From chemistry-request@server.ccl.net Mon Mar 26 07:42:17 2001 Received: from musset.lpct.u-bordeaux.fr ([147.210.56.51]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QCgHD14730 for ; Mon, 26 Mar 2001 07:42:17 -0500 Received: from lpct.u-bordeaux.fr (localhost [127.0.0.1]) by musset.lpct.u-bordeaux.fr (8.9.3/jtpda-5.3.2) with ESMTP id OAA00537 for ; Mon, 26 Mar 2001 14:48:08 +0200 (METDST) Sender: fleurat@lpct.u-bordeaux.fr Message-ID: <3ABF3A87.79F24B2D@lpct.u-bordeaux.fr> Date: Mon, 26 Mar 2001 14:48:07 +0200 From: Paul Fleurat-Lessard Reply-To: fleurat@lpct.u-bordeaux.fr Organization: LPCM X-Mailer: Mozilla 4.61 [en] (X11; I; HP-UX B.10.01 9000/735) X-Accept-Language: fr-FR, en MIME-Version: 1.0 To: CCL Subject: Queuing systems Content-Type: text/plain; charset=iso-8859-2 Content-Transfer-Encoding: 7bit Hi, A few weeks ago, I asked about the best queuing system available for a cluster of PC. This topic has already been discussed, and PBS (www.openpbs.com) seems to be a good one, especially when using patches to improve it. (see archives for PBS to have some examples.) Another system is Codine 5.2.2 which was released for FREE (http://www.sun.com/gridware/). We have not yet tested these systems so I cannot say more by now. Paul. -- ------------------------------------------------------------------------- Fleurat-Lessard Paul | fleurat@lpct.u-bordeaux.fr Laboratoire de Physico Chimie Moleculaire, | Phone: (33)(0)5 56 84 63 09 Universite Bordeaux I, 33400 Talence, FRANCE| Fax : (33)(0) 5 56 84 66 45 _________________________________________________________________________ From chemistry-request@server.ccl.net Mon Mar 26 08:08:50 2001 Received: from schrodinger.com (root@[192.156.98.99]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QD8nD14950 for ; Mon, 26 Mar 2001 08:08:49 -0500 Received: from tandoori.schrodinger.com (root@tandoori.schrodinger.com [216.216.237.99]) by schrodinger.com (8.8.5/8.8.5) with ESMTP id FAA11250; Mon, 26 Mar 2001 05:08:30 -0800 Received: from sally.schrodinger.com (shenkin@sally.schrodinger.com [216.216.237.92]) by tandoori.schrodinger.com (8.8.5/8.8.5) with ESMTP id IAA13253; Mon, 26 Mar 2001 08:08:27 -0500 Received: from localhost (shenkin@localhost) by sally.schrodinger.com (8.8.8/8.8.5) with ESMTP id IAA09453; Mon, 26 Mar 2001 08:08:27 -0500 X-Authentication-Warning: sally.schrodinger.com: shenkin owned process doing -bs Date: Mon, 26 Mar 2001 08:08:27 -0500 (EST) From: Peter Shenkin To: Paul Fleurat-Lessard cc: CCL Subject: Re: CCL:Queuing systems In-Reply-To: <3ABF3A87.79F24B2D@lpct.u-bordeaux.fr> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hi, We have been asking around ourselves, and find that in the commercial world, the most popular is LSF, by platform computing. Mucho dinero, however. -P. On Mon, 26 Mar 2001, Paul Fleurat-Lessard wrote: > Date: Mon, 26 Mar 2001 14:48:07 +0200 > From: Paul Fleurat-Lessard > To: CCL > Subject: CCL:Queuing systems > > Hi, > > A few weeks ago, I asked about the best queuing system available for a > cluster of PC. This topic has already been discussed, and PBS > (www.openpbs.com) seems to be a good one, especially when using patches to > improve it. (see archives for PBS to have some examples.) > Another system is Codine 5.2.2 which was released for FREE > (http://www.sun.com/gridware/). > > We have not yet tested these systems so I cannot say more by now. > > Paul. > -- > ------------------------------------------------------------------------- > Fleurat-Lessard Paul | fleurat@lpct.u-bordeaux.fr > Laboratoire de Physico Chimie Moleculaire, | Phone: (33)(0)5 56 84 63 09 > Universite Bordeaux I, 33400 Talence, FRANCE| Fax : (33)(0) 5 56 84 66 45 > _________________________________________________________________________ > > -= This is automatically added to each message by mailing script =- > CHEMISTRY@ccl.net -- To Everybody | CHEMISTRY-REQUEST@ccl.net -- To Admins > MAILSERV@ccl.net -- HELP CHEMISTRY or HELP SEARCH > CHEMISTRY-SEARCH@ccl.net -- archive search | Gopher: gopher.ccl.net 70 > Ftp: ftp.ccl.net | WWW: http://www.ccl.net/chemistry/ | Jan: jkl@ccl.net > > > > > From chemistry-request@server.ccl.net Mon Mar 26 08:55:23 2001 Received: from vytautas.vdu.lt (mail@[193.219.38.33]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QDt1D15190 for ; Mon, 26 Mar 2001 08:55:04 -0500 Received: from vaidila.vdu.lt ([193.219.38.52] ident=root) by vytautas.vdu.lt with esmtp for chemistry@ccl.net id 14hXRl-0004Vw-00; Mon, 26 Mar 2001 15:54:13 +0200 Received: from vaidila.vdu.lt (vejas.vdu.lt [193.219.38.82]) by vaidila.vdu.lt (8.9.3/8.9.3) with ESMTP id PAA22666 for ; Mon, 26 Mar 2001 15:54:00 +0200 (GMT) Message-ID: <3ABEA155.CC363B94@vaidila.vdu.lt> Date: Mon, 26 Mar 2001 03:54:29 +0200 From: "art'" X-Mailer: Mozilla 4.7 [en] (Win98; I) X-Accept-Language: en MIME-Version: 1.0 CC: chemistry@ccl.net Subject: CCL: flexibaale docking error in AD3.0 References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hi, I run flexible docking with AutoDock 3.0 under Suse 6.1 Linux with 3 branches. During run I see in console: "WARNING! The population appears to have converged, so this run will shortly terminate." In my *.dlg (result) file after assigning translations and torsions I see strange strings: "eval.cc: ERROR! energy is not a number!" Finally, at the end of *.dlg file I see "+NaN kcal/mol" for final docking energies. The vdW columns of results in PDBQ format at the end of *.dlg file have enormous real numbers, as 3225.98 or 8761.24 .... She same structure without torsions in rigid docking gave "normal" energies and no problems. What is wrong ? A.Ziemys From