From chemistry-request@server.ccl.net Wed Dec 16 11:10:03 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id LAA02033 for ; Wed, 16 Dec 1998 11:10:03 -0500 Received: from majuelo.sdi.uam.es (majuelo.sdi.uam.es [150.244.9.25]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id KAA03172 Wed, 16 Dec 1998 10:54:59 -0500 (EST) Received: from roble.sdi.uam.es (roble.sdi.uam.es [150.244.9.71]) by majuelo.sdi.uam.es (8.8.5/8.8.5) with SMTP id QAA29906 for ; Wed, 16 Dec 1998 16:53:48 +0100 (MET) Received: from brihuega ([150.244.37.59]) by chopo.sdi.uam.es with SMTP for chemistry@www.ccl.net; Wed, 16 Dec 1998 16:53:42 +0100 Message-Id: <1.5.4.16.19981216154917.1dcf8742@bosque.sdi.uam.es> X-Sender: depaz@bosque.sdi.uam.es X-Mailer: Windows Eudora Light Version 1.5.4 (16) Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Wed, 16 Dec 1998 16:49:17 +0100 To: chemistry@www.ccl.net From: Dr Jose Luis Garcia de Paz Subject: CIS and solvent calculation Hi All, Season Greetings! Does anybody knows a program able to perform a calculation of the excited state (using CIS model) of a molecule into a solvent (i.e using the Onsager model)? Thank you very much. Regards Jose _____________ Dr. Jose Luis Garcia de Paz |_ / Departamento de Quimica Fisica | | Facultad de Ciencias, C-XIV-602 | 0 / Universidad Autonoma de Madrid < | Telephon: +34-91-397.4957 or 4263 Ctra de Colmenar Km. 15 |_ ___/ Fax: +34-91-397.4512 E-28049 Madrid (SPAIN) \/ E-mail: alt E-mail: http://www.uam.es/qfa/depaz.html http://www.adi.uam.es/~depaz From chemistry-request@server.ccl.net Wed Dec 16 11:10:05 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id LAA02046 for ; Wed, 16 Dec 1998 11:10:05 -0500 Received: from alchemy.chem.utoronto.ca (alchemy.chem.utoronto.ca [142.150.224.224]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id MAA26440 Tue, 15 Dec 1998 12:57:54 -0500 (EST) Received: (from elewars@localhost) by alchemy.chem.utoronto.ca (8.7.4/8.7.3) id MAA02387 for chemistry@www.ccl.net; Tue, 15 Dec 1998 12:57:51 -0500 (EST) Date: Tue, 15 Dec 1998 12:57:51 -0500 (EST) From: "E. Lewars" Message-Id: <199812151757.MAA02387@alchemy.chem.utoronto.ca> To: chemistry@www.ccl.net Subject: CALCULATING A RATE COSTANT Tued 1998 Dec 15 Subject: Calculating rate constant Hello, Has anyone any suggestions for a fairly easy way to calculate a rate constant using Gaussian? I want to avoid the intricacies of RRKM theory and just get a k_rate accurate to within a factor of, say, 10. For example, consider HNC ---> HCN At some temp T HNC in, say, the gas phase (or nonpolar solvent?) has a certain halflife (=ln2/k_r). It's easy to calculate a good number for the activation energy, but what's a simple way to estimate a realistic k_r (and thus t_1/2)? Thanks E. Lewars ================ From chemistry-request@server.ccl.net Wed Dec 16 11:10:06 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id LAA02049 for ; Wed, 16 Dec 1998 11:10:05 -0500 Received: from euch4e.chem.emory.edu (euch4e.chem.emory.edu [170.140.59.12]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id NAA28344 Tue, 15 Dec 1998 13:17:14 -0500 (EST) Received: from euch4e.chem.emory.edu ([170.140.49.160]) by euch4e.chem.emory.edu (AIX4.2/UCB 8.7/8.7) with ESMTP id NAA44800 for ; Tue, 15 Dec 1998 13:17:23 -0500 (EST) Message-ID: <36769AF4.AE3E6E34@euch4e.chem.emory.edu> Date: Tue, 15 Dec 1998 13:23:07 -0400 From: Maricel Reply-To: maricel@euch4e.chem.emory.edu Organization: Emory University X-Mailer: Mozilla 4.04 (Macintosh; I; PPC) MIME-Version: 1.0 To: CHEMISTRY@www.ccl.net Subject: Re: electric field gradient unit Content-Type: