From TAILOR@ch.pw.edu.pl Tue Aug 24 07:12:54 1993 Received: from coi.pw.edu.pl ([148.81.28.1]) for TAILOR@ch.pw.edu.pl by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA08027; Tue, 24 Aug 1993 06:36:37 -0400 Received: from mailgw.ch.pw.edu.pl ([148.81.69.10]) by coi.pw.edu.pl with SMTP (5.59/1.62, rha) id AA06445; Tue, 24 Aug 93 12:30:25 +0200 Received: From WCHEM/WORKQUEUE by mailgw.ch.pw.edu.pl via Charon 3.4 with IPX id 100.930824123721.640; 24 Aug 93 12:43:06 -0500 Message-Id: To: chemistry@ccl.net From: "ALFRED; Warsaw Univesity of Technology." Date: 24 Aug 93 12:37:17 MET-1 Subject: Need help - WEBNET X-Mailer: Pegasus Mail v2.1a. X-Pmrqc: 1 Hi I`m looking for a program "WEBNET", connecting computers throught serial ports. Please help me. Alfred =-==--==--==--==--==--< >--==--==--==--==--==-= Andrzej Tokarzewski Warsaw University of Technology Department of Chemistry p-mail:(internet) TAILOR@CH.PW.EDU.PL =-==--==--==--==--==--< >--==--==--==--==--==-= %%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%##%%## From roberto@demilo.boston.sgi.com Tue Aug 24 04:23:22 1993 Received: from sgi.sgi.com (SGI.COM) for roberto@demilo.boston.sgi.com by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA09340; Tue, 24 Aug 1993 08:23:39 -0400 Received: from relay.sgi.com by sgi.sgi.com via SMTP (920330.SGI/910110.SGI) for CHEMISTRY@ccl.net id AA27949; Tue, 24 Aug 93 05:23:36 -0700 Received: from sgibos.boston.sgi.com by relay.sgi.com via SMTP (920330.SGI/920502.SGI) for @sgi.sgi.com:CHEMISTRY@ccl.net id AA06488; Tue, 24 Aug 93 05:23:32 -0700 Received: from demilo.boston.sgi.com by sgibos.boston.sgi.com via SMTP (920330.SGI/911001.SGI) for @sgi.com:virtual@quantum.larc.nasa.gov id AA08772; Tue, 24 Aug 93 08:23:27 -0400 Received: from localhost.boston.sgi.com by demilo.boston.sgi.com via SMTP (930416.SGI/920502.SGI) for @sgibos.boston.sgi.com:CHEMISTRY@ccl.net id AA22245; Tue, 24 Aug 93 08:23:23 -0400 Message-Id: <9308241223.AA22245@demilo.boston.sgi.com> To: "Don H. Phillips" Cc: CHEMISTRY@ccl.net Subject: Re: Seeking words of wisdom <199308172144.AA08056@oscsunb.ccl.net> In-Reply-To: Your message of "Wed, 18 Aug 93 13:33:45 EDT." <9308181733.AA00465@quantum.larc.nasa.gov> Reply-To: roberto@boston.sgi.com Date: Tue, 24 Aug 93 08:23:22 EDT From: Roberto Gomperts Your message dated: Wed, 18 Aug 93 13:33:45 EDT |> In response to: |> >I am wondering what a good investment could be. That is, we have words |> >that a mini-CRAY is in the market, and that HP-750 workstations can |> >give about half the speed of a real CRAY |> |> Dr. Omar G. Stradella responds: |> >>You should also consider the new line of symmetric multiprocessors from |> >>Silicon Graphics, the Power Challenge Series. |> >> ... |> >>... (deleted specs and peak speed info for sgi and cray boxes) |> |> There are a lot of wonderful new machines out there but info on |> advertised peak speeds is nearly worthless in distinguishing between |> them. Information on sustained performance on a typical problem |> using Gaussian92, GAMESS, etc. is helpful to this audience. |> Actually, a few typical cases which differ in the vectorization |> of the code used (SCF, gradient, CI, etc) is even better since that |> shows the effects of differing cache designs. |> |> Is any info on the Power Challenge Series, Cray desk(top/side), etc |> compared to garden variety risc boxes available out there? |> |> Don Phillips |> Dear netters, Since this has been explicitly asked I will post at the end of this message a table with timings for a set of jobs ran and posted originally by Richard Walsh comparing Cay C90 execution times with PC times. My table compares the currently available SGI-Challenge systems running at 100/50 MHz and 150/75 MHz with the C90 timings from Richard Walsh. But first some