chemistry-request@server.ccl.net Mon Mar 26 12:27:18 2001 Received: from mail.chem.tamu.edu ([165.91.176.8]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QHRID16858 for ; Mon, 26 Mar 2001 12:27:18 -0500 Received: from mail.chem.tamu.edu ([165.91.177.32]) by mail.chem.tamu.edu (Post.Office MTA v3.5.3 release 223 ID# 0-64333U1000L100S0V35) with ESMTP id edu for ; Mon, 26 Mar 2001 11:27:15 -0600 Sender: yubo@ccl.net Message-ID: <3ABF7C75.B02B387B@mail.chem.tamu.edu> Date: Mon, 26 Mar 2001 17:29:25 +0000 From: Yubo Fan Organization: Chemistry Department, Texas A&M University X-Mailer: Mozilla 4.7C-SGI [en] (X11; I; IRIX64 6.5 IP28) X-Accept-Language: zh-CN, en MIME-Version: 1.0 To: CHEMISTRY@ccl.net Subject: Step Size Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hi, everyone, I want to try some geometry optimizations at the smaller level of step size by G98. But the options, opt=(stepsize=N) or IOp=(39/N), cannot work properly. Any advice? Thanks in advance. I will summerize. Yubo -- ============================================================ Yubo Fan Email: yubofan@mail.chem.tamu.edu Department of Chemistry Tel: 1-979-845-7222 Texas A&M University College Station, TX 77843 ============================================================ From chemistry-request@server.ccl.net Mon Mar 26 04:52:41 2001 Received: from linus.imcm.cas.cz ([147.231.77.250]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2Q9qeD12945 for ; Mon, 26 Mar 2001 04:52:41 -0500 Received: from thales (thales.imcm.cas.cz [147.231.77.130]) by linus.imcm.cas.cz (8.9.3/8.9.3) with ESMTP id LAA05924 for ; Mon, 26 Mar 2001 11:52:39 +0200 Message-Id: <200103260952.LAA05924@linus.imcm.cas.cz> From: "Petr Toman" To: chemistry@ccl.net Date: Mon, 26 Mar 2001 11:53:04 +0200 MIME-Version: 1.0 Content-type: text/plain; charset=ISO-8859-2 Subject: CCL:ZINDO in G98 & Molecular Orbitals -- SUMMARY Reply-to: toman@imc.cas.cz Priority: normal X-mailer: Pegasus Mail for Win32 (v3.12a) Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from Quoted-printable to 8bit by server.ccl.net id f2Q9qfD12946 Dear CCLers, I received several answers to my question, but only the suggestion from Jens Antony solved the problem. The Pop=Full or Pop=Regular options do not work with ZINDO. Yours, Petr Toman ------- Forwarded message follows ------- Datum odeslání: Fri, 23 Mar 2001 17:53:03 +0100 (MET) Od: Jens Antony Komu: Petr Toman Věc: Re: CCL:ZINDO in G98 & Molecular Orbitals Dear Petr, try IOp(5/33=2). Best regards, Jens ------- End of forwarded message ------- From chemistry-request@server.ccl.net Mon Mar 26 07:37:08 2001 Received: from sina.com ([202.106.187.170]) by server.ccl.net (8.11.0/8.11.0) with SMTP id f2QCb6D14678 for ; Mon, 26 Mar 2001 07:37:07 -0500 Received: (qmail 6308 invoked by uid 99); 26 Mar 2001 12:36:25 -0000 Message-ID: <20010326123625.6305.qmail@sina.com> From: cgliu To: chemistry@ccl.net Cc: yihan@bastille.cchem.berkeley.edu, yangmh@cz3.nus.edu.sg Subject: Again about CASSCF geometry optimization Date: Mon, 26 Mar 2001 20:36:25 +0800 X-Mailer: SinaMail 3.0Beta (FireToad) X-Priority: 3 Dear cclers: I got some suggestions for my last question on this topic. I am sorry I had not stated it clearly. I am doing a CASSCF optimization of Decapentaene( C10H12), to find the gometry of the ground state and the first excited state structure. The program I used is Gaussian 98 For the first excited state structure. I took the following steps: 1) UHF optimization to find a rough geomtry which will be used as a starting geomtry in the next step 2) CASSCF(10,10)/3-21g single point calculation to find if it is necessary to alter the sequence of the orbitals. 3) The active space is compose of the ten pi and pi* molecular orbitals. 3) CASSCF(10,10)/3-21g guess=(alter,read) geom=check opt iop(1/8=3) to do geometry optimization with very small step size. Unfortunately, after 2 steps of optimization. The abnormal result of MC-SCF result was reached, which looks like: Step scaled by 0.1350669311289174 ITN= 1 MaxIt= 64 E= -364.8936812240 DE=-3.65D+02 Acc= 1.00D-08 Lan= 1 ITN= 2 MaxIt= 64 E= -366.1983086710 DE=-1.30D+00 Acc= 1.00D-08 Lan= 1 ITN= 3 MaxIt= 64 E= -367.3423713309 DE=-1.14D+00 Acc= 1.00D-08 Lan= 1 ITN= 4 MaxIt= 64 E= -368.3968198326 DE=-1.05D+00 Acc= 1.00D-08 Lan= 1 ITN= 5 MaxIt= 64 E= -369.5356108067 DE=-1.14D+00 Acc= 1.00D-08 Lan= 1 ITN= 6 MaxIt= 64 E= -370.4283645295 DE=-8.93D-01 Acc= 1.00D-08 Lan= 1 ITN= 7 MaxIt= 64 E= -371.0982314618 DE=-6.70D-01 Acc= 1.00D-08 Lan= 1 ITN= 8 MaxIt= 64 E= -371.5249216994 DE=-4.27D-01 Acc= 1.00D-08 Lan= 1 ITN= 9 MaxIt= 64 E= -372.1397979687 DE=-6.15D-01 Acc= 1.00D-08 Lan= 1 ITN= 10 MaxIt= 64 E= -372.4663398452 DE=-3.27D-01 Acc= 1.00D-08 Lan= 1 ITN= 11 MaxIt= 64 E= -372.6890832012 DE=-2.23D-01 Acc= 1.00D-08 Lan= 1 TTN= 12 MaxIt= 64 E= -372.7988003755 DE=-1.10D-01 Acc= 1.00D-08 Lan= 1 ITN= 13 MaxIt= 64 E= -372.8776810015 DE=-7.89D-02 Acc= 1.00D-08 Lan= 1 ITN= 14 MaxIt= 64 E= -372.9287242959 DE=-5.10D-02 Acc= 1.00D-08 Lan= 1 ITN= 15 MaxIt= 64 E= -372.9442106552 DE=-1.55D-02 Acc= 1.00D-08 Lan= 1 ITN= 16 MaxIt= 64 E= -372.9578164278 DE=-1.36D-02 Acc= 1.00D-08 Lan= 1 ITN= 17 MaxIt= 64 E= -372.9619686829 DE=-4.15D-03 Acc= 1.00D-08 Lan= 1 ITN= 18 MaxIt= 64 E= -372.9642236498 DE=-2.25D-03 Acc= 1.00D-08 Lan= 1 ITN= 19 MaxIt= 64 E= -372.9652195918 DE=-9.96D-04 Acc= 1.00D-08 Lan= 1 ITN= 20 MaxIt= 64 E= -372.9656045045 DE=-3.85D-04 Acc= 1.00D-08 Lan= 1 ITN= 21 MaxIt= 64 E= -372.9661566455 DE=-5.52D-04 Acc= 1.00D-08 Lan= 1 ITN= 22 MaxIt= 64 E= -372.9661928826 DE=-3.62D-05 Acc= 1.00D-08 Lan= 1 ITN= 23 MaxIt= 64 E= -372.9661961859 DE=-3.30D-06 Acc= 1.00D-08 Lan= 1 ITN= 