Author: Slater Wold
Date: 15:47:13 10/19/01
Go up one level in this thread
On October 19, 2001 at 11:12:32, Robert Hyatt wrote: >On October 18, 2001 at 23:48:08, Slater Wold wrote: > >>Eval with 2 CPUs: >> >>Deep Fritz - W,S >>4qknr/r1b2pp1/2Q1p3/2PpP1Bp/3P1N1N/8/P4PPP/5RK1 w - - 0 1 >> >>Analysis by Deep Fritz: >> >>1.Qxe8+ Kxe8 >> ± (0.81) Depth: 1/3 00:00:00 >>1.Qxe8+ Kxe8 2.Ra1 >> ± (0.91) Depth: 2/6 00:00:00 >>1.Qxe8+ Kxe8 2.Ra1 Kd7 >> ± (0.78) Depth: 3/11 00:00:00 >>1.Qxe8+ Kxe8 2.Ra1 f6 3.exf6 gxf6 >> ± (0.72) Depth: 4/10 00:00:00 1kN >>1.Qxe6! >> ± (0.75) Depth: 4/15 00:00:00 4kN >>1.Qxe6! fxe6 >> +- (2.06) Depth: 4/15 00:00:00 4kN >>1.Qxe6 fxe6 2.Nfg6+ Qxg6 3.Nxg6+ >> +- (2.06) Depth: 5/15 00:00:00 8kN >>1.Qxe6 fxe6 2.Nfg6+ Qxg6 3.Nxg6+ >> +- (2.06) Depth: 6/15 00:00:00 14kN >>1.Qxe6 fxe6 2.Nfg6+ Qxg6 3.Nxg6+ >> +- (2.06) Depth: 7/17 00:00:00 35kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Rc1 Ne7 >> +- (2.22) Depth: 8/19 00:00:00 97kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Rc1 Ne7 6.Bxe7 Kxe7 >> +- (2.22) Depth: 9/19 00:00:00 235kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Rc1 Ne7 6.Bxe7 Kxe7 >> +- (2.28) Depth: 10/21 00:00:00 518kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Ng6 Kf7 6.Nf4 g6 >> +- (2.53) Depth: 11/25 00:00:01 1358kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Ra4 5.Rc1 Rxd4 6.Ng6 Ra4 >> +- (2.53) Depth: 12/27 00:00:02 2721kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Ne7 5.Bxe7 Kxe7 6.Ra1 Ra4 >> +- (2.56) Depth: 13/29 00:00:06 7203kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Ng6 Kf7 6.Nf4 Ba5 >> +- (2.47) Depth: 14/31 00:00:13 15955kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Ng6 Kf7 6.Nf4 Ba5 >> +- (2.47) Depth: 15/35 00:00:30 35300kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Ra4 5.Rd1 Ne7 6.Bxe7 Kxe7 >> +- (2.53) Depth: 16/37 00:01:17 91981kN >> >>(W, 18.10.2001) >> >>Eval with 1 CPU: >> >>Deep Fritz - W,S >>Analysis by Deep Fritz: >> >>1.Qxe8+ Kxe8 2.Ra1 Kd7 >> ± (0.78) Depth: 3/11 00:00:00 >>1.Qxe8+ Kxe8 2.Ra1 f6 3.exf6 gxf6 >> ± (0.72) Depth: 4/10 00:00:00 1kN >>1.Qxe8+ Kxe8 2.Ra1 f6 3.exf6 gxf6 >> ± (0.72) Depth: 4/10 00:00:00 1kN >>1.Qxe8+ Kxe8 2.Ra1 f6 3.exf6 gxf6 >> ± (0.72) Depth: 4/10 00:00:00 1kN >>1.Qxe6! >> ± (0.75) Depth: 4/15 00:00:00 4kN >>1.Qxe6! fxe6 >> +- (2.06) Depth: 4/15 00:00:00 4kN >>1.Qxe6 fxe6 2.Nhg6+ Qxg6 3.Nxg6+ >> +- (2.06) Depth: 5/15 00:00:00 8kN >>1.Qxe6 fxe6 2.Nhg6+ Qxg6 3.Nxg6+ >> +- (2.06) Depth: 6/15 00:00:00 14kN >>1.Qxe6 fxe6 2.Nhg6+ Qxg6 3.Nxg6+ >> +- (2.06) Depth: 7/17 00:00:00 35kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Rc1 Ne7 >> +- (2.22) Depth: 8/19 00:00:00 96kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Rc1 Ne7 6.Bxe7 Kxe7 >> +- (2.22) Depth: 9/19 00:00:00 235kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Rc1 Ne7 6.Bxe7 Kxe7 >> +- (2.28) Depth: 10/21 00:00:00 518kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Ng6 Kf7 6.Nf4 g6 >> +- (2.53) Depth: 11/28 00:00:01 1341kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Ra4 5.Rc1 Rxd4 6.Ng6 Ra4 >> +- (2.53) Depth: 12/29 00:00:02 2697kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Ne7 5.Bxe7 Kxe7 6.Ra1 Ra4 >> +- (2.56) Depth: 13/32 00:00:06 7357kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Ng6 Kf7 6.Nf4 Ba5 >> +- (2.47) Depth: 14/34 00:00:13 15576kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Rxa2 5.Ng6 Kf7 6.Nf4 Ba5 >> +- (2.47) Depth: 15/35 00:00:31 37201kN >>1.Qxe6 Qxe6 2.Nxe6+ fxe6 3.Ng6+ Ke8 4.Nxh8 Ra4 5.Rd1 Ne7 6.Bxe7 Kxe7 >> +- (2.53) Depth: 16/36 00:01:20 96223kN >> >>(W, 18.10.2001) >> >> >>Review: >> >>2 CPUs: 1,194,558 nps >> >>1 CPU: 1,202,787 nps >> >>And just for reference: >> >>Fritz 6 "Fritzmark" = 1330k nps >> >>Interesting, ah? >> >> >> >>Slate > > >First, there are two ways to compute NPS: > >NPS=TotalNodes/CpuTime; > >NPS=TotalNodes/ElapsedTime; > >Either one is perfectly reasonable. The first gives "average NPS per >processor", the second gives "average NPS overall". > >In your case above, it doesn't look like DF got _anything_ from the second >processor, looking at the time taken to reach depth 15. 30 seconds vs 31 >seconds suggests that either (a) your machine is not using the second >processor (this usually happens when someone tries to use windows 98, etc, >rather than win2K). Or else something else is running and using one of >the two processors heavily... Win2k reports that the exe is getting 50% with 1 thread, and anywhere from 60% - 80% with 2. And the rest going to idle time. I've done this test on several positions, and it's always the same. 1 CPU is faster. It's pretty weird. No other engines have this problem.
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.