clarifications/confirmations: As Michael Frisch and Mark Berger stated, Gaussian 92 has been certified for the Challenge and Onyx series running IRIX 5.0 (and higher). The first certified version was D.2. Power Challenge is a system that will be available beginning of next year, so obviously Gaussian has not been certified on this machine. However, barring any catastrophic events, Gaussian should be submitted for certification on the Power Challenge shortly after the final version of the OS and Compilers have been firmed up. Also, as Richard Walsh suggested, the Power Challenge will be binary compatible with the current Challenge, i.e. executables build for the Challenge/Onyx systems will run correctly on the Power Challenge. The drawback is that not all of the potential performance of the new machine will be utilized. Please feel free to contact Mark Mark Berger or myself if further clarification/informaton is needed. -- Roberto Roberto Gomperts roberto@sgi.com phone: (508) 562 4800 Fax: (508) 562 4755 "Gracias a la vida, que me ha dado tanto ..." Violeta Parra April/May 1993 Gaussian 92 rev D.2 Results for Challenge 100/50 MHz. ================================= Summary. ======= A number of "typical" Gaussian calculations have been performed on different number of processors and with different memory usage on the new SGI "Challenge" computer running at 100/50 MHz and 150/75MHz. When available comparisons are made with a Cray C90. Some preliminary conclusions. . Very small differences between CPU time and Wall Clock time, degradation is in general less than 3% . Minimal degradation when running throughput using all available processors (less than 3%) . Comparison with Cray C90: 1 processor runs can be as much a 25x slower than the C90, but 12 processor runs can be only 3.5x slower than the Cray . Differences between Direct and In-core calculations on the Challenge are not as big as in the Cray. Average ratio for the Cray is: 2.5x, for the Challenge (1 processor): 1.68x. This ratio becomes even smaller on the Challenge when parallelism is used: In-core speed-ups are poorer than Direct SCF speed-ups. NEW === I ran the tests on a 4 processor Challenge running at 150/75 MHz. This machine had smaller memory so instead of 16MW I used 12MW (== 96 MBytes) for the large calculations. Test 1 ====== Test 178 from Gaussian's QA suite: Single Point Direct SCF calculation: 300 basis sets, extensive Population Analysis * Increased memory to 16 MW (== 128 MBytes) Times in minutes 100/50 MHz 150/75 MHz proc CPU SpUp C90fr WClock CPU SpUp C90fr 1 103 ( 1.00) 4.1% 105 69 ( 1.00) 6.2% 2 56 ( 1.83) 7.6% 57 38 ( 1.81) 11.2% 4 31 ( 3.32) 13.8% 32 21 ( 3.28) 20.3% 8 19 ( 5.42) 22.5% 20 13 ( 5.31) 32.9% 12 15 ( 6.86) 28.5% 16 10 ( 6.90) 42.8% proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor C90fr: Percentage of a Cray C90. Base time of C90: 4.28 min CPU time WClock: Wall Clock Time. I ran 5 simultaneous copies of this job each using 4 MW (== 32 MBytes) of memory and 4 processors, on a 20 processor challenge. CPU time of slowest job: 33 min Wall Clock time for completion of all jobs: 34 min. Test2 ===== Skeleton of test178 of Gaussian's QA suite. Changed to become a Force calculation with standard Population Analysis. Command line: #P RHF/6-31G** scf=direct force * Increased memory to 16 MW (== 128 MBytes) This test gives an idea of the to be expected performance of geometry optimizations of large molecular systems. Times in minutes 100/50 MHz 150/75 MHz proc CPU SpUp WClock CPU SpUp 1 339 ( 1.00) 344 229 ( 1.00) 2 181 ( 1.87) 186 123 ( 1.86) 4 96 ( 3.53) 99 69 ( 3.31) 8 55 ( 6.16) 58 38 ( 6.02) 12 42 ( 8.07) 44 28 ( 8.17) 16 24 ( 9.54) 20 21 (10.90) Test3 ===== RHF/6-31G* Frequencies on propellane: 87 basis functions. A) Using 4 MW ( == 32 MBytes) SCF is NOT in-core Times