24 MaxIt= 64 E= -372.9661963454 DE=-1.60D-07 Acc= 1.00D-08 Lan= 1 ITN= 25 MaxIt= 64 E= -372.9661964451 DE=-9.97D-08 Acc= 1.00D-08 Lan= 1 ITN= 26 MaxIt= 64 E= -372.9661964510 DE=-5.87D-09 Acc= 1.00D-08 Lan= 1 ... DO AN EXTRA-ITERATION FOR FINAL PRINTING ( 1) EIGENVALUE -0.37296620E+03 ( 1) 0.8546868 ( 232)-0.1605760 ( 18) 0.1273223 ( 214)-0.1203281 ( 29) 0.1085574 ( 327) 0.1014540 ( 348)-0.0957454 ( 8257)-0.0880574 ( 1545) 0.0854231 ( 7879)-0.0851818 ( 7881)-0.0818381 ( 20) 0.0795558 ( 68)-0.0769911 ( 1771) 0.0741894 ( 8780) 0.0631177 ( 334) 0.0609166 ( 244)-0.0540400 ( 75)-0.0520658 ( 2018) -0.0480310 ( 63)-0.0473743 ( 8283) 0.0436091 ( 375)-0.0435983 ( 359) 0.0423989 ( 265)-0.0423721 ( 1797)-0.0420108 ( 1626) -0.0403186 ( 1856) 0.0401087 ( 253) 0.0399931 ( 2039) 0.0399314 ( 8801)-0.0386601 ( 7898) 0.0378718 ( 8269)-0.0377589 ( 234) 0.0369503 ( 8787) 0.0363671 ( 8908)-0.0355514 ( 2077)-0.0352455 ( 8030) 0.0348422 ( 8540)-0.0347887 ( 236)-0.0330726 ( 8004) -0.0328432 ( 8643)-0.0324733 ( 298)-0.0324499 ( 1888)-0.0316934 ( 8775)-0.0311804 ( 8839) 0.0292843 ( 1564)-0.0292114 ( 403) -0.0287844 ( 88)-0.0286648 ( 8656) 0.0286427 ( 8142)-0.0286416 ( Final one electron symbolic density matrix: 12345 1 0.000000D+00 2 0.000000D+00 0.000000D+00 3 0.000000D+00 0.000000D+00 0.000000D+00 4 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 5 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 6 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 7 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 8 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 9 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 10 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 6 7 8 9 10 6 0.000000D+00 7 0.000000D+00 0.000000D+00 8 0.000000D+00 0.000000D+00 0.000000D+00 9 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 10 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 0.000000D+00 MCSCF converged. =============================> The above matrix elements are all zero. The followed gradient calculation was of course failed. It produced a very large negative eigenvalue. The optimization go unreasonably next, and failed in several steps. I would appreciate your advice about it. By the way, is unrestricted DFT an acceptable method for the calculation of the first triplet state geometry? I computed some structures and found that the geometries obtained from DFT were very close the the CASSCF result with respect to the CIS calculation. Best regards, Chungen Liu ______________________________________ Chungen Liu, Dr. Chemistry Department Nanjing University =================================================================== (http://mail.sina.com.cn) (http://newchat.sina.com.cn) From chemistry-request@server.ccl.net Mon Mar 26 08:54:45 2001 Received: from ucidoor.unitedcatalysts.com ([208.23.162.2]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QDsjD15175 for ; Mon, 26 Mar 2001 08:54:45 -0500 Received: by ucidoor.unitedcatalysts.com; (8.8.8/1.3/10May95) id IAA10696; Mon, 26 Mar 2001 08:54:44 -0500 (EST) Received: from 10.1.0.50 by ucismtp02.unitedcatalysts.com (InterScan E-Mail VirusWall NT); Mon, 26 Mar 2001 08:54:02 -0500 (Eastern Standard Time) Received: from lvlxch01.unitedcatalysts.com ([10.16.100.88]) by lvlmail.unitedcatalysts.com (PMDF V6.0-24 #41777) with ESMTP id <0GAT00DMA54C6C@lvlmail.unitedcatalysts.com> for chemistry@ccl.net; Mon, 26 Mar 2001 08:50:36 -0500 (EST) Received: by lvlxch01.unitedcatalysts.com with Internet Mail Service (5.5.2650.21) id ; Mon, 26 Mar 2001 08:54:43 -0500 Content-return: allowed Date: Mon, 26 Mar 2001 08:54:43 -0500 From: "Shobe, Dave" Subject: spin-orbit coupling To: "'CCL'" Message-id: <157A51F55AAAD3119CD70008C7B1629DDAAD98@lvlxch01.unitedcatalysts.com> MIME-version: 1.0 X-Mailer: Internet Mail Service (5.5.2650.21) Content-type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by server.ccl.net id f2QDsjD15176 How do I calculate the splitting of the ground electronic state due to spin-orbit coupling? Does this have a noticeable effect on the heat capacity or entropy? --David Shobe Süd-Chemie Inc. phone (502) 634-7409 fax (502) 634-7724 email dshobe@sud-chemieinc.com From chemistry-request@server.ccl.net Mon Mar 26 09:26:44 2001 Received: from mail2.intermedia.net ([206.40.48.152]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QEQhD15418 for ; Mon, 26 Mar 2001 09:26:43 -0500 Received: from mentecolectiva.com (unverified [200.24.89.178]) by mail2.intermedia.net (Rockliffe SMTPRA 4.5.4) with ESMTP id for ; Mon, 26 Mar 2001 06:26:42 -0800 Message-ID: <3ABF5122.E181E512@mentecolectiva.com> Date: Mon, 26 Mar 2001 09:24:34 -0500 From: Mauricio Esguerra Neira X-Mailer: Mozilla 4.7 [en] (Win98; I) X-Accept-Language: en MIME-Version: 1.0 To: chemistry@ccl.net Subject: Summary: Short Budget Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Short Budget Summary On Tue, 13 March 2001 09:27:20 -0800 I wrote the following question: Hello all, I would like to ask for your advice on buying a PC for running gaussian 98 and gamess-us jobs under Slackware Linux 7.1. My budget is U$4000 and I have been looking at Dell computers for the fastest machine I can get with this money. I am undecided on whether to buy a dual pentium III 933MHz or a single pentium IV 1.5 GHz processor both with 1Gb PC800 RDRAM; if any of you out there has any information as to which of this two options would work better, or other suggestion on how to invest my money in order to obtain the fastest performing machine I would greatly appreciate it. ------------------------------------------------------------------------------------------------------------------- Thank you all for responding to my question, it has been very helpful for making an informed choice of the hardware I am going to buy. Respectfully. Mauricio Esguerra Neira Grupo de Quimica Teorica Chemist I received the following answers Date: Tue, 13 Mar 2001 20:34:07 +0000 From: Yubo Fan Organization: Chemistry Department, Texas A&M University Hi, I prefer a dual pentium III 933MHz with 1GB SDRAM (I don't think PC800 RDRAM is much faster than PC133 SDRAM and the bus speed of PIII/933 is 133 MHz). I list the prices of the most important components for your calculations. 