technology...they would have. it doesnt take a >>>>rocket scientist to see that two threads are better than one for multiple >>>>applications. So i will keep the 2600 dollar pair of xeons if you don't mind. >>>>Besides i wasnt looking to save money i was looking for a good machine that >>>>didn't_have_to be overclocked to get 2500 kNs. Had I been searching for an >>>>economical system i would have chosen the AMD. But I would have lived with what >>>>I had and not tried to fry eggs on it. :-)) Not everyone here wants 100.00 >>>>cpu's. Thats why there are two markets... >>> >>>As koibito pointed out two XP's in SMP configuration are slower than two MP's. >>>Also I've already done the comparison. Koibito (dual athlons @ 2.31) gets more >>>nodes/second than hyatts box. Also, my old AthlonXP 1.6 @ 1.86GHz beat Hyatts >>>dual Xeon running Crafty? Does that make my machine better than his? Sure >>>doesn't. Programs, books, hash sizes, etc were all different. If you take the >>>exact same program with the same settings faster will better. There's no way >>>that single P4 is faster than his dual and that single P4 gets barely half of >>>what my cpu gets in chess stuff. >>> >>>Also, did koibito play a few hundred games against that P4? Did they use the >>>exact same program and settings? You can't make ANY conclusions about speed >>>based off what you've seen in those 1-2 games you saw. If you do then go ahead >>>and assume my old XP 1.86 is better than hyatts xeons.. it's not true but go >>>ahead anyway, you did the same with the dual Athlon box. ;) >> >>actually i never said her machine was faster than his duals...you werent paying >>attention. i said her single cpu was beating him every game...not that it was >>faster...what I_did_say was that his kNs there was waaay lower than what he >>claimed it was here. Nothing more. Everyone is posting unreasonably high figures >>here. Tell us what the machine really does within the bounds of the laws of >>physics...not what you_wish_it would do. I wish mine would get 10,000 kNs but it >>wont. And until something comes out that will get that speed I will be happy to >>stick with whatever is best at the time and not have to roast marshmallows on my >>processors. :-)) > >What engine what he using? Some engines get REALLLLYYY low nodes/second, some >get really high nodes/second. Want to see an example? Here is my CPU running >Goliath Light v1.5. >http://speedycpu.dyndns.org/fastgl.jpg >6,250kN/s, does this mean my PC is 2x faster than the 8 cpu Deep Junior box used >in the Kasparov game? Nope. The point I'm trying to make is your aren't doing >controlled tests. One PC gets 800kn/s in Shredder 7, the other gets 900kN/s in >Deep Fritz, which is faster? The Shredder 7 PC, actually. See, you need to use >the SAME engine, SAME position, SAME hash tables, etc. We've already done this. >Mine got ~1720kn/s, Hyatt got ~2100, koibito got ~2400kn/s. > >You can take Deep Blue and put TSCP on it and then run Chesstiger 15 on a 486. >When the 486 running 2000 nodes/second beats the pc doing 200 million, what are >you going to say? Probably, "486 is my hero!!! I'm going to get a dual 486 and >own everyone! HAHA" :P Well now you are stepping off into an area that I_do_ know. He was not running a slow engine. He was running Deep Fritz 7 in 3+2 blitz. Deep Fritz, while not as fast as Deep Junior 7, still belongs in the category of fast engines. Now perhaps he may say that he was running a super large hash and it slowed him down...who knows? But running a hash over 64MB in 3+2 would be totally unnecessary and_would_ kill his speed and performance. Sara ran a Fritzmark for me yesterday on her AMD 2600 single cpu machine. She posted 1230kNs. That equates to 2275 kNs with the 2600MP Dual Box. A VERY respectable speed for any machine. But the processor that everyone compares the 2600xp to is the p4 3.06 and_several_ benchmarks with deep fritz have shown it to be in the 1350 to 1385 range with Deep Fritz 7. A difference of only 155kNs over the single 2600xp. But in a dual machine the difference is noticable, if slight. The point I am trying to make is if you have a factory AMD system_today_you will have to overclock the_high_end AMD chip to outpost the 3.06 P4 in kNs. And I believe Bob has been having _this_same discussion with Aaron for_several_ days now. Overclocking high-end chips in_not_safe.
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.