text/plain; charset=us-ascii; x-mac-type="54455854"; x-mac-creator="4D4F5353" Content-Transfer-Encoding: 7bit On Tue, 15 Dec 1998 15:49:12 +0100 (MEZ) Jerry wrote: > Dear CClers, > > In an old article [C. B. Harris, Inorg. Chem., vol 7, 1517 (1968)] the > electric field gradient at the Co nucleus of (C5H5)2CoClO4 is reported as > qzz = 5.85 x 10^15 esu/cm^3. Could anyone tell me how I can convert this > quantity to the quadrupole coupling constant, e^2Qqzz, in MHz? The > quadrupole moment of 59Co is 0.404 x 10^28 m^2. > > Many thanks, > Jerry > > ****************************************************************** > * Jerry Chun Chung CHAN chan@uni-muenster.de * > * Universitaet Muenster phone: 0049-251-83-29156 * > * Institut fuer Physikalische Chemie fax: 0049-251-83-29159 * > * Schlossplatz 4-7 * > * D-48149 Muenster * > * Germany * > ****************************************************************** > Dear Jerry, Please, notice that the whole expression for the nuclear quadrupolar coupling constant (NQCC) is X = e**2Qq/h (in frequency units) where h is Planck's constant (= 6.6260755 x 10E-34 J sec.), e is the electron charge (= 1.602188 x 10E-19 Coul = = 4.803242 x 10E-10 ESU), eQ is the nuclear electric quadrupole moment, and eq is the electric field gradient (EFG). It should be noted that your "qzz" (the largest principal component of the diagonalized EFG tensor) is already "eq" in the above expression. Having all these data into account, you will be able to easily convert the EFG into the NQCC tensor (in MHz) by applying the following conversion factor: 1 a.u. of qzz = e**(-1)a_o**(-2)E_h = = 9.71736 x 10E21 V m-2 = = 3.24136 x 1oE15 esu cm-3 (4 pi epsilon_o)**(-1) More details about the quantities of your interest and their conversion can be found in J. Chem. Phys. 1998, 109, 7176-7184. Hope it helps, Maricel **************************************************** Dr. Maricel Torrent, Research Associate * Present address: Cherry L. Emerson Center for Scientific Computation and Department of Chemistry, Emory University, Atwood Hall, 1515 Pierce Drive, Atlanta, Georgia 30322, USA Phone: +1-404-727-6634 Fax: +1-404-727-7412 E-mail: maricel@euch4e.chem.emory.edu * Private address: 765-5 Houston Mill Road, N.E. Atlanta, Georgia 30329, USA Phone: +1-404-325-1011 **************************************************** From chemistry-request@server.ccl.net Wed Dec 16 11:10:09 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id LAA02067 for ; Wed, 16 Dec 1998 11:10:08 -0500 Received: from hera.math.uni.wroc.pl (hera.math.uni.wroc.pl [156.17.86.1]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with SMTP id IAA07651 Tue, 15 Dec 1998 08:13:22 -0500 (EST) Received: from wchuwr.chem.uni.wroc.pl [156.17.103.193] by hera.math.uni.wroc.pl with esmtp (Exim 1.70 #1) id 0zpuHg-0003uF-00; Tue, 15 Dec 1998 14:13:04 +0100 Received: from ICHUWR/SpoolDir by wchuwr.chem.uni.wroc.pl (Mercury 1.43); 15 Dec 98 14:11:49 MET Received: from SpoolDir by ICHUWR (Mercury 1.43); 15 Dec 98 14:11:46 MET From: "Adrian Radomir Jaszewski" Organization: University of Wroclaw (Chemistry) To: chemistry@www.ccl.net Date: Tue, 15 Dec 1998 14:11:39 MET Subject: EPR-2 and EPR-3 basis sets X-Confirm-Reading-To: "Adrian Radomir Jaszewski" X-pmrqc: 1 Priority: normal X-mailer: Pegasus Mail/Windows (v1.22) Message-ID: <8134013B8A@wchuwr.chem.uni.wroc.pl> Dear CCLers, I know that EPR-2 basis set has been published in 'Recent Advances in Density Functional Methods' by Vincenzo Barone (edited by D. P. Chong); World Scientific, Singapore, 1995. Unfortunately, I haven't an access to this book. Vincenzo Barone has also published exponents and contraction coefficients of the EPR-3 