From chemistry-request@ccl.net Wed Mar 4 08:49:54 1992 Date: Wed, 4 Mar 1992 05:55 CST From: Andy Holder Subject: A gadfly forever..... To: CHEMISTRY@ccl.net Status: RO I haven't run this out yet, but I'm sure that AMPAC using the highly successful AM1 method could perform this calculation in a much shorter time than an ab initio calculation with reasonable results. I think the discussion should refocus on "What can be done now that couldn't be done before?" Semi- empirical methods can now do calculations on reasonably sized polymer chains or biomolecules efficiently. Oh, well. Andy Holder =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= DR. ANDREW HOLDER Assistant Professor of Computational/Organic Chemistry Department of Chemistry || BITNET Addr: AHOLDER@UMKCVAX1 University of Missouri - Kansas City || Internet Addr: aholder@vax1.umkc.edu Spencer Chemistry, Room 502 || Phone Number: (816) 235-2293 Kansas City, Missouri 64110 || FAX Number: (816) 235-1717 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= From chemistry-request@ccl.net Wed Mar 4 10:36:20 1992 Date: Wed, 4 Mar 92 9:09:04 EST From: rs0thp@rohmhaas.COM (Dr. Tom Pierce) Subject: Re: Performance comparisons question To: chemistry@ccl.net Status: RO > >Date: Wed, 26 Feb 92 08:30:08 -0800 > >From: jev@chem.ucla.edu (Jayasankar Variyar) > Dear Netters, > I have access to a Cyber-840(?), an Apollo 9000, and and 486/33(running Dos/Windows3.0). I would like to get the performance ratings for these machines realtive to a VAX-780. I will be running MD simulations (simple pair potentials) on these machines. > Also what would be the best Fortran compiler for the PC? I can get Microsoft Fortran 5.1, but how does it compare to Lahey/ Watfor compilers? > Jayasankar E. Variyar > ----------- End Forwarded Message ----------- I am not familar with a Cyber-840, the Apollo 9000 is an Alliant FX-8 I think, and I just benched Lahey and Watfor on 486 and 386 machines. Ballpark estimates of relative speeds depend almost completely on the code that is used. Caveat Emptor. For my small dynamics code (particle simulation and growth for a polymer reactor, I obtained the following comparisons. Alliant FX-80 10 minutes 386/33MHz/387 20 Minutes 486/33Mhz 10 Minutes est. MicroVax time 50 minutes I tested Lahey's compilers (The Big one F77EM32 and the 640K version ) and the Watcom compiler. Basically the Lahey worked easier (BOth versions) I was unable to get the Watcom to work (I am still trying tho). My final comment is that with the Lahey EM32 compiler with virtual memory, and the speed of a 386/33MHz/387 cpu, PCs are useful for large calculations. -- -------------------------------------------------------------------- Sincerely, Thomas Pierce Computer Applications Research | rs0thp@rohmhaas.com Internet Bldg 64C, Rohm and Haas Co. | rs0thp@rohvm1 Bitnet P.O. Box 219 | (215)-785-8989 Voice Bristol, PA 19007 | (215)-781-4204 Fax Official Disclaimer: "The opinions expressed are those of the writer and not the Rohm and Haas Company." From chemistry-request@ccl.net Wed Mar 4 10:36:29 1992 Date: Wed, 4 Mar 92 09:40:36 EST From: states@ncbi.nlm.nih.GOV (David States) Subject: Re: A gadfly forever..... (How far can you push semi-empirical To: CHEMISTRY@ccl.net, AHOLDER@VAX1.UMKC.EDU Status: RO |> Andy Holder writes: |> |> ... I think the discussion should refocus |> on "What can be done now that couldn't be done before?" Semi- |> empirical methods can now do calculations on reasonably sized |> polymer chains or biomolecules efficiently. To me this is a real opportunity. Empirical energy simulations of biological macromolecules have improved incrementally over the past decade, but many significant terms in the potential energy function are ignored, particularly polarization effects (charge induced dipole terms, cooperativity in hydrogen bonding etc.). These terms are significant compared to kT. It does not make alot of sense to me to run longer and longer simulations when you know the underlying potential function is flawed. Semi-empirical methods offer a real hope of addressing some of these issues. So what are the largest peptide systems that have been evaluated with semi-empirical methods? Have any been done with Monte Carlo averaging over nuclear conformations? Has anyone been bold enough to attempt dynamical simulations based on semi-empirical PE functions and derivatives? David States National Center for Biotechnology Information / National Library of Medicine From chemistry-request@ccl.net Wed Mar 4 10:49:23 1992 Date: Tue, 3 Mar 92 23:17:52 -0500 From: fredvc@esvax.dnet.dupont.com Subject: RE: the future of ab initio