in minutes proc CPU SpUp C90fr WClock 1 48 ( 1.00) 8.9% 49 2 25 ( 1.92) 17.1% 26 proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor C90fr: Percentage of a Cray C90. Base time of C90: 4.29 min CPU time WClock: Wall Clock Time. B) Using 16 MW ( == 128 MBytes) SCF is in-core Times in minutes 100/50 MHz 150/75 MHz proc CPU SpUp C90fr CPU SpUp C90fr 1 27.8 ( 1.00) 5.3% 19.1 ( 1.00) 7.7% 2 16.8 ( 1.65) 8.8% 11.5 ( 1.66) 12.8% 4 10.5 ( 2.64) 14.1% 7.2 ( 2.65) 20.5% 8 7.3 ( 3.81) 20.2% 12 6.3 ( 4.41) 23.4% proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor C90fr: Percentage of a Cray C90. Base time of C90: 1.48 min CPU time Test4 ===== RMP2/6-31G* Single point energy on acetone: 72 basis functions A) Using 2 MW ( == 16 MBytes) SCF is NOT in-core Times in seconds 100/50 MHz proc CPU SpUp C90fr WClock 1 339 ( 1.00) 12.8% 343 2 266 ( 1.27) 16.2% 272 4 231 ( 1.46) 18.7 243 proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor C90fr: Percentage of a Cray C90. Base time of C90: 43.27 sec CPU time B) Using 16 MW ( == 128 MBytes) SCF is in-core Times in seconds 100/50 MHz 150/75 MHz proc CPU SpUp C90fr CPU SpUp C90fr 1 157 ( 1.00) 9.7% 103 ( 1.00) 14.7% 2 122 ( 1.28) 12.4% 83 ( 1.24) 18.3% 4 108 ( 1.45) 14.1% 72 ( 1.43) 21.1% proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor C90fr: Percentage of a Cray C90. Base time of C90: 15.23 sec CPU time Test5 ===== RMP4/6-31G* Single point energy on acetaldehyde: 53 Using 4 MW ( == 32 MBytes) SCF is in-core Times in seconds 100/50 MHz 150/75 MHz proc CPU SpUp C90fr WClock 1 690 ( 1.00) 9.2% 706 2 452 ( 1.52) 13.9% 465 4 324 ( 2.13) 19.4% 344 8 266 ( 2.59) 23.6% 299 12 242 ( 2.85) 26.0% 283 proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor C90fr: Percentage of a Cray C90. Base time of C90: 62.94 sec CPU time Test6 ===== RHF/6-31+G CI-Singles excited states on benzene: 90 basis functions A) Using 2 MW ( == 16 MBytes) SCF is NOT in-core Times in seconds 100/50 MHz 150/75 MHz proc CPU SpUp C90fr WClock CPU SpUp C90fr 1 908 ( 1.00) 9.7% 924 600 ( 1.00) 14.7% 2 753 ( 1.20) 11.7% 783 480 ( 1.25) 18.3% 4 652 ( 1.39) 13.5% 685 420 ( 1.42) 21.0% 8 587 ( 1.54) 15.0% 619 12 587 ( 1.54) 15.0% 647 proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor C90fr: Percentage of a Cray C90. Base time of C90: 88.23 sec CPU time B) Using 16 MW ( == 128 MBytes) SCF is in-core Times in seconds 100/50 MHz 150/75 MHz proc CPU SpUp CPU SpUp 1 823 ( 1.00) 540 ( 1.00) 2 713 ( 1.15) 480 ( 1.12) 4 652 ( 1.26) 420 ( 1.28) proc: Number of processors CPU: user+system cpu time SpUp: Speed Up with respect to 1 processor (*) This test case gives different number of SCF cycles for varying number of processors and memory sizes (between 13 and 19 cycles). All results are consistent however at all levels: RHF, MP2 and CIS. Roberto Gomperts From WILLIAMS%XRAY2@ULKYVX.LOUISVILLE.EDU Tue Aug 24 08:40:17 1993 Received: from ulkyvx.louisville.edu for WILLIAMS%XRAY2@ULKYVX.LOUISVILLE.EDU by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA14035; Tue, 24 Aug 1993 13:41:08 -0400 Received: from DECNET-MAIL (WILLIAMS@XRAY2) by ULKYVX.LOUISVILLE.EDU (PMDF V4.2-11 #3949) id <01H24J57L81S95N7M3@ULKYVX.LOUISVILLE.EDU>; Tue, 24 Aug 1993 13:40:17 EST Date: Tue, 24 Aug 1993 13:40:17 -0500 (EST) From: WILLIAMS%XRAY2@ULKYVX.LOUISVILLE.EDU Subject: SGI Z-buffer To: chemistry@ccl.net Message-Id: <01H24J57M0ZM95N7M3@ULKYVX.LOUISVILLE.EDU> X-Vms-To: ULKYVX::IN%"chemistry@ccl.net" X-Vms-Cc: WILLIAMS Mime-Version: 1.0 Content-Type: TEXT/PLAIN; CHARSET=US-ASCII Content-Transfer-Encoding: 7BIT I am planning to purchase an entry level Silicon Graphics Indigo. I have no illusions about memory requirements and plan to specify 32MB RAM and at least a 1GB disk, so I can run G92 efficiently. However, I am uncertain about graphics program requirements