2 PIII/933 $500 1GB SDRAM(4x256MB) $450 MSI 694D VIA Dual Socket 370 motherboard $140 40GB ATA100 Hard Drive (7200 rpm) < $150 others: monitor, video card, network card housing, mouse and keyboard So, you can build two dual PIII/933 w/1GB memory Linux machine as I suggest. Of course, you just need one monitor for them. You can remote-control them when you run QM jobs on them. Remember, this type of computer is very good for DFT calculation. Don't try POST-SCF calculations (ie CCSD) on them, or you will be very disappointed. I hope my suggestion gives you some help Sincerely Yubo Date: Tue, 13 Mar 2001 14:46:54 -0600 (CST) From: Mat Halls Dear Mauricio The most cost-effective solution hands-down is a AMD Thunderbird 1.2GHz machine with 1.5Gb SDRAM and a fast harddrive like IBM Deskstar. Best Wishes. Mat Halls Date: Tue, 13 Mar 2001 14:52:08 -0600 (CST) From: "Fred P. Arnold" Organization: Northwestern U. Hello, Short suggestion: PIII (PIV changed the float unit), Server, with 64 bit PC slots at 66MHz, and then put an Ultra-160 SCSI adapter in that, with 2 disks striped together. Since G98, etc, write a fair number of files, this will help your disk performance mightily. -fred Frederick P. Arnold, Jr. NUIT, Date: Tue, 13 Mar 2001 13:29:37 -0800 From: "Phillip D. Matz" Hello, Do you already have a license for G98? If so then that will help. Also, are you getting the binaries from Gaussian, Inc. or the source code? If you have the source code then you will need to buy a compiler (Portland Group compiler costs ~$300 US). Why am I asking these questions? because the software costs can eat up your $4,000 budget really quick. Purchasing any Tier 1 computers (Dell, Gateway, Toshiba, IBM, etc.) is probably the quickest way to throw your money away. Since you are purchasing hardware to perform scientific calculations I can only appeal to the logical side of you with respect to buying the best hardware for the least money. A Pentium4 will not outperform an Athlon or a PIII to the extent that justifies its large pricetag relative to an Athlon or PII system. Since you will be in Linux you could put together a beowulf (4 nodes say) that will far outperform anything you could by from Dell. Bottom line is don't purchase from Dell (first thing), don't buy a PIV for G98 calculations (second thing), and consider a beowulf if you have that kind of money to spend, you could really put together a nice system for 4 thousand dollars. Respectfully, Phillip Matz matz@wsunix.wsu.edu Date: 13 Mar 2001 13:50:28 PST From: Alan.Shusterman@directory.reed.edu (Alan Shusterman) Organization: Department of Chemistry Reed College --- You wrote: I have been looking at Dell computers for the fastest machine I can get with this money. --- end of quote --- Why Dell? Why Pentium? One year ago (which is an eternity when it comes to computer purchases), I saved a lot of money by buying a computer with an AMD Athlon processor instead of a computer with a Pentium processor. Alan Shusterman Date: Tue, 13 Mar 2001 17:17:51 -0500 From: lowther@att.net Eugene.Leitl@lrz.uni-muenchen.de wrote: Unless you use programs that support dual processors, you will not get the same performance out of two as you will one. Currently, I'd go with a good Athlon processor. Ken Date: Tue, 13 Mar 2001 23:43:06 +0100 From: "art'" Hi, FIC is now shipping their first motherboard supporting the DDR memory technology and the Socket A Athlon and Duron processors. It will support older and newer Athlons with 266MHz FSB. The AD11 supports up to 1GB of PC1600 or 2100 DDR SDRAM via two DIMM slots. The AD11 also features; AGP 4X, ATA 100 and integrated audio. The price is about $200. That is a new product, and I have no news about the benchmarks of its tests. DDR memory shoud be very quick as r/wr functions are performed in another way from SDRAM's. The 1.2 GHz Athlon is about $ 280 and mauch cheaper than PIII or P4 .... and Athlons usually have higher performance about 20 %. Espetially in floating point intensive calculations. Here is only my opinion, but the price/performance for Athlons is better. The save money you can use in obtaining good SCSI system if presume high HDD usage during calculations. good luch Arturas Z. Date: Wed, 14 Mar 2001 00:51:13 +0100 (NFT) From: Matthias Mann Hallo, buy two separate systems with AMD Athlon processors for this money. Dual systems are not effective in most cases. Greetings, Matthias -- Dr. Matthias Mann Computational Chemistry, TU Dresden Matthias.Mann@chemie.tu-dresden.de Date: Tue, 13 Mar 2001 20:04:38 -0500 (EST) From: Mark Hahn > I would like to ask for your advice on buying a PC for running gaussian 98 > and gamess-us jobs under Slackware Linux 7.1. My budget is U$4000 and I > have been looking at Dell computers for the fastest machine I can get with > this money. Dell is a reasonable starting point: their machines are solid. their primary disadvantage is that they don't sell anything with AMD Athlon chips, and Dell is never the cheapest. > I am undecided on whether to buy a dual pentium III 