Author: M Hurd
Date: 15:41:54 01/27/06
Go up one level in this thread
On January 27, 2006 at 13:01:33, Bernardo Wesler wrote: >I noticed that Rybka improves its performance against F9,Sh9 and TogaII if it >plays with low hash tables sizes; for example having 1gb as a total ram, Rybka >plays better with 32 mb hash than with 128, 256 or 512...... >Have anybody else noticed this? >If so, why? >Thx Hello Bernado Have a look at the following position : [D]r3r1k1/1bq2np1/1p1ppp1p/7B/nPP1PP1N/6P1/4Q2P/B1R1R1K1 w - - 0 34 Analysis with Rybka using 32 mb hash time to find Bxf6 = 2o mins 54 secs Analysis with Rybka using 256 mb hash time to find Bxf6 = 9 mins 53 secs New game - Rybka 1.01 Beta 13 32-bit r3r1k1/1bq2np1/1p1ppp1p/7B/nPP1PP1N/6P1/4Q2P/B1R1R1K1 w - - 0 1 Analysis by Rybka 1.01 Beta 13 US 32-bit: 32 mb Hash 1.Bd4 = (-0.05) Depth: 3 00:00:00 1.Qg4 ² (0.36) Depth: 3 00:00:00 1.Qg4 ² (0.45) Depth: 4 00:00:00 1.Qg4 Qe7 ² (0.47) Depth: 5 00:00:00 1.Qg4 Qe7 2.Bd4 ² (0.47) Depth: 6 00:00:00 11kN 1.Qg4 Qe7 2.Bd4 Ra7 ² (0.49) Depth: 7 00:00:00 29kN 1.Qg4 Qe7 2.Bd4 Ra7 3.Ra1 ² (0.54) Depth: 8 00:00:01 83kN 1.Qg4 Qe7 2.Bd4 Bc6 3.Rc2 Ra7 ² (0.53) Depth: 9 00:00:04 214kN 1.Qg4 Qe7 2.Bd4 Bc6 3.Ra1 b5 4.Ng6 ² (0.59) Depth: 10 00:00:08 398kN 1.Qg4 Qe7 2.Bd4 Bc6 3.Ra1 b5 4.Ng6 Qd8 ² (0.63) Depth: 11 00:00:20 998kN 1.Qg4 Qe7 2.Bd4 Rac8 3.f5 e5 4.Be3 ² (0.58) Depth: 12 00:00:33 1642kN 1.Qg4 Qe7 2.Bd4 Rac8 3.Qd1 Qd7 4.Rc2 b5 5.Qg4 Qe7 6.Qd1 Qd7 7.Qg4 Qe7 ² (0.59) Depth: 13 00:01:01 3022kN 1.Qg4 Qe7 2.Rc2 Bc6 3.Bd4 Bd7 4.f5 e5 5.Be3 Rec8 6.Ra1 ² (0.69) Depth: 14 00:02:42 7712kN 1.Qg4 Qe7 2.Rc2 Rac8 3.Ra2 Ra8 4.Bd4 ± (0.72) Depth: 15 00:06:28 18045kN 1.Qg4 Qe7 2.Rc2 Bc6 3.Rd2 Bb7 4.Ng6 Qd8 5.e5 f5 ± (0.93) Depth: 16 00:18:19 50018kN 1.Bxf6 gxf6 2.Qg4+ Kh8 3.Qg6 Re7 4.Qxf6+ Kg8 5.Qg6+ ± (1.10) Depth: 16 00:20:54 56938kN New game - Rybka 1.01 Beta 13 32-bit, Blitz:10' Swadlincote r3r1k1/1bq2np1/1p1ppp1p/7B/nPP1PP1N/6P1/4Q2P/B1R1R1K1 w - - 0 1 Analysis by Rybka 1.01 Beta 13 US 32-bit: 256 mb hash 1.Bd4 = (-0.05) Depth: 3 00:00:00 1.Qg4 ² (0.36) Depth: 3 00:00:00 1.Qg4 ² (0.45) Depth: 4 00:00:00 1.Qg4 Qe7 ² (0.47) Depth: 5 00:00:00 1.Qg4 Qe7 2.Bd4 ² (0.47) Depth: 6 00:00:00 11kN 1.Qg4 Qe7 2.Bd4 Ra7 ² (0.49) Depth: 7 00:00:00 29kN 1.Qg4 Qe7 2.Bd4 Ra7 3.Ra1 ² (0.54) Depth: 8 00:00:01 83kN 1.Qg4 Qe7 2.Bd4 Bc6 3.Rc2 Ra7 ² (0.53) Depth: 9 00:00:04 214kN 1.Qg4 Qe7 2.Bd4 Bc6 3.Ra1 b5 4.Ng6 ² (0.59) Depth: 10 00:00:08 398kN 1.Qg4 Qe7 2.Bd4 Bc6 3.Ra1 b5 4.Ng6 Qd8 ² (0.63) Depth: 11 00:00:21 995kN 1.Qg4 Qe7 2.Bd4 Rac8 3.f5 e5 4.Be3 ² (0.58) Depth: 12 00:00:33 1622kN 1.Qg4 Qe7 2.Rc2 Rac8 3.Ra2 ² (0.64) Depth: 13 00:01:19 3917kN 1.Qg4 Qe7 2.Rc2 Rac8 3.Ra2 Ra8 4.Bd4 Bc6 5.Rea1 b5 6.Ng6 ± (0.77) Depth: 14 00:03:01 8855kN 1.Qg4 Qe7 2.Rc2 Rac8 3.Ra2 Ra8 4.Bd4 Bc6 5.Rea1 b5 6.Ng6 Qd8 ± (0.72) Depth: 15 00:04:18 12373kN 1.Qg4 Qe7 2.Rc2 Rac8 3.Ra2 Ra8 4.Bd4 Bc8 5.f5 e5 6.Be3 Bd7 7.Rc1 ² (0.69) Depth: 16 00:07:46 21488kN 1.Bxf6 gxf6 2.Qg4+ Kh8 3.Qg6 Re7 4.Qxf6+ Kg8 5.Bg4 Bc8 6.Qg6+ Kh8 7.f5 Ne5 ± (1.10) Depth: 16 00:09:53 27020kN Depth 14 and below is quicker with 32 mb hash whereas Depth 15 and above is quicker with 256 mb hash on my machine with this articular position. I do not have enough ram to try 512 mb but would guess that would find Bxf6 quicker again. So if you play fast games a lower hash size will be more suitable wheras for long games bigger is better if you have enough ram to spare. Hope this helps. Regards Mike
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.