for chemistry software. Specifically, can I run commonly available graphics programs on the entry level machine, or is a Z-buffer necessary? I just want the graphics program to run, not necessarily at top speed. Graphics upgrades on the Indigo seem quite expensive. Thank you for your help. Don Williams Chemistry Dept. University of Louisville Louisville, KY 40292 dewill01@ulkyvx.louisville.edu From shaoweng@helix.nih.gov Tue Aug 24 11:48:14 1993 Received: from helix.nih.gov for shaoweng@helix.nih.gov by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA15877; Tue, 24 Aug 1993 15:48:20 -0400 Received: by helix.nih.gov (5.64/1.35(helix-1.0)) id AA23936; Tue, 24 Aug 93 15:48:14 -0400 Date: Tue, 24 Aug 93 15:48:14 -0400 From: shaoweng@helix.nih.gov (Shaomeng Wang) Message-Id: <9308241948.AA23936@helix.nih.gov> To: CHEMISTRY@ccl.net Subject: AM1-SM1 program Cc: shaoweng@helix.nih.gov Dear netters: Does anyone have information on how to get AM1-SM1 program running on SGI machine and how much it will cost? Thank you in advance! Shaomeng Wang LMC,NCI,NIH Bethesda, MD20892 From eberlin@iqm.unicamp.br Tue Aug 24 12:32:09 1993 Received: from styx.iqm.unicamp.br (iqm.unicamp.br) for eberlin@iqm.unicamp.br by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA16662; Tue, 24 Aug 1993 16:39:08 -0400 Return-Path: Received: by styx.iqm.unicamp.br (Sendmail 4.1/1.07) id AA06571; Tue, 24 Aug 93 17:32:09 EST Date: Tue, 24 Aug 93 17:32:09 EST From: Marcos Eberlin Message-Id: <9308242032.AA06571@styx.iqm.unicamp.br> To: chemistry@ccl.net Subject: Heat of Formatiom from Gaussian92 Dear Netters, I have performed a geometry/energy calculation at the MP2/6-31G(d,p) level and a frequency calculation at 298K for the CH2=S=OH+ ion by using Gassian92. My question is: How can I calculate the heat of formation of this ion from the results? I would like to compare it with the experimental value. Thanks for helping. Prof. Marcos N. Eberlin State University of Campinas Campinas, SP Brazil EBERLIN@IQM.UNICAMP.BR From MA@ASUCHM.LA.ASU.EDU Tue Aug 24 07:05:19 1993 Received: from asuchm.la.asu.edu for MA@ASUCHM.LA.ASU.EDU by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA17040; Tue, 24 Aug 1993 17:05:21 -0400 Date: Tue, 24 Aug 1993 14:05:19 -0700 (MST) From: MA@ASUCHM.LA.ASU.EDU To: chemistry@ccl.net Message-Id: <930824140519.29499@ASUCHM.LA.ASU.EDU> Subject: change of e-mail address please change my e-mail address from ma@asuchm.la.asu.edu to hongma@ncsc.org Thanks hong ma From keogh@chipmunk.cita.utoronto.ca Tue Aug 24 13:54:50 1993 Received: from chipmunk.cita.utoronto.ca for keogh@chipmunk.cita.utoronto.ca by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA17841; Tue, 24 Aug 1993 17:53:46 -0400 Received: by chipmunk.cita.utoronto.ca id AA08978; Tue, 24 Aug 93 17:54:50 EDT Date: Tue, 24 Aug 93 17:54:50 EDT From: Bill Keogh Message-Id: <9308242154.AA08978@chipmunk.cita.utoronto.ca> To: chemistry@ccl.net Subject: He-H2 surface; He basis functions Dear fellow computational chemists: We are interested in studying the molecular dynamics of the He-H2 system for comparison with the H-H2 system. We would like to have the most accurate, up-to-date analytical He-H2 potential energy surface available, and we would like to compute some ab initio points to test the accuracy of He-H2 surface. Does anyone know of a more recent or more accurate surface than the one described by Wilson, Kapral and Burns in 1974? (Potential energy surface for the Hydrogen Molecule-Helium system; C.W.Wilson Jr., R.Kapral, G.Burns; Chem.Phys.Lett. 24 488 1974) We are also looking for a bigger (and hopefully better) set of Gaussian basis functions for the He atom. (We already have a good set of basis functions for the H2 molecule.) We are hoping for a set containing roughly 20 to 40 elementary