From chemistry-request "-at-" ccl.net Wed Apr 28 13:14:42 2004 Received: from nightshade.noc.ucla.edu (nightshade.noc.ucla.edu [169.232.48.18]) by server.ccl.net (8.12.8/8.12.8) with ESMTP id i3SIEfkh007127 for ; Wed, 28 Apr 2004 13:14:41 -0500 Received: from azalea.noc.ucla.edu (azalea.noc.ucla.edu [169.232.47.12]) by nightshade.noc.ucla.edu (8.12.11/8.12.11) with ESMTP id i3SIG38Y025058 for ; Wed, 28 Apr 2004 11:16:03 -0700 Received: from ucla.edu (Yang.seas.ucla.edu [128.97.82.197]) (authenticated bits=0) by azalea.noc.ucla.edu (8.12.11/8.12.11) with ESMTP id i3SIFwF0011355 for ; Wed, 28 Apr 2004 11:16:03 -0700 Message-ID: <408FF44F.4070707{at}ucla.edu> Date: Wed, 28 Apr 2004 11:13:35 -0700 From: Douglas Sievers User-Agent: Mozilla Thunderbird 0.5 (X11/20040208) X-Accept-Language: en-us, en MIME-Version: 1.0 To: chemistry{at}ccl.net Subject: Summary: Gaussian 03 32-bit binaries on AMD 64 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Probable-Spam: no X-Spam-Hits: -4.901 X-Scanned-By: vscan.smtp.ucla.edu X-Spam-Status: No, hits=0.0 required=7.5 tests=none autolearn=no version=2.61 X-Spam-Checker-Version: SpamAssassin 2.61 (1.212.2.1-2003-12-09-exp) on servernd.ccl.net All, Here is a summary of replies to my original message (bottom). Summarizing an answer, the 32-bit binaries run without a hitch on SuSe linux on AMD64. Also, there is a response from Gaussian stating that they will not test or support Fedora, and they also suggest using SuSe linux. Reply 1 --------------------------- Hi Doug, we have a cluster built with 2 GHz AMD Opteron processors. We have run the 32-bit binaries of Gaussian98 on this system without any problems. Other program's 32-bit binaries have also run without a hitch on this system (as advertised by AMD). We are running SuSE Enterprise and like it alot (system admin is very intuitive). good luck, Kirk Peterson ---------------------------- Reply 2 (from Gaussian support) ---------------------------- Dr. Sievers, We have tested 32 bit binaries and with some versions of Linux they do run. It is a big waste since you won't have access to one of the best features, large disk files and larger memory because of 64 bit addressing. We are finishing the Opteron port and it will be available in our next minor revision. It will be available as an additional platform addendum to your existing license at that time. We have tested with SuSE Linux. Given that Fedora is not going to be a supported OS from RedHat we don't plan on testing or supporting it. I highly recommend you look into SuSE. We are using the Portland Group compilers but if you stick with a binary license that is not an issue. We will have details on what we tested and support at release. ---------------------------- Reply 3 ---------------------------- Dear Doug, I got forwarded your e-mail by a friend. We are running 32bit binaries on 64bit processors. We have not had any major problems with the recent operating system releases SuSe 64 bit linux. We have used binaries supplied on the Gaussian CD (i.e. precompiled binaries) and they work fine. We have used old G98 binaries from a 32bit Red Hat system and they work fine. The only thing that might be a problem is that we have not been able to get a file bigger than 2GB. This is a limitation of the 32 bit binaries. Does this help? Regards Mark Vincent P.S. I should state there was an initial problem with the 64b operating system, but it is fine now. ----------------------------------------------------- > Hi CCLers, > > I am planning to build a workstation with an AMD Athlon 64 Processor (3400+ to be specific). My university > has a site-license for Gaussian 03, but we do not have the source code (at least I do not believe we do). > > I have binaries available for Intel and AMD 32-bit architecture under Linux. Supposedly, the Athlon 64 > processors can run 32 bit and 64 bit applications. I was wondering if anyone has had experience running > 32-bit Gaussian binaries on one of these 64-bit(32-bit compatible) systems. > > I am not so much worried about optimizing performance of Gaussian, as I am in the operability of > GaussView and Gaussian itself. I am planning to run the calculations on a separate cluster, unless they > are small or test calculations that could be handled by a single CPU efficiently. > > Please include what OS you are using in the reply. I would likely choose Fedora Core 1 for AMD64 as > my platform, although I am open to suggestions for alternatives. > > Thank you all! > > Doug Sievers