From ccl@www.ccl.net Wed Apr 8 09:23:07 1998 Received: from ccl.net for ccl@www.ccl.net by www.ccl.net (8.8.3/950822.1) id IAA11864; Wed, 8 Apr 1998 08:48:28 -0400 (EDT) Received: from liposome.genebee.msu.su (eugene@liposome.genebee.msu.su [158.250.34.92]) by ccl.net (8.8.6/8.8.6/OSC 1.1) with ESMTP id IAA11000 for ; Wed, 8 Apr 1998 08:47:46 -0400 (EDT) Received: (from eugene@localhost) by liposome.genebee.msu.su (8.8.5/8.8.5) id QAA16012; Wed, 8 Apr 1998 16:47:16 +0400 From: Eugene Leitl MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Date: Wed, 8 Apr 1998 16:47:15 +0400 (MSD) To: Michael Hodgson Cc: "Brent H