933MHz or a single > pentium IV 1.5 GHz processor both with 1Gb PC800 RDRAM; if any of you out PIII/rdram means the i840 chipset, which is universally considered crappy. it is not even vaguely comparable to the PIV/i850's delivered bandwidth. imagine dividing a mundane 600 MB/s among two CPUs versus one CPU drinking > from a river of 1.6 GB/s... considering that the rdram is the vast majority of the price, I think it makes sense to buy a processor/chipset that can actually use it. my understanding is that these codes are extremely bandwidth-intensive, and the fact that the PIV has unexceptional FPU performance is completely hidden by the bandwidth issue. (in other words, PIII's are very bandwidth starved.) Date: Tue, 13 Mar 2001 20:24:34 -0600 From: "Donald B. Kinghorn" Organization: Parallel Quantum Solutions Hi, Believe me, you would be throwing your money away with the Dell. You can build a better machine yourself for half the money. I do system design (among other things) for Parallel Quantum Solutions http://www.pqs-chem.com ... I am not impressed with Dell or Gateway. First, we could build you a system with our parallel quantum chemistry code on it if you are interested. We usually build larger clusters but we have built a few single node dual processor machines. ... If you want to build your own you can save loads of money and have a high quality machine. This is what I would suggest for a great g98 box: Supermicro 370 DLE motherboard -- Serverworks chipset, we've seen very good performance with theses boards [ $300] (2) Intel PIII 1000MHz cpu's about $309 each [$618] (4) PC133 256MB SDRAM Registered ECC CL=2 Micron (Crucial) memory -- 1 GB of high quality low latency memory ... you will see a difference in g98 ... buy it from Crucial online $120 each [part number CT32M72S4R7E] [$480] (2) Western Digital 30GB 7200rpm hard drives and a Promise ATA/100 controller ... $135 for the drives and $35 for the controller -- these things are very fast on the ATA/100 controller, use a 30GB scratch drive for g98 ... [$305] Video card you'll need a PCI card for this motherboard (it has 64bit/66MHz PCI but no AGP) You should be able to find a TNT2 based card with at least 16MB for around $60 (don't use an ATI rage 128 pro there is a conflict with this motherboard stick with a good TNT2 based card) [$60] Network card -- intel ether pro 100 is on the motherboard [0] zip disk [$50] floppy disk [$15] CDRW [$160] case -- maybe IN-WIN ATX server case with 300Watt power supply [$100] Mouse and Keyboard [$35] 19 inch monitor [$300] Lets see that's about $2500 for a great machine and the fun of building it yourself--dual PIII1GHz on a high quality motherboard a gigabyte of high quality ram and 60GB of fast disk space good video and a 19" monitor ... not bad :-) I got most of these prices off the web from MicroPro http://www.mpipc.com Good Luck -Don P.S. you have my permission to post this back to the CCL if you like ... Dr. Donald B. Kinghorn PQS LLC http://www.pqs-chem.com Date: Wed, 14 Mar 2001 08:52:59 +0100 (CET) From: GUTIERREZ Fabien Hi There is no doubt always buy a dual pentium even at lower frequency . We are using the same kind of computer you wanna buy eg precision 420 from DELL under redhat 7 and it works really fine even with gaussian on both processors . see ya Fabien Date: Wed, 14 Mar 2001 10:30:36 +0100 From: Eugene Leitl I'm running Guassian-98 and GAMESS-US (also GAMESS-UK). I'm using athlon 750's that behave like intel's running at 1.15 times the clock speed. Price/performance dictatest that I use athlon's. By the way, I've been using Debian very successfully with these software packages. Art Edwards Date: Wed, 14 Mar 2001 12:41:09 +0100 From: Eugene Leitl Remember, G98 thrashes cache and so needs high speed bandwidth to ram. All my tests have shown that the Thunderbird 750 is the best value for the money with its 133MHz bus for the jobs we're doing. If you are doing smaller molecules, they might fit in L2 cache, especially on the larger cache on P3's. However, without benchmarking your exact jobs its hard to tell. The extra money for Thunderbirds at higher speed in my tests have indicated that its not worth the money (even when extra cabinet space/network infrastructure is included in the cost, along with cost of mainboard, ram, etc). For the $ in fact, few things are as efficient in terms of Mhz speed vs speed of calculation as my Celeron450 for G98 stuff that we're working on - only the Tbird 700 and 750 are as 'efficient' per Mhz as the Celeron. The cpu sits idle, basically, wait stating to death, while ram fetches the next hunk of data. Need a really fast bus. This is why 133Mhz bus dual CPU boards dont hold high hopes for me for what we're doing. I have yet to get back results from DDR tests for comparison. /kc Date: Wed, 14 Mar 2001 17:22:23 +0000 From: Yubo Fan Organization: Chemistry Department, Texas A&MUniversity > Hi, > > FIC is now shipping their first motherboard supporting the DDR memory technology > and the Socket A Athlon and Duron processors. It will support older and newer > Athlons with 266MHz FSB. The AD11 supports up to 1GB of PC1600 or 2100 DDR SDRAM > via two DIMM slots. The AD11 also features; AGP 4X, ATA 100 and integrated audio. > The price is about $200. That is a new product, and I have no news about the > benchmarks of its tests. > I don't think DDR memory is much faster than PC133 SDRAM. You can find the comparison data on the homepage of www.amd.com. CPU is always the most important for calculations. > > DDR memory shoud be very quick as r/wr functions are performed in another way from > SDRAM's. The 1.2 GHz