To: %chem@ccl.net Status: RO From: ESDS01::"m10!frisch@uunet.UU.NET" "Michael Frisch" 3-MAR-1992 19:06:21.41 To: chemistry@ccl.net CC: Subj: Re: The future of ab initio > What would you do if you could ROUTINELY perform ab initio > calculations (let say at the level of HF or HF/MP2) with 500 > contracted basis functions? What kind of projects and > applications would you run? > > Jan Labanowski > Ohio Supercomputer Center ------------------------------------------------------------------------------- >>HF/6-31G* on C30H62 with no symmetry (574 basis functions) takes well under >>a day in Gaussian 90 on an inexpensive workstation (RS/6000 model 530). >>Shouldn't that question be "What applications do you do now that such >>calculations have been possible for a couple of years?" >> >>Michael Frisch >>Gaussian, Inc. ------------------------------------------------------------------------------- We have certainly come a long way from the time when calculation of the rotational barrier in ethane was worthy of publication as a JACS Communication. Dr. Frisch certainly provides a good benchmark for where we are. However, one would really like to complete such a calculation in 30 minutes or less!! Such a time scale, within a factor of 2-3, is what I believe to be the intended frame of reference for the question. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ FREDERIC A. VAN-CATLEDGE Scientific Computing Division || Office: (302) 695-1187 Central Research & Development Dept. || FAX: (302) 695-9658 The Du Pont company || P. O. Box 80320 || Internet: fredvc@esvax.dnet.dupont.com Wiilmington DE 19880-0320 || ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ From chemistry-request@ccl.net Wed Mar 4 12:47:57 1992 Date: Wed, 4 Mar 92 08:35:06 PST From: case@scripps.EDU (David Case) Subject: Re: A gadfly forever..... (How far can you push semi-empirical To: chemistry@ccl.net Status: RO David States wrote: > To me this is a real opportunity. Empirical energy simulations of > biological macromolecules have improved incrementally over the past > decade, but many significant terms in the potential energy function are > ignored, particularly polarization effects (charge induced dipole > terms, cooperativity in hydrogen bonding etc.). These terms are > significant compared to kT. It does not make alot of sense to me to > run longer and longer simulations when you know the underlying > potential function is flawed. Semi-empirical methods offer a real > hope of addressing some of these issues. > I too would enjoy hearing from people that know about this. When I wrote a review a few year ago [in "Conformational Analysis of Medium-Sized Heterocycles", edited by Richard Glass] semiempirical methods seemed to have very poor performance for non-bonded interactions and hydrogen bonds -- things that are crucial to most peptide and protein simulations. Errors in barriers to rotation about bonds, or in ring conformations (e.g. planar cylcopentane) were also common. But at that time, not much had been published with AM1, although that appeared to give improved results in these areas. Are semi-empirical methods now meeting Dave States' desires? ===================================================================== David A. Case | internet: case@scripps.edu Dept. of Molecular Biology | bitnet: case%scripps.edu@sdsc The Scripps Research Institute | fax: 619-554-6717 10666 N. Torrey Pines Rd. | phone: 619-554-9768 La Jolla CA 92037 USA | ===================================================================== From chemistry-request@ccl.net Wed Mar 4 13:45:27 1992 Date: Wed, 4 Mar 92 08:35:38 PST From: d3f012@gator.pnl.GOV Subject: large scale semi-empirical calculations To: chemistry@ccl.net Status: RO > |> Andy Holder writes: > |> > |> ... I think the discussion should refocus > |> on "What can be done now that couldn't be done before?" Semi- > |> empirical methods can now do calculations on reasonably sized > |> polymer chains or biomolecules efficiently. > > David States writes: > > To me this is a real opportunity. Empirical energy simulations of > biological macromolecules have improved incrementally over the past > decade, but many significant terms in the potential energy function are > ignored, particularly polarization effects (charge induced dipole > terms, cooperativity in hydrogen bonding etc.). These terms are > significant compared to kT. It does not make alot of sense to me to > run longer and longer simulations when you know the underlying > potential function is flawed. Semi-empirical methods offer a real > hope of addressing some of these issues. > > So what are the largest peptide