basis set for hydrogen and carbon atom in J.Chem.Phys. 105 (24), 1996, p.11060. In this paper he gives additional exponents of diffuse s functions for non-hydrogen atoms in the EPR-2 basis set. I would like to know all exponents and contraction coefficient of the EPR-2 and EPR-3 basis sets for hydrogen, carbon, nitrogen and oxygen atoms (with exponents of diffuse functions). I will be very grateful for any information on this theme. Best regards, Adrain =============================================== Adrian R. Jaszewski Faculty of Chemistry UNIVERSITY OF WROCLAW 14 Joliot-Curie St. 50-383 Wroclaw Poland e-mail: adrian@wchuwr.chem.uni.wroc.pl ================================================ From chemistry-request@server.ccl.net Wed Dec 16 11:10:07 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id LAA02053 for ; Wed, 16 Dec 1998 11:10:06 -0500 Received: from moorea.uni-muenster.de (MOOREA.UNI-MUENSTER.DE [128.176.126.33]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id JAA08793 Tue, 15 Dec 1998 09:49:41 -0500 (EST) Received: from localhost (chan@localhost) by moorea.uni-muenster.de (8.8.7/8.7.5) with ESMTP id OAA20794 for ; Tue, 15 Dec 1998 14:49:14 GMT Date: Tue, 15 Dec 1998 15:49:12 +0100 (MEZ) From: "Jerry C.C. Chan" To: CHEMISTRY@www.ccl.net Subject: electric field gradient unit In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear CClers, In an old article [C. B. Harris, Inorg. Chem., vol 7, 1517 (1968)] the electric field gradient at the Co nucleus of (C5H5)2CoClO4 is reported as qzz = 5.85 x 10^15 esu/cm^3. Could anyone tell me how I can convert this quantity to the quadrupole coupling constant, e^2Qqzz, in MHz? The quadrupole moment of 59Co is 0.404 x 10^28 m^2. Many thanks, Jerry ****************************************************************** * Jerry Chun Chung CHAN chan@uni-muenster.de * * Universitaet Muenster phone: 0049-251-83-29156 * * Institut fuer Physikalische Chemie fax: 0049-251-83-29159 * * Schlossplatz 4-7 * * D-48149 Muenster * * Germany * ****************************************************************** From chemistry-request@server.ccl.net Wed Dec 16 13:27:18 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id NAA02763 for ; Wed, 16 Dec 1998 13:27:17 -0500 Received: from mx.seanet.com (dns2.seanet.com [199.181.164.2]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id NAA08400 Wed, 16 Dec 1998 13:24:35 -0500 (EST) Received: from biggeek ([150.215.151.125]) by mx.seanet.com (8.8.8/Seanet-8.7.3) with SMTP id KAA00521; Wed, 16 Dec 1998 10:24:21 -0800 (PST) Message-ID: <001201be2921$50dd42d0$7d97d796@biggeek.wrq.com> From: "Mark Thompson" To: , "Dr Jose Luis Garcia de Paz" Subject: Re: CCL:CIS and solvent calculation Date: Wed, 16 Dec 1998 10:24:21 -0800 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 4.72.2106.4 X-MimeOLE: Produced By Microsoft MimeOLE V4.72.2106.4 Yes, ArgusLab will do this using the semiempirical INDO/s method. It's free. Download from: http://www.seanet.com/~mthompson/ArgusLab/index.htm Mark Thompson -----Original Message----- From: Dr Jose Luis Garcia de Paz To: chemistry@www.ccl.net Date: Wednesday, December 16, 1998 8:25 AM Subject: CCL:CIS and solvent calculation >Hi All, > >Season Greetings! > >Does anybody knows a program able to perform >a calculation of the excited state (using CIS model) >of a molecule into a solvent (i.e using the Onsager >model)? > >Thank you very much. Regards >Jose > > _____________ >Dr. Jose Luis Garcia de Paz |_ / >Departamento de Quimica Fisica | | >Facultad de Ciencias, C-XIV-602 | 0 / >Universidad Autonoma de Madrid < | Telephon: +34-91-397.4957 or 4263 >Ctra de Colmenar Km. 15 |_ ___/ Fax: +34-91-397.4512 >E-28049 Madrid (SPAIN) \/ E-mail: > alt E-mail: >http://www.uam.es/qfa/depaz.html >http://www.adi.uam.es/~depaz > > >-------This is added Automatically by the Software-------- >-- Original Sender Envelope Address: chemistry-request@www.ccl.net >-- Original Sender From: Address: depaz@uam.es >CHEMISTRY@www.ccl.net: Everybody | CHEMISTRY-REQUEST@www.ccl.net: Coordinator >MAILSERV@www.ccl.net: HELP CHEMISTRY or HELP SEARCH | Gopher: www.ccl.net 73 >Anon. ftp: www.ccl.net | CHEMISTRY-SEARCH@www.ccl.net -- archive search > Web: http://www.ccl.net/chemistry.html > > From chemistry-request@server.ccl.net Wed Dec 16 15:59:20 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id PAA02846 for ; Wed, 16 Dec 1998 15:59:20 -0500 Received: from kappa.ucdavis.edu (kappa.ucdavis.edu [169.237.38.55]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with SMTP id PAA23414; Wed, 16 Dec 1998 15:56:36 -0500 (EST) Received: (from daizadeh@localhost) by kappa.ucdavis.edu (950413.SGI.8.6.12/950213.SGI.AUTOCF) id MAA25524; Wed, 16 Dec 1998 12:56:36 -0800 Date: Wed, 16 Dec 1998 12:56:36 -0800 (PST) From: Iraj Daizadeh To: chemistry@www.ccl.net, chemistry-request@www.ccl.net Subject: Normal Mode Coordinate Constraint Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hello. We would like to perform geometry optimization keeping particular normal mode coordinates, Q_i, constant. Is this possible in G94? Please direct reponses to emile@omega.ucdavis.edu Thank you, Emile From chemistry-request@server.ccl.net Wed Dec 16 16:56:48 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id QAA02905 for ; Wed, 16 Dec 1998 16:56:48 -0500 Received: from mcmail.cis.McMaster.CA (mattacf@mcmail.CIS.McMaster.CA [130.113.20.6]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id QAA25134 Wed, 16 Dec 1998 16:54:04 -0500 (EST) Received: from localhost (mattacf@localhost) by mcmail.cis.McMaster.CA (8.8.8/8.8.8) with SMTP id QAA15786 for ; Wed, 16 Dec 1998 16:53:57 -0500 (EST) Date: Wed, 16 Dec 1998 16:53:57 -0500 (EST) From: "C.F. Matta" To: chemistry@www.ccl.net Subject: SUMMARY: G94 under Linux Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hi there, Many thanks to all those who took the time to respond to my question: D. Mariano A. Vera, Joseph W. Bausch, John McKelvey, Luis Antonio Barbosa, Morten Langgrd, Bernd Melchers, Douglas J. Fox, and Stefan Fau. Their responses are listed below. The replies seem to point to the motherboard speed as the likely bottleneck. If it is replaced, I will inform you of the result - refering to this series of emails. Thank you all again. Cherif Matta Grad. student, Chemistry Dept. McMaster University, Hamilton, Ontario, Canada, L8S 4M1 =========================================================================== ORIGINAL QUESTION: ================== I am running Gaussian94 and Atoms-In-Molecules (AIM) jobs on a PC and they all seem to run too slow (both in real time and in CPU time)for the hardware specs. Where is the bottleneck, and what hardware part(s) can be changed to speed the calculations? I will summarize. Thank you in advance. SPECIFICATIONS: =============== Operating system: Red Hat Linux 5.1 Processor: single Intel Pentium II MMX - 300MHz Mother board: AX6BC External bus clock: 66MHz Cache: 512K RAM: 320M ( 2x128M + 64M) 10-nanoseconds DIMMs Swap space: 237M (in two partitions, 117M and 120M) Hard drive: 6.4G IDE (Quantum) ========================================================================== REPLIES: ======== (1) From: "Lic. Domingo M. Vera" =============================================================== I am running G94 in two Linux boxes (Intel Pentium platformsB) and I find quite surprising your trouble since we are very happy with the perfomances. I have two observations about your hardware, for I/O intensive jobs you have two bottlenecks: a 66MHz bus and IDE disks. An SCSI controller is the natural choice in order to take the 100% of the possibilities of your processor running g94. I suggest the following: 1) Consider to recompile the kernel without K6 and R1000 processors bux fixing options (this is not required since you have an Intel Pentium). 