Athlon is about $ 280 and mauch cheaper than PIII or P4 .... > and Athlons usually have higher performance about 20 %. Espetially in floating > point intensive calculations. > I agree Athlon is faster and cheaper than PIII but there isn't dual processor supported motherboard available now. Maybe two months later, this type of motherboard is available commercially. Double speed is always better than 20% faster. > > Here is only my opinion, but the price/performance for Athlons is better. The save > money you can use in obtaining good SCSI system if presume high HDD usage during > calculations. > SCSI hard drive is nearly useless for DFT calculations, I think. PC is not good for POST-SCF calculations. So, you shouldn't waste money on SCSI system. Spend your fund on CPUs and memory. Date: Wed, 14 Mar 2001 22:22:37 +0100 From: "art'" Hi, The bottleneck, usually, is the M/B bus, RAM and L2 Cache. If to choose the best and the most expensive CPU with highest core performance, what is the overall performance when M/B and RAM are far behind with speed and performance ? Athlon has better choice choosing M/B and Cache. So DDRAM, which performs two r or wr functions per clock, could serve well increasing the total performance too. Maybe I am wrong somethere, but let me know ;) Good day A.Ziemys Date: Thu, 15 Mar 2001 12:05:30 +0100 (MET) From: Eugene Leitl On Wed, 14 Mar 2001, art' wrote: > The bottleneck, usually, is the M/B bus, RAM and L2 Cache. If to choose the best and the The bottleneck is specific to the problem. Some codes run entirely within L2 (or even L1) cache, some thrash cache, so bandwith to memory becomes relevant, some stream a lot to nonvolatile storage or across network (where latency or bandwidth may be the bottleneck). The only meaningful benchmark is your application. > most expensive CPU with highest core performance, what is the overall performance when > M/B and RAM are far behind with speed and performance ? Athlon has better choice choosing > M/B and Cache. So DDRAM, which performs two r or wr functions per clock, could serve well > increasing the total performance too. Date: Thu, 15 Mar 2001 10:32:12 -0800 From: "Phillip D. Matz" Hi Eugene, > (where latency or bandwidth may be the bottleneck). The only meaningful > benchmark is your application. Not to start a flamer here or anything, but this thread has a very specific application in mind - Gaussian98. All of this discussion (at least what I've seen thus far) is all related to the hardware factors that affect G98 computation efficiency. So your post is correct in that it doesn't say anything that is incorrect, but at the same time give us credit for knowing what the topic is for what we are still talking about. Respectfully, Phillip Matz Date: Thu, 15 Mar 2001 21:25:21 +0100 From: "art'" Good day CCL'ers, The talk turned to something other way from initial topic, so I should some excuse to Yubo Fan. About Athlon and DDR .... Yes, I have surfed internet right for this topic about DDR benchmarks and Athlon. The improvement of DDR to common memory is not very impressive. I think the essence of DDR is very good, as the new efficient PC memory is market will have its place and we all are interested in high speed and in price affordable RAM. The problem about DDR and Athlon could be that AMD started DDR implementation late compared to Intel. There is possibility, that DDR techniques could became more efficient in Athlon (AMD) systems in the future, as good CPU needs good RAM :)). But this is just philosophy ... I would like to ask about dual Athlon motherboards ... Could anyone point to manufacture of dual M/B for Athlon or post the link to pages concerning dual Athlon systems ? Best regards A.Ziemys > Athlons seem to not be able to make much use from DDR SDRAM's higher > bandwidth, but Gaussian98 might be different in this case. It will be > interesting to compare it with Gaussian98 benchmarks on dual-Athlong DDR > systems (a well-known Beowulfer is currently running a few benchmarks, > hopefully it will include Gaussian98). Date: Thu, 15 Mar 2001 20:26:14 +0000 From: Yubo Fan Organization: Chemistry Department, Texas A&M University Hi, I really agree with Phillip. PC is a very good choice to do QM computations. But there are still some problems when you run jobs on Linux and PC. Sometimes, the system will jammed or respond very very slowly if you run some large DFT jobs (> 1000 basis functions) for a long time. Anyway, UNIX workstations can finish some jobs more efficiently than PCs. I love PCs because it's far cheaper and the speed of running DFT calculations is good. Yubo Date: Thu, 15 Mar 2001 22:55:51 +0100 From: Eugene.Leitl@lrz.uni-muenchen.de "Phillip D. Matz" wrote: > Not to start a flamer here or anything, but this thread has a very specific > application in mind - Gaussian98. All of this discussion (at least what Oh, we're still talking about Gaussian98? Sorry, I thought the thread sidetracked. Of course in this case the problem is not underdefined. > I've seen thus far) is all related to the hardware factors that affect G98 > computation efficiency. So your post is correct in that it doesn't say > anything that is incorrect, but at the same time give us credit for knowing > what the topic is for what we are still talking about. Athlons seem to not be able to make much use from DDR SDRAM's higher bandwidth, but Gaussian98 might be different in this case. It will be interesting to compare it with Gaussian98 benchmarks on dual-Athlong DDR systems (a well-known Beowulfer is currently running a few benchmarks, hopefully it will include Gaussian98). Date: Thu, 15 Mar 2001 