Author: Robert Hyatt
Date: 18:56:54 06/22/04
Go up one level in this thread
On June 22, 2004 at 08:02:16, Djordje Vidanovic wrote: >On June 21, 2004 at 16:30:15, Drexel,Michael wrote: > >>On June 21, 2004 at 12:51:52, Robert Hyatt wrote: >> >>>On June 21, 2004 at 09:15:59, Joachim Rang wrote: >>> >>>>On June 21, 2004 at 07:19:34, Drexel,Michael wrote: >>>> >>>>>On June 21, 2004 at 06:43:00, Joachim Rang wrote: >>>>> >>>>>>On June 21, 2004 at 06:14:58, Uri Blass wrote: >>>>>> >>>>>>>On June 21, 2004 at 05:51:40, Joachim Rang wrote: >>>>>>> >>>>>>>>On June 21, 2004 at 05:24:33, Uri Blass wrote: >>>>>>>> >>>>>>>>>I talk with a person and he tells me that Xeon2 is new technology and it is >>>>>>>>>better than opteron but it simply does not fit the price that I agree to pay for >>>>>>>>>it. >>>>>>>>> >>>>>>>>>He tells me that there are graphs that tell that Xeon2 is better. >>>>>>>>>When I tell him that people in this forum told me that pentium4 is relatively >>>>>>>>>bad he tells me that he does not trust them(note that he does not deal with >>>>>>>>>chess programs). >>>>>>>>> >>>>>>>> >>>>>>>>that's the point. Xeons are fine for certain applications and in generally I >>>>>>>>would say not worse than Opterons (but not better neither and much more >>>>>>>>expansive). >>>>>>>> >>>>>>>>For certain applications Xeons will outperform Opterons significantly and for >>>>>>>>others (such as chess programs) Opterons will outperform Xeons significantly. >>>>>>>>That is what all people in that board are telling you, so no need to ask other >>>>>>>>people who know nothing about chess performance. >>>>>>>> >>>>>>>> >>>>>>>>>He also claims that the Athlon64 is the same quality as the pentium4 that I can >>>>>>>>>get. >>>>>>>>>He agrees that the opteron is better than normal pentium4 that I can get. >>>>>>>>> >>>>>>>> >>>>>>>>this is all wrong. You can't no more make such general statements, since the >>>>>>>>performance for different purposes very widely. In general the P4 has also its >>>>>>>>merits and performs in some multimedia applciation very well, but for chess... >>>>>>>> >>>>>>>>>I am interested to know if there is some graph that shows that the opteron and >>>>>>>>>even the athlon64 is better than the xeon2. >>>>>>>>> >>>>>>>> >>>>>>>>look at the benchmarks given in this board and here: >>>>>>>> >>>>>>>>http://www.beepworld.de/members39/computerschach2/chessmarks.htm >>>>>>> >>>>>>>As far as I can see the leader is xeon and not athlon and the xeon is more than >>>>>>>twice faster in nps than second place so the reason is not having 2 processors >>>>>>>against one. >>>>>> >>>>>> >>>>>>This is misleading: the 1st entry is with two processors and HT enabled so >>>>>>apparently using 4 processors. Fritz than pushes the nps in the sky but that >>>>>>does not mean that it reaches greater depth (probably the contrary is true). >>>>>> >>>>>>DualXeon 3,565 2953 kn/s Jens H. (Deep Fritz8, 2 CPU) >>>>>> >>>>>>The "realistic" value is the 4th one: >>>>>> >>>>>>DualXeon 3,565 1326 kn/s Jens H. (Fritz8; 1 CPU) >>>>>> >>>>>>And that is an overclocked Xeon, so a standard Xeon @ 3.2 GHz would give 1190 >>>>>>knps. Compare that to 1405 with AMD 64 @ 2.2 GHz! >>>>> >>>>>I don't think you get 1405 kn/s on a "normal" A3400+. >>>>>There is something wrong. >>>>>Two others reported ~1330 kn/s for AMD 64 @ 2.2 Ghz. >>>>> >>>>>Michael >>>> >>>> >>>>Perhaps they were unlucky? Other Mobo? Better RAM? I mean 5% difference is >>>>normal if RAM and Mobo differ. >>>> >>>>regards Joachim >>> >>> >>>I find the above numbers interesting. Here is a single-cpu Opteron 248 run >>>(2.2ghz) on the DTS paper positions. These logs are available on my ftp box... >>> >>>total positions searched.......... 24 >>>number right...................... 24 >>>number wrong...................... 0 >>>percentage right.................. 100 >>>percentage wrong.................. 0 >>>total nodes searched.............. 15728711815 >>>average search depth.............. 14.1 >>>nodes per second.................. 2184528 >>> >>>Crafty averages about 1/1000th the Opteron clock rate for NPS. IE 2.2ghz >>>produces 2.2M nodes per second. Either somehow I have gotten way faster than >>>Fritz, which seems unlikely, or the 64 bit opteron likes Crafty much better than >>>a 32 bit application... >> >>Crafty is faster than Fritz on 32 bit. >>For example: >> >>[D] 1r5r/3b1pk1/p2p1np1/p1qPp3/2N1PbP1/2P2PN1/1PB1Q1K1/R3R3 b - - 0 1 >> >>Crafty 19.12 (Dann Corbit) : 1243 kN/s >>Fritz 8.026 : 1035 kN/s >>Fritz 5.32 : 1976 kN/s >> >>Athlon XP 2.2Ghz,64 MB hash after 30 sec search. >> >>Michael > > > >I wouldn't generalise. Crafty 19.13 (ChessBase, 4 threads) is slower than Deep >Fritz 8 (4 threads), at least while analysing the given position: > >Crafty 19.13 1769 Kn/s >Deep Fritz 8 1999 Kn/s > >30 second search, infinite analysis mode, dual Xeon 2.4, 64 MB hash, Win XP Pro. Turn off hyperthreading for Crafty and test with mt=2. that's the best way to run it on newer PIV's since recent changes related to NUMA were made.
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.