From owner-chemistry@ccl.net Mon Jan 3 09:05:00 2022 From: "Detlev Conrad Mielczarek detlevcm~!~googlemail.com" To: CCL Subject: CCL:G: Estimated job time in Gaussian Message-Id: <-54567-220103052832-828-YdLAKeclsz81pc6UEe4i3A a server.ccl.net> X-Original-From: Detlev Conrad Mielczarek Content-Language: en-US Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=UTF-8; format=flowed Date: Mon, 3 Jan 2022 11:28:20 +0100 MIME-Version: 1.0 Sent to CCL by: Detlev Conrad Mielczarek [detlevcm_._googlemail.com] Estimating compute times is its own area of research and typically very hard to do - not just in ab initio quantum chemistry but also in other fields, say CFD. Under certain circumstances, it is possible to look at the time used for the initial iterations and thus extrapolate the required time for a set number of iterations to get a rough idea. E.g. if I know I complete 200 CFD iterations a day, I will possibly do 1000 in 5 days. Having said that, CFD convergence is possibly a lot more straightforward than ab initio calculations. Just the choice of method and basis will have a dramatic impact on the calculation time. Then small perturbations in the starting geometry can equally impact the calculation time for more complex geometries as its can impact the trajectory selected during the optimization. For very simple molecules, it may be possible to establish correlations and thus make accurate predictions, but these simple molecules will either require seconds (for teaching purposes with simple methods and small basis sets) or be subject to very high accuracy calculations in research where time matters less. Still, this could be a subject for a research project... Detlev ORCiD 0000-0002-4720-353X On 31/12/2021 23:14, Andrew DeYoung andrewdaviddeyoung*gmail.com wrote: > Sent to CCL by: "Andrew DeYoung" [andrewdaviddeyoung\a/gmail.com] > Hi, > > Does anyone know if it is possible for the Gaussian package to estimate the > time remaining in a relatively pedestrian calculation (e.g., Opt Freq)? I'm > not seeing a discussion of this in the documentation. I'm running Gaussian > 2016 (G16). > > Thanks in advance for any insight you can provide, > > Andrew > > Andrew DeYoung, PhD > Department of Chemistry > Carnegie Mellon University > andrewdaviddeyoung###gmail.com> > From owner-chemistry@ccl.net Mon Jan 3 09:40:00 2022 From: "Milica Feldt mfeldt*_*uni-muenster.de" To: CCL Subject: CCL:G: Virtual Winter School for Computational Chemistry Message-Id: <-54568-220103061134-7111-GmEvQi5rV2umswCkWUQZyQ-,-server.ccl.net> X-Original-From: "Milica Feldt" Date: Mon, 3 Jan 2022 06:11:32 -0500 Sent to CCL by: "Milica Feldt" [mfeldt ~ uni-muenster.de] Dear colleague, Registrations are now open for the 8th Virtual Winter School for Computational Chemistry which will take place 21-25th February 2022. Full information and the registration page can be found here: www.winterschool.cc Registration is free for all participants. The Virtual Winter School for Computational Chemistry covers a broad range of topics in computational and theoretical chemistry. The extended lecture format allows ample time to cover both introductory material on topics as well as the latest research developments. There will also be two hands on workshops on how to use quantum chemical programs, including Gaussian and ADF this year. Confirmed speakers for the 2022 edition include: Professor Roald Hoffman (Cornell University) Dr Joaquin Barroso (National Autonomous University of Mexico) Dr Stephane Irle (Oak Ridge National Laboratory) Professor Satoshi Maeda (Hokkaido University) Professor Daniel Crawford (Virginia Tech) Professor Anastasia V. Bochenkova (Lomonosov Moscow State University) Dr Nicole Holzmann (Riverlane) Professor Gyrgy M Keser (Research Center for Natural Sciences) Professor Takeshi Yanai (Nagoya University) Professor Jeremy Harvey (KU Leuven) Professor Samer Gozem (Georgia State University) Professor Carla de Figuria (University College London) We look forward to welcoming you at the 2022 event. Virtual Winter School for Computational Chemistry Organising Committee From owner-chemistry@ccl.net Mon Jan 3 10:42:00 2022 From: "Rzepa, Henry S h.rzepa~~imperial.ac.uk" To: CCL Subject: CCL: Estimated job time in Gaussian Message-Id: <-54569-220103102234-14463-n006ofG+K5qW+zu613JYww---server.ccl.net> X-Original-From: "Rzepa, Henry S" Content-ID: <5C1D9C209DFE794BAF78FD98B89549E4---eurprd06.prod.outlook.com> Content-Language: en-US Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="us-ascii" Date: Mon, 3 Jan 2022 15:22:19 +0000 MIME-Version: 1.0 Sent to CCL by: "Rzepa, Henry S" [h.rzepa*imperial.ac.uk] In terms of estimating time, geometry optimisation is in itself a black art. Sometimes the accuracy limits, either set by default or by the user, can greatly influence the time taken, as can the coordinate system chosen. It is always useful to "tail" the job output and take a look at the number of geometry cycles and whether the preset convergence criteria are close to being met, or in fact oscillating (which can happen). Convergence of other algorithms (eg 2nd derivatives, VCD etc etc) can depend greatly on integral accuracy (ie acc2e=14 often solves that problem) and other presets. So there is really no general answer to estimating job times. Henry > On 3 Jan 2022, at 10:28, Detlev Conrad Mielczarek detlevcm~!~googlemail.com wrote: > > > Sent to CCL by: Detlev Conrad Mielczarek [detlevcm_._googlemail.com] > Estimating compute times is its own area of research and typically very hard to do - not just in ab initio quantum chemistry but also in other fields, say CFD. >