2) Are you sure that the kernel detects your 320MB (issue the 'free' command)? Some systems needs the following options in the lilo.conf file as boot parameter append=`mem=320M` 3) I have another Pentium at home, a 300Mz one with IDE disks (very similar to your configuration). You can send me a (short) job for testing. 4) Consider an upgrade to Red Hat Linux 5.2. I done an upgrade last week so I'm not sure but it seems to be faster running gaussian jobs. Naturally, you must recompile the code. D. Mariano A. Vera PhD student at Computational Chemistry and Electron Transfer Group Department of Organic Chemistry. Faculty of Chemistry Sciences. Universidad Nacional de Cordoba, Cordoba. ARGENTINA. Voice: 54-51-334170 Fax: 54-51-333030 mariano@zeus.fcq.unc.edu.ar (2) From: "Joseph W. Bausch" ==================================================== I need to know some run times for standard jobs to comment further. Perhaps you can run all seven of the G94 test jobs that Gaussian recommends to run, and summarize those times. If you do so, some of these jobs are actually many small calcs wrapped into one, so be sure to sum up all the run times in the output. If you do this, I can tell you what a "good" PII 300 MHz system should give, because I have these numbers tabulated. My guess would be if in fact your machine is slow, that the problem is the motherboard. Anyway, let me know if I can be of further help. (3) From: "Joseph W. Bausch" =================================================== >CPU times are summarized below: > > log file coming Cherif's Joseph's 300 PII comming with G94 machine machine > ---------------- ------- -------------- >test 001 4.2 sec 3.5 sec 3.9 >test 028 10.4 sec 12.1 sec 11.2 >test 094 39.3 sec 47.3 sec 38.7 >test 155 1 min 4.0 sec 2 min 23.1 sec 221.4* >test 194 1 min 1.4 sec 2 min 58.9 sec 167.4 >test 296 job1 23.6 sec 28.5 sec 757* (all jobs) > job2 18.9 sec 27.4 sec > job3 31.9 sec 45.9 sec > job4 14.1 sec 35.1 sec >test 302 52.0 sec 59.1 sec 52 > >TOTAL 319.8 sec 580.9 sec * Note, I don't think all parts of your jobs 155 and 296 ran. You should check this. What I can see seems to be that your machine is about what is expected for a PII 300 MHz. My machine I think had a scsi hard disk, and this might give slightly better numbers than say an IDE based drive. (4) From: John McKelvey ================================================= Change to fast-wide SCSI disk, perhaps? I WOULD BE QUITE INTERESTED IN YOUR REPLIES, AS I AM FACING THE SAME TYPE OF ISSUE... EXCEPT WITH JAGUAR [ A _LOT_ FASTER THAN g98 ON AN SGI IN SINGLE PROCESSOR MODE FOR MY WORK] (5) From: Luis Antonio Barbosa ======================================================== It is interesting to know that Gaussian94 is slow in you case. Do you mind to tell me the size of your system? I would like to ask also some questions: * Which compiler did you use to create the executable code? * Which math libraries did you use? I checked in the Gaussian web, at least for the new version of Gaussian (98), it requires a special compiler and math libraries. (6) From: "Morten Langgrd" ========================================================== You