Author: Enrique Irazoqui
Date: 01:42:44 11/21/00
Go up one level in this thread
On November 21, 2000 at 04:06:26, Gregor Overney wrote: >On November 19, 2000 at 18:19:34, Robert Hyatt wrote: > >>On November 19, 2000 at 10:07:17, Randy Schmidt wrote: >> >>>Of course Bob's comments are interesting... >>> >>>Yesterday I put together (in web browser only) a dream system >>>from Dell. It had two 1-gig P3's and 1024 of RAM...total cost >>>around $6500. Now according to what I am reading, it would be >>>better to spend around $3500 and wait until the 1.6 gigs are >>>released? I can save $3000, wait 3 months and I have about >>>the same speed of machine? >> >>basically, yes. Of course if you do any computation at all, and you do >>multiple things, the second cpu will work quite well. The 1.7X figure I >>quoted was for my chess engine using 2 cpus. 2 compiles will run 2x >>faster, for example. > >Interesting. After the P4/1.5G got announced I made a "paper-based" comparison >of a dual P3/1G with Dell's new 330 Workstation. The numbers are: > >P4/1.5G SpecINT2000 535 (one CPU) >P3/1.0G SpecINT2000 418 (per CPU) > >(SpecINT2000 includes Crafty in its benchmark.) > >Those values are obtained using SIMD and SIMD 2. A feature that is not yet >supported by VC++ 6SP4. Let us take Bob's suggested value of 1.7 for Crafty then >the dual P3/1.0G offers a value of 710 (a p6/200 gets a value of 75). Since my >old p6/200 calculates 100,000 nodes/sec, I expect a dual P3/1.0G to achive >around 900,000 nodes/sec when using Crafty. Right? With the Chessbase version of Crafty 17.14 running on a dual PIII-933 with 464 MB hash, I got 1000k NPS analyzing a more or less average middlegame position. It would be interesting to see if Bob's Crafty 17.14 gives a similar analysis. Ply depth 38 moves both 3rr1k1/pp1nqp1p/2pb1np1/8/3N4/PP2P2P/1BQ1NPP1/R4RK1 b - - 0 1 Analysis by Crafty 17.14 P4: 22...Ne4 ³ (-0.49) Depth: 1/2 00:00:03 22...Ne4 23.Rfd1 = (-0.23) Depth: 2/3 00:00:03 22...c5 23.Nf3 ³ (-0.39) Depth: 2/3 00:00:03 22...c5 23.Nb5 Bb8 = (-0.21) Depth: 3/6 00:00:03 22...Ne4 23.Rfd1 Ndc5 ³ (-0.30) Depth: 3/6 00:00:03 22...Qe5 23.Nf3 Qe4 ³ (-0.44) Depth: 3/6 00:00:03 22...Qe5 23.Nf3 Qe4 24.Rac1 ³ (-0.37) Depth: 4/13 00:00:03 22...Be5 23.Rad1 c5 24.Nf3 Bxb2 25.Qxb2 ³ (-0.41) Depth: 4/13 00:00:03 22...Be5 23.Rad1 c5 24.Nf3 Bxb2 25.Qxb2 ³ (-0.41) Depth: 5/13 00:00:03 22...Be5 23.Rad1 c5 24.Nb5 Bxb2 25.Qxb2 a5 ³ (-0.39) Depth: 6/13 00:00:03 59kN 22...Be5 23.Rfd1 Ne4 24.Nc3 Bg7 25.Nxe4 Qxe4 26.Rac1 = (-0.22) Depth: 7/16 00:00:03 293kN 22...Qe5 23.Nf3 Qe4 24.Qxe4 Nxe4 25.Rad1 Bf8 26.Nf4 h6 ³ (-0.27) Depth: 7/16 00:00:04 806kN 22...Nb6 23.Rfd1 Rd7 24.Nf3 Red8 25.Bxf6 Qxf6 26.Qe4 ³ (-0.28) Depth: 7/16 00:00:04 929kN 22...Nb6 23.Rfd1 Rd7 24.b4 Red8 25.Rac1 Ne4 26.b5 = (-0.25) Depth: 8/19 00:00:04 1137kN 22...Be5 23.Rfd1 Ne4 24.Nc3 Ndf6 25.Nde2 Nxc3 26.Nxc3 h6 ³ (-0.26) Depth: 8/19 00:00:06 2781kN 22...Nd5 23.Rfd1 Nc5 24.Nc3 Nxc3 25.Bxc3 Ne4 26.Ba5 Rd7 ³ (-0.27) Depth: 8/19 00:00:06 3185kN 22...Nd5 23.Rad1 N7b6 24.a4 Nb4 25.Qc3 Be5 26.e4 Bg7 = (-0.23) Depth: 9/20 00:00:09 6224kN 22...Be5 23.Rfd1 Ne4 24.Nc3 Bg7 25.Nxe4 Qxe4 26.Rac1 Qxc2 27.Rxc2 Ne5 ³ (-0.27) Depth: 9/20 00:00:13 9785kN 22...Be5 23.Rad1 Ne4 24.Nf4 Nd6 25.Qd2 Qf6 26.Rfe1 Ne4 27.Qc1 ³ (-0.29) Depth: 10/20 00:00:20 16549kN 22...Be5 23.Rfd1 Ne4 24.Nc3 Bg7 25.Rac1 Ndc5 26.b4 Nxc3 27.Bxc3 Ne4 28.a4 ³ (-0.29) Depth: 11/20 00:00:38 34457kN 22...Be5 23.Rad1 Ne4 24.Nf4 Bg7 25.b4 Qf6 26.Rd3 Ne5 27.Rdd1 Ng5 28.Qc5 ³ (-0.30) Depth: 12/20 00:01:22 80076kN 22...Be5 23.Rad1 Ne4 24.Nf4 Ng5 25.b4 c5 26.bxc5 Nxc5 27.Qc4 b5 28.Qc2 Bxf4 29.exf4 Nge6 30.Nxb5 Nxf4 31.Rxd8 Rxd8 ³ (-0.26) Depth: 13/25 00:02:30 150361kN Enrique > A dual P4/1.5G should be the >first _dual_ CPU-based Intel system that reaches more than one mega-nodes when >running Crafty. > >But if you can wait a couple of months, just buy a dual P4/1.5G or, maybe, a >dual Itenium/1.5GHz........if you can wait, you'll probably never buy a system, >because at the same time Dell ships your system, it's already outperformed by >the next one in line. > > >Gregor
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.