systems that have been evaluated > with semi-empirical methods? > > Have any been done with Monte Carlo averaging over nuclear conformations? > > Has anyone been bold enough to attempt dynamical simulations based > on semi-empirical PE functions and derivatives? > > David States > National Center for Biotechnology Information / National Library of Medicine > We routinely do INDO/s SCF/CI calculations on the chromophores of bacterial photosynthetic reaction centers that involve ~600+ atoms, 1500+ electrons, and a few thousand configurations (single-excited) in the CI. These calculations also have included portions of the surrounding protein. We have been doing Molecular Dynamics with forces from a semi-empirical Hartree Fock solution. Mark A. Thompson Molecular Science Research Center Pacific Northwest Laboratory Richland, WA. d3f012@pnlg.pnl.gov From chemistry-request@ccl.net Wed Mar 4 15:34:59 1992 Date: Wed, 4 Mar 1992 13:25:11 -0500 From: zheng@retina.chem.psu.EDU (Ya-Jun Zheng) To: JKL@MPS.OHIO-STATE.EDU Status: RO To those who are interested in the performance of molecular mechanics on hydrogen bonding interactions. The following reference may be very interesting. J. Mol. Struct. 1992, 265, 179. For system like CH3COO-...3H2O, the molecular mechanics results can be 10-14 kcal/mol off the experimental value. Yajun Zheng From chemistry-request@ccl.net Wed Mar 4 16:15:53 1992 Date: Wed, 4 Mar 1992 15:26 EST From: "DOUGLAS A. SMITH" Subject: Midwest Theo. Chem. Conference To: chemistry@ccl.net Status: RO Can anyone tell me when and where the next Midwest Theoretical Chemistry Conference will be held? Who should I contact for more details? Doug Smith Assistant Professor of Chemistry The University of Toledo Toledo, OH 43606-3390 voice 419-537-2116 fax 419-537-4033 email fax0236@uoft02.utoledo.edu From chemistry-request@ccl.net Wed Mar 4 17:34:08 1992 Date: Wed, 4 Mar 92 13:19:59 -0600 From: berkley@wubs.wustl.EDU (Berkley Shands) Subject: Recent Benchmarks To: chemistry@ccl.net Status: RO I've seen some number flying around for different machines. Here are some for an "Active Analog" model of the ACE inhibitors. Runtime summary of Constrained Search (CSR) V4.5F from The Center For Molecular Design, Washington University - St. Louis, Mo. Optimization level as indicated, single precision floating point when available. VAXen run VMS 5.5 with VIP, all others are running some flavor of BSD or SYSV Unix. Single stream execution, with Autosegmentation, Sweeping and log files enabled. Angle step order is lowest to highest (0:359). Fast ordering of selected molecules. Benchmarks of the complete ACE series of 71 molecules Resolutions: 0.25 0.25 0.20 0.20 0.20 "csr ace.run -u4 -m1 -M0 -Tabsolute -Oforward -lace.log -G5000" Smallest molecule has three rotatable bonds, largest has nine. Uniform four degree scan factor at all rotations. The SPARC machines used an aproximate square-root. All others used the standard system call. This is equivalent to a 20% speedup on SPARC machines. Unbundled C compiler using hardware squareroot ran 31% SLOWER. ******************************************************************************* Top CPU Top Elps Top Sys Chg CPU Chg Elps XuVax2 CPU Type and notes ======= ======== ======= ======= ======== ====== ================== 7008 7647 318 6314 6597 1.00 uVAX2 16mb VMS 5.5+VIP 7183 7583 0 6347 6509 1.00 uVAX2 16mb VMS 5.5+RMS 1678 2083 186 1474 1609 4.30 VS3520 32mb VMS 5.5+VIP 1779 2071 0 1547 1623 4.10 VS3520 32mb VMS 5.5+RMS 630 852 33 586 651 10.83 IRIS 4d/20gt (IP6) 16mb R2000@12Mhz 386 479 3 338 347 18.77 IBM RS-6000-320 24mb (AIX 3.1) 337 420 7 286 292 22.19 SUN 4mp Sparc 690 64Mb 361 441 8 280 331 22.55 SUN 4/c SPARCstation 2 16mb 272 340 10 247 278 25.56 ESV M120 R3000A@25Mhz 289 400 5 239 294 26.41 IBM RS-6000-320H 32mb (AIX 3.2) 205 283 6 189 240 33.40 IRIS Indigo (IP12) 48mb R3000A@35Mhz 198 249 36 177 212 35.67 IRIS 4d/380 (IP7) 128mb R3000A@35Mhz 172 192 1 140 154 45.10 IBM RS-6000-550 64mb (AIX 3.2) ******************************************************************************* Notes: "Top CPU" is the recorded total CPU for all forked processes, including overhead and I/O. "Top Elps" is the recorded elapsed time from command line parsing to the exit of the last child processes. "Top Sys" is the recorded UNIX system overhead from the times() function (VMS has no equivalent) "Chg CPU" is the algorithmic chargable time from initial rotations to termination (child process CPU) "Chg Elps" is the elapsed time spent rotating and computing the algorithm. "XuVax2" is the scale factor to a Micro-VAXII measured in chargable CPU time. Berkley