14:41:49 -0800 From: "Phillip D. Matz" > bandwidth, but Gaussian98 might be different in this case. It will be Yeah, I'm really curious how well G98 takes advantage of the extra bandwidth of DDR SDRAM. My own tests regarding the Athlon with PC100 and PC133 (CAS2 and 3) indicated that G98 was indifferent to the CAS latency but really benefited by the increase in effective bandwidth in going from PC100 to PC133 (ie by changing the memory FSB asynchronously with the KT133 chipset). > systems (a well-known Beowulfer is currently running a few benchmarks, I doubt Dr. Brown will be benchmarking G98 on the dual-athlon rig but it would be nice (I don't think he has a license but I may be wrong here). Of course I should be asking him rather than lamenting about it here on the CCL I suppose. At any rate it would be nice to have some numbers on a P4 system with the dual-channel rambus just to see if the extra bandwidth really boosts the effective IPC for G98 code (although I doubt it will lower the price/performance enough to justify the cost) simply as a matter of curiosity. From chemistry-request@server.ccl.net Mon Mar 26 13:12:39 2001 Received: from ccl.net ([192.148.249.4]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QICdD17349 for ; Mon, 26 Mar 2001 13:12:39 -0500 Received: from krakow.ccl.net (krakow.ccl.net [192.148.249.195]) by ccl.net (8.9.3/8.9.3/OSC 2.0) with ESMTP id NAA07831; Mon, 26 Mar 2001 13:12:33 -0500 (EST) Date: Mon, 26 Mar 2001 13:12:32 -0500 (EST) From: Jan Labanowski To: chemistry@ccl.net cc: Jan Labanowski Subject: Thank you, and: Please help me keep CCL running... Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear CCL subscribers... I want to thank all of you who decided to support CCL with your donations, contribution of materials, lively discussions, and participation. Thank you for making the CCL possible... This is another infrequent appeal for support of CCL. Since I do not do it every other day, please read through it. CCL needs your support to pay student wages, cover the cost of operation (hardware, software, network access, disposables), and facilities (office space, phone, utilities). The OSC, CCL parent organization, is picking up the tab at the moment. The parent organization is aggressively pursuing its leadership in various computing and communication areas, and nobody on this planet can say with certainty what priorities will emerge in years ahead. To operate in a more stable environment CCL has to share (or better cover) costs of its operation and have some safety cushion. I wish I was able to present the CCL to the OSC management as a self-funding project/program rather than humbly request support every year. OSC will always be here to help CCL, but CCL must contribute to its own existence. It would make life easier for everyone involved if CCL was an asset. I am not sure how much stronger can my public statement be... Of course, I hear advice to go after BIG SPONSORS. Even if it was easy and the BIG SPONSORS were just waiting to give me cash, I have a problem with having a few key supporters. Even if the supporters were angels, it would endanger CCL's non-alliance and independence and would bring "politics". I am not sure if I would like this and/or could actually do this dance. Also, do not tell me to go to the XXX agency for support. They will not do it, and there is even more politics there, since CCL would have to take money from one of you. We can stay as an independent and open forum only if we do not have to count on a handful of big supporters, but are kept in operation by the large number of smaller donations from CCL subscribers/users. You see, if each of you gave me $50 or so, I would not have to get any additional support, and could even set up the rainy days fund. Every donation counts... I understand that many of you just cannot help (e.g., your currency cannot be changed to US Dollars), but most of you can... You can support CCL in 2 ways: 1) GIVE CCL A DONATION which is a convenient way of support for individuals and corporations. Such a donation is tax deductible in many countries (we are an educational institution as a part of Ohio State University). If you need a formal solicitation letter to convince people in your organization than let me know. I will craft a letter about CCL and benefits of supporting it. You are even encouraged to tell me what you want to hear... After we receive your donation, we will thank you for your generosity with a formal letter needed for tax purposes, and we will acknowledge your donation in the CCL web site. Donations need not to be excessive. We will accept a million {:-)}, but many smaller donations will do as well (or better)... 2) BUY STUFF FROM US. This is the most convenient way to support us when using your research funds, i.e., when you need a formal invoice/billing process. And since we are helping directly in your research, there is no problem in justifying your expenditure. The easiest (and we expect, also the most popular) solution is to buy from us one or more sets of CDs which represent the current content of CCL archive. They are priced at $250, and contain about 2 Gigabytes of stuff: messages, data, software, and documents. You can buy one set for work, and one set for home to browse in the evening. We are also willing to share our experience in setting up Web sites, mailing lists, help you maintain sites, and possibly house it, etc. We would be happy to share our experience with nonprofit organizations, associations, or with our subscribers. Of course, details have to be discussed, since we are not in competition with commercial vendors