will probably receive hundreds of answers to your question since there may be many possible bottlenecks involved. Regarding G94: Slow performance are most likely caused by: bad compilation, inefficient use of the "mem" settings or your IDE drive. The IDE drive is only a question if your job uses the read-write files extensively. I would recommend that you get hand on this paper: "Computational Chemistry on Commodity-Type Computers", Nicklaus, M.C. et al., J. Chem. Inf. Comput. Sci. 38(5), 893-905 (1998). http://pubs.acs.org/journals/jcisd8/index.html. In this paper you will find a lot of useful discussion of Linux related compilations of G94 and benchmark results. From these figures you can easily judge if your systems has a particular bad or just normal preformance. Morten Langgrd, Dept. 814, Computational Chemistry. H. Lundbeck A/S, Ottiliavej 9 Dk-2500 Valby, Copenhagen. Denmark Phone: (+45) 36 44 24 25 tone 2936, Fax: (+45) 36 30 13 85 e-mail: mol@lundbeck.com (7) From: Bernd Melchers ======================================================= If you have i/o intensive jobs (great rwf or checkpoint files), you should verify that you configured your kernel for Busmaster-DMA drivers, if it is not the case (try "hdparm -d /dev/hda" and read the manpage hdparm(1)). Or you could use (expensive) SCSI-disks. You should configure the memory usage of gaussian (sometimes less memory results in increasing speed). You could use another compiler (g77 or f2c + pgcc or f2c + egcs). This is an option for experienced users only, as pgcc and egcs may have some compiler errors and gaussian is not tested for these compilers. See: ftp://ftp.goof.com/pub/pcg/source/ ftp://egcs.cygnus.com:/pub/egcs/ Sometimes I read, that some glibc implementations has slow but more precise math-functions. You can try to compile gaussian on a libc-5 system and compare the speed results. Bernd Melchers | melchers@FU-Berlin.DE Freie Universitaet Berlin | "We don't write software, we compose it." AG Macromolecular Modelling - Prof. Dr. E.W. Knapp for more information see http://userpage.chemie.fu-berlin.de/~melchers/ (8) From: Doug Fox ==================================================== I don't see how to address your question without some details on what performance you see and what performance you expect. For example we improved performance between G94 and G98 by improving the compiler, we utilize the Portland Group compiler, which understands the Pentium II instruction set. This is was good for a fairly large factor, up to about a factor of 2 in some places, but if you are missing a factor of 10 there may be more to this. On a similar note does the performance difference vary with problem size? How? Douglas J. Fox Director of Technical Support help@gaussian.com (9) From: Doug Fox ==================================================== A couple of quick comments. Memory bandwidth, i.e. how quickly you can access memory is better for the workstations than for PC based systems. This is not an issue for small cases but as the problem size goes up it will become a bottleneck. Also for any given case there is an optimal memory and increasing memory beyond that may actually cause a small decrease due to overhead in allocating memory that is never used. Generally this is small but on very small cases it can be evident. In general water is not a good timing case any more for estimating general performance. Pick something with 10-15 minutes minimum on the faster platform. A factor of 3 does not seem out of range for an Octane over a 300MHz Pentium. The very small cases willdo better but they are not too interesting. Douglas