basis functions (e.g. 5s3p1d to 10s6p2d); a larger basis set would be even better, provided it contains contraction coefficients, such that it can be contracted to the desired size. Can anyone point us to a reference which contains such a set of basis functions? Thanks in advance for any and all help. ... Bill Dr. William J. Keogh Department of Chemistry and Canadian Institute for Theoretical Astrophysics University of Toronto, Canada Email address: keogh@cita.utoronto.ca From shenkin@still3.chem.columbia.edu Tue Aug 24 14:45:29 1993 Received: from mailhub.cc.columbia.edu for shenkin@still3.chem.columbia.edu by oscsunb.ccl.net (5.65c+KVa/930601.1506) id AA18440; Tue, 24 Aug 1993 18:45:44 -0400 Received: from merhaba.cc.columbia.edu by mailhub.cc.columbia.edu with SMTP id AA21056 (5.65c+CU/IDA-1.4.4/HLK for chemistry@ccl.net); Tue, 24 Aug 1993 18:45:37 -0400 Received: from still3.chem.columbia.edu by merhaba.cc.columbia.edu with SMTP id AA08334 (5.65c+CU/IDA-1.4.4/HLK for WILLIAMS%XRAY2%ULKYVX.LOUISVILLE.EDU@cunixf.cc.columbia.edu); Tue, 24 Aug 1993 18:44:10 -0400 Received: by still3.chem.columbia.edu (920330.SGI/920502.SGI.AUTO) for @cunixf.cc.columbia.edu:chemistry@ccl.net id AA03944; Tue, 24 Aug 93 18:45:29 -0400 Date: Tue, 24 Aug 93 18:45:29 -0400 From: shenkin@still3.chem.columbia.edu (Peter Shenkin) Message-Id: <9308242245.AA03944@still3.chem.columbia.edu> To: WILLIAMS%XRAY2@ULKYVX.LOUISVILLE.EDU, chemistry@ccl.net Subject: Re: SGI Z-buffer Entry-level graphics is fast enough to rotate small molecules interactively, even in a CPK representation. Trouble is, it looks ugly, due to low-resolution graphics, few bit planes and dithering. Entry-level graphics does Z buffering in software. The next step up is XS graphics. This has no Z buffer, even in software. In answer to your question: Yes, this makes it a poor choice for an application that tries to rotate molecules and give a "nice" 3D representation. The next step up beyond this is XS-24. This also has no Z buffer, even in software, and is therefore also a poor choice. The first step beyond entry-level that gives you a Z buffer at all is XS-24Z. This provides a hardware Z buffer and very nice graphics, indeed. More expensive models than this offer greater speed, and in some cases bells and whistles, such as "fog", "texture mapping", and other things which molecular visualization programs could, in principle, use to good effect, but, for the most part, do not bother to implement; for one thing, they wouldn't work on the lower models that lots of people have. This is YATTMCWTAOOGL (Yet Another Thing That May Change With The Advent Of OpenGL). Bottom line: get at least XS-24Z if you can. If you can't, get entry-level graphics. If you're going to want to do a lot of visualization of large molecules, such as proteins, try to get even a faster graphics engine, such as XZ or Elan. SGI sells a special package that includes an Indigo R4000 XZ with 32 Mb RAM and 1.2 Gb disk. It's imaginable that the special pricing of the package might make this not much more expensive than the same machine with XS-24Z, but I don't really know. By the way, your choice of 32 Mb and 1.2 Gb is a reasonable minimum configuration for molecular modeling. For ab-initio quantum calculations you'd probably want to increase main memory considerably. For molecular modeling of large molecules, e.g. proteins, you'd probably want to increase swap space to 100 Mb from the default 40 Mb once you receive the machine. Hope this helps, -P. ************************f*u*cn*rd*ths*u*cn*gt*a*gd*jb************************ Peter S. Shenkin, Box 768 Havemeyer Hall, Dept. of Chemistry, Columbia Univ., New York, NY 10027; shenkin@still3.chem.columbia.edu; (212) 854-5143 ********************** Wagner, Beame, Screvane in '93! **********************