of such services, and our involvement should support our mission: education and research. In short... Please consider supporting CCL. Details are available in our site: http://www.ccl.net/chemistry/aboutccl/supporting and there is also a link to it on our home page at: http://www.ccl.net If you have questions, please contact me directly. I will also be thankful for suggestions and advice on how to improve our solicitation for CCL support. Some of you, who know me personally, are aware that I am not a salesman type and I would hate to bug you directly. At the same time I am committed to keep CCL alive since it is in my opinion a useful resource. Please do not force me to do "direct marketing" since I am not sure I could take it... Please help... Jan K. Labanowski | phone: 614-292-9279, FAX: 614-292-7168 Ohio Supercomputer Center | Internet: jkl@ccl.net 1224 Kinnear Rd, | http://www.ccl.net/chemistry.html Columbus, OH 43212-1163 | http://www.ccl.net/ From chemistry-request@server.ccl.net Mon Mar 26 15:23:28 2001 Received: from sirene.rz.uni-duesseldorf.de ([134.99.128.2]) by server.ccl.net (8.11.0/8.11.0) with ESMTP id f2QKNRD18589 for ; Mon, 26 Mar 2001 15:23:27 -0500 Received: from modem.rz.uni-duesseldorf.de (Isis100.urz.uni-duesseldorf.de [134.99.138.100]) by neptun.rz.uni-duesseldorf.de (Sun Internet Mail Server sims.4.0.2000.05.17.04.13.p6) with SMTP id <0GAT000IGNATV0@neptun.rz.uni-duesseldorf.de> for chemistry@ccl.net; Mon, 26 Mar 2001 22:23:27 +0200 (MET DST) Date: Mon, 26 Mar 2001 22:23:34 +0200 From: petra imhof Subject: Re: CCL:Again about CASSCF geometry optimization In-reply-to: <20010326123625.6305.qmail@sina.com> X-Sender: imhof@mail.rz.uni-duesseldorf.de To: cgliu Cc: chemistry@ccl.net Message-id: <3.0.6.32.20010326222334.007da100@mail.rz.uni-duesseldorf.de> MIME-version: 1.0 X-Mailer: QUALCOMM Windows Eudora Light Version 3.0.6 (32) Content-type: text/plain; charset=us-ascii Content-transfer-encoding: 7BIT At 20:36 26.03.01 +0800, you wrote: >Dear cclers: > >I got some suggestions for my last question on this topic. I am sorry I had not stated it clearly. > >I am doing a CASSCF optimization of Decapentaene( C10H12), to find the gometry of the ground state and the first excited state structure. > >The program I used is Gaussian 98 > Though it will not help you: you are not the first one who tries in vain to run a CASSCF calculation with this number of CSFs using Gaussian. "Promises" of CASSCF calculations with up to 12 orbitals are not kept -or maybe with only two electrons? At least to my knowledge no one ever published the iops necessary for this. I wonder how M. Klene, M.A. Robb, P. Celani and M. J. Frisch managed the (14,14) calculation they reported as example for parallelization (J. Chem. Phys. 113 (2000) 5653). (If anyone "official" of Gaussian Inc. is reading- if you please have the answer...) Sorry for not helping but only complaining, Petra http://www-public.rz.uni-duesseldorf.de/~imhof/ From chemistry-request@server.ccl.net Mon Mar 26 16:28:55 2001 Received: from chemsun.chem.umn.edu ([128.101.157.2]) by server.ccl.net (8.11.0/8.11.0) with SMTP id f2QLSsD18889 for ; Mon, 26 Mar 2001 16:28:55 -0500 Received: (qmail 46051 invoked by uid 1120); 26 Mar 2001 15:28:56 -0600 Received: from localhost (sendmail-bs@127.0.0.1) by localhost with SMTP; 26 Mar 2001 15:28:56 -0600 Date: Mon, 26 Mar 2001 15:28:56 -0600 (CST) From: Lakshmi Devi Kesavan To: chemistry@ccl.net Subject: CCL:Semi-empirical parameters for FE In-Reply-To: <20010326123625.6305.qmail@sina.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hello all, I am looking for PM3 parameters for the transition metal 'Fe'. Thanking you in advance, Lakshmi Kesavan. ************************* Lakshmi S Devi Kesavan Graduate Student Department of Chemistry 207, Pleasant Street S.E, University of Minnesota Minneapolis, MN 55455-0431 Off # : 612-625-5325 ************************* From chemistry-request@server.ccl.net Mon Mar 26 21:13:29 2001 Received: from mail.delanet.com ([63.250.63.182]) by server.ccl.net (8.11.0/8.11.0) with SMTP id f2R2DSD19879 for ; Mon, 26 Mar 2001 21:13:28 -0500 Received: (qmail 46074 invoked by uid 0); 26 Mar 2001 23:29:16 -0000 Date: 26 Mar 2001 23:29:15 -0000 Message-ID: <20010326232915.46073.qmail@mail.delanet.com> From: inter7@mail.delanet.com To: CCL Subject: CCL:Queuing systems Content-Type: text/plain; charset=iso-8859-2 Content-Transfer-Encoding: 7bit Sender: "Computational Chemistry List" Errors-To: chemistry-request@ccl.net Precedence: bulk Hi, A few weeks ago, I asked about the best queuing system available for a cluster of PC. This topic has already been discussed, and PBS (www.openpbs.com) seems to be a good one, especially when using patches to improve it. (see archives for PBS to have some examples.) Another system is Codine 5.2.2 which was released for FREE (http://www.sun.com/gridware/). We have not yet tested these systems so I cannot say more by now. Paul. -- ------------------------------------------------------------------------- Fleurat-Lessard Paul | fleurat@lpct.u-bordeaux.fr Laboratoire de Physico Chimie Moleculaire, | Phone: (33)(0)5 56 84 63 09 Universite Bordeaux I, 33400 Talence, FRANCE| Fax : (33)(0) 5 56 84 66 45 _________________________________________________________________________ -= This is automatically added to each message by mailing script =- CHEMISTRY@ccl.net -- To Everybody | CHEMISTRY-REQUEST@ccl.net -- To Admins MAILSERV@ccl.net -- HELP CHEMISTRY or HELP SEARCH CHEMISTRY-SEARCH@ccl.net -- archive search | Gopher: gopher.ccl.net 70 Ftp: ftp.ccl.net | WWW: http://www.ccl.net/chemistry/ | Jan: jkl@ccl.net