J. Fox Director of Technical Support help@gaussian.com (10) From: Stefan Fau ================================================= I'd like to throw in the experiences one of my colleagues has made. G94 was not nice to compile on our (non-standard-)LINUX PCs, using several FORTRAN compilers. It had unexplainable performance losses with DFT, MP3 and MP4 (it would use up to 20fold times) or would not run at all. Our hints are: use a Fortran compiler and PII optimized BLAS routines (pointer at GAUSSIAN.com) With _G98_ (and new PCs) we made better experiences. On a 400 MHz PII with 512 MB RAM (compiled using PII optimized BLAS routines and PGF77) the runtimes are comparable to rs10k-machines (-20% to +30%, depending on the link). Dr. Stefan Fau Fachbereich Chemie, AK Frenking Philipps-Universitt Marburg 35032 Marburg, Germany fau@chemie.uni-marburg.de ========================================================================== From chemistry-request@server.ccl.net Wed Dec 16 19:01:34 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id TAA02958 for ; Wed, 16 Dec 1998 19:01:33 -0500 Received: from alchemy.chem.utoronto.ca (alchemy.chem.utoronto.ca [142.150.224.224]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id SAA27843 Wed, 16 Dec 1998 18:58:48 -0500 (EST) Received: (from elewars@localhost) by alchemy.chem.utoronto.ca (8.7.4/8.7.3) id SAA20784 for chemistry@www.ccl.net; Wed, 16 Dec 1998 18:58:48 -0500 (EST) Date: Wed, 16 Dec 1998 18:58:48 -0500 (EST) From: "E. Lewars" Message-Id: <199812162358.SAA20784@alchemy.chem.utoronto.ca> To: chemistry@www.ccl.net Subject: SUMMARY OF ANSWERS TO PAULING POINT 1998 Dec 15 PAULING POINT Thanks to all who replied to my question about the term _Pauling Point_: who first used it and if the Pauling is Linus Pauling. Apparently no one knows who first used it, but there is agreement that it refers to Linus Pauling. E. Lewars ANSWERS, MY SUMMARY (if any one has been omitted, it's my carelessness) #1 Masao Masamura Yes, 99% sure (it's Linus Pauling) #2 Alan Shusterman The idea is that a graph of understanding vs. effort rises steeply then goes down a bit then rises again. Pauling was an expert in getting good understanding >from moderate effort. #3 Michael Nolan Probably find it in Pauling's "The Nature of the Chemical Bond" #4 Georg Schreckenbach L Pauling has been great at locating such points. Attempts by his successors to improve on this tend to make matters worse. ======== From chemistry-request@server.ccl.net Wed Dec 16 19:02:37 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id TAA03002 for ; Wed, 16 Dec 1998 19:02:37 -0500 Received: from alchemy.chem.utoronto.ca (alchemy.chem.utoronto.ca [142.150.224.224]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id SAA27859 Wed, 16 Dec 1998 18:59:51 -0500 (EST) Received: (from elewars@localhost) by alchemy.chem.utoronto.ca (8.7.4/8.7.3) id SAA20828 for chemistry@www.ccl.net; Wed, 16 Dec 1998 18:59:52 -0500 (EST) Date: Wed, 16 Dec 1998 18:59:52 -0500 (EST) From: "E. Lewars" Message-Id: <199812162359.SAA20828@alchemy.chem.utoronto.ca> To: chemistry@www.ccl.net Subject: SUMMARY OF ANSWERS TO DIAGONALIZATION OF LAGRANGIAN MATRIX 1998 Dec 15 PAULING POINT Thanks to all who replied to my question about diagonalizing the matrix of undetermined Lagrangian multipliers in one derivation of the Hartree-Fock eqs. Essentially I asked how do we know that this matrix _is_ diagonalizable (not all square matrices are). E. Lewars ANSWERS, MY SUMMARY (if any one has been omitted, it's my carelessness) #1 Alan Shusterman The matrix is Hermitian because its elements are integrals involving a Hermitian operator (the