From noy@tci002.uibk.ac.at Sat Jan 7 01:34:28 1995 Received: from uibk.ac.at for noy@tci002.uibk.ac.at by www.ccl.net (8.6.9/930601.1506) id BAA19592; Sat, 7 Jan 1995 01:27:44 -0500 Received: from tci002.uibk.ac.at by uibk.ac.at with SMTP id AA29514 (5.65c/IDA-1.4.4 for ); Sat, 7 Jan 1995 07:27:19 +0100 Received: by tci002.uibk.ac.at (AIX 3.2/UCB 5.64/CFIBK-2e.AIX) id AA23103; Sat, 7 Jan 1995 07:27:17 +0100 From: noy@tci002.uibk.ac.at (Teerakiat Kerdcharoen) Message-Id: <9501070627.AA23103@tci002.uibk.ac.at> Subject: What makes Gaussian fast ? To: chemistry@ccl.net Date: Sat, 7 Jan 1995 07:27:17 +0100 (NFT) X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 1326 Dear cyberchemists, I have posted this question a few days ago and I will summarize the respond and have them posted to the net very soon. I just would like to add some informations here as requested by some readers, - The review article which I read that Gaussian scale at N**2.7 not at N**4 is Richard A. Friesner (1991) "New Methods for Calculations on Large Molecules". Annu.Rev.Phys.Chem. 42: 341-67. - My system is Li(I)6(NH3). Dunning's DZP bases are being used to perform Hartree-Fock calculation. The calculation is undertaken on IBM RS6000/550 and here is the result: G92 30 minutes 59.2 seconds E = -344.688161 GAMESS 4 hours 48 minutes E = -344.6887325 One can see that the Gaussian is much faster than GAMESS. I have tried running parallel GAMESS on a cluster and it is still slower than non-parallel Gaussian. The total energy is in Hartree unit. One can see that Gaussian speed up at the loss of accuracy but I am still satisfied because the interaction energy (comparative energy) from both programs remain the same (consistency). take care, Teerakiat From JAS@CHEM.UNI.WROC.PL Sat Jan 7 10:34:37 1995 Received: from sun1000.ci.pwr.wroc.pl for JAS@CHEM.UNI.WROC.PL by www.ccl.net (8.6.9/930601.1506) id JAA20923; Sat, 7 Jan 1995 09:43:54 -0500 Received: from hera.math.uni.wroc.pl (hera.math.uni.wroc.pl [156.17.86.1]) by sun1000.ci.pwr.wroc.pl (8.6.9/8.6.9) with SMTP id PAA06714 for ; Sat, 7 Jan 1995 15:45:32 +0100 Received: from chem.uni.wroc.pl (ichuwr.chem.uni.wroc.pl) by hera.math.uni.wroc.pl (5.0/SMI-SVR4) id AA28069; Sat, 7 Jan 1995 15:40:32 --100 Received: from ICHUWR/SMTPQueue by chem.uni.wroc.pl (Mercury 1.11); Sat, 7 Jan 95 15:44:31 +0100 Received: from Mailqueue by ICHUWR (Mercury 1.11); Sat, 7 Jan 95 15:44:19 +0100 From: "Janusz Dzielendziak" To: chemistry@ccl.net Date: Sat, 7 Jan 1995 15:44:17 GMT+1 Subject: Molecular Dynamics of molten salts X-Pmrqc: 1 Priority: normal X-Mailer: Pegasus Mail v3.1 (R1) Message-Id: <4D25996A83@chem.uni.wroc.pl> content-length: 307 Dear Netters, I'm interested in molecular dynamics (or the other simulation) in electrochemistry of molten salts. Maybe somebody help me. Thanx in advance, Janusz Dzielendziak Institute of Chemistry University of Wroclaw Joliot-Curie 14 50-383 Wroclaw phone: +48 71 204345 e-mail: jas@chem.uni.wroc.pl From m10!frisch@uunet.uu.net Sat Jan 7 12:34:37 1995 Received: from relay3.UU.NET for m10!frisch@uunet.uu.net by www.ccl.net (8.6.9/930601.1506) id MAA21339; Sat, 7 Jan 1995 12:03:24 -0500 Received: from uucp2.UU.NET by relay3.UU.NET with SMTP id QQxxrg23984; Sat, 7 Jan 1995 12:03:30 -0500 Message-Id: Received: from m10.UUCP by uucp2.UU.NET with UUCP/RMAIL ; Sat, 7 Jan 1995 12:03:24 -0500 Received: by m10.gaussian.com; Sat, 7 Jan 95 11:27:43 EST Date: Sat, 7 Jan 95 11:27:43 EST From: m10!frisch@uunet.uu.net (Michael Frisch) Subject: Re: CCL:What makes Gaussian fast ? To: chemistry@ccl.net In-Reply-To: uunet!tci002.uibk.ac.at!noy (Teerakiat Kerdcharoen), Sat, 7 Jan 1995 07:27:17 +0100 (NFT) Dear cyberchemists, I have posted this question a few days ago and I will summarize the respond and have them posted to the net very soon. I just would like to add some informations here as requested by some readers, - The review article which I read that Gaussian scale at N**2.7 not at N**4 is Richard A. Friesner (1991) "New Methods for Calculations on Large Molecules". Annu.Rev.Phys.Chem. 42: 341-67. - My system is Li(I)6(NH3). Dunning's DZP bases are being used to perform Hartree-Fock calculation. The calculation is undertaken on IBM RS6000/550 and here is the result: G92 30 minutes 59.2 seconds E = -344.688161 GAMESS 4 hours 48 minutes E = -344.6887325 One can see that the Gaussian is much faster than GAMESS. I have tried running parallel GAMESS on a cluster and it is still slower than non-parallel Gaussian. The total energy is in Hartree unit. One can see that Gaussian speed up at the loss of accuracy but I am still satisfied because the interaction energy (comparative energy) from both programs remain the same (consistency). When doing just a single-point SCF calculation, Gaussian 92 defaults to doing the calculation to sufficient accuracy for the purposes for which people normally do single-point calculations -- energy comparisons to a few tenths of a kcal, potential-derived charges to a few digits, etc. If you say SCF=Tight on a single-point calculation, you will get the full convergence. (Full convergence is the default for anything but a HF single-point calculation -- HF optimizations, MP2 energies, etc.) Full convergence is about a factor of 2 slower than the cheap convergence. The G92 (SCF=Tight) and GAMESS energies should agree to 8 figures at least. Mike Frisch Gaussian, Inc. -------