Hamiltonian). #2 Darko Babic The matrix in my case is Hermitian. #3 Eric Bittner Think like this: L = T -V T and V are Hermitian, so L is Hermitian. ============ From chemistry-request@server.ccl.net Wed Dec 16 23:37:33 1998 Received: from www.ccl.net (www.ccl.net [192.148.249.5]) by server.ccl.net (8.8.7/8.8.7) with ESMTP id XAA03288 for ; Wed, 16 Dec 1998 23:37:33 -0500 Received: from pnl.gov (galaxy.pnl.gov [130.20.64.11]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id XAA28802 Wed, 16 Dec 1998 23:34:44 -0500 (EST) From: d3e102@emsl.pnl.gov Received: from odyssey.emsl.pnl.gov by pnl.gov (PMDF V5.1-12 #28154) with ESMTP id <01J5F2Z5KBDS97WOVL@pnl.gov> for chemistry@www.ccl.net; Wed, 16 Dec 1998 20:34:37 PST Received: from ames.emsl.pnl.gov (ames.emsl.pnl.gov [198.128.80.135]) by odyssey.emsl.pnl.gov (8.8.5/8.8.5) with ESMTP id UAA11710 for ; Wed, 16 Dec 1998 20:34:34 -0800 (PST) Received: (from d3e102@localhost) by ames.emsl.pnl.gov (8.8.5/8.8.5) id UAA11110 for chemistry@www.ccl.net; Wed, 16 Dec 1998 20:34:31 -0800 (PST) Date: Wed, 16 Dec 1998 20:34:31 -0800 Subject: A Computational Results Database To: chemistry@www.ccl.net Message-id: <9812162034.ZM11108@ames.emsl.pnl.gov> MIME-version: 1.0 X-Mailer: Z-Mail Lite (3.2.0 26may94) Content-type: text/plain; charset=us-ascii A database of ab initio electronic structure calculations is available at URL: http://www.emsl.pnl.gov:2080/proj/crdb/. Please feel free to try it and let me know what your impressions are. This database currently contains over 30,000 pieces of information from calculations performed with small to very large basis sets, using everything >from simple Hartree-Fock theory to full CI. At the moment, we have data on 211 molecules and 41 atoms. New entries are being added at the rate of approximately 5000 per year. The types of information include: o Optimized Structures (Selected Internals and Cartesians) o SMILES, IUPAC and CAS Indices o Electronic Configuration o Harmonic Frequencies o Relative Energetics (Dissociation Energies, etc.) o Total Energies (at Optimized or Expt. Geometries) o Literature Citations o Experimental values. The current implementation is somewhat limited in the flexibility of the searches that can be performed. However, we hope to be able to release another version of the interface sometime next year that will permit greater freedom in the way one searches and more types of information. There are many areas in which the database could be expanded. Perhaps the most obvious is the lack of density functional results. We would like to create a web page to allow people to submit the results of their published work in a semi-automated manner. Approximately 5200 entries corresponding to calculations with the aug-cc-pVQZ, aug-cc-pV5Z and aug-cc-pV6Z basis sets have been locked out of this initial release of the database until papers related to these results are accepted for publication. We anticipate that the excluded results will be made available sometime in late January. I personally want to thank Ms. Karen Schuchardt for implementing the web interface. I would also like to thank the following people for contributing the results of their work: Drs. Thom Dunning, Jr., Dave Dixon, Kirk Peterson, Dave Woon, Ricky Kendall and Sotiris Xantheas. I've benefitted from numerous conversations with Drs. Thom Dunning, Jr. and Michel Dupuis. Any mistakes you find (and you WILL find some) are almost certainly mine. Dave -- David Feller | Mail Stop K1-96 Environmental Molecular Sciences Laboratory | Box 999 Battelle Pacific Northwest National Lab | Richland, WA 99352 | e-mail:d3e102@emsl.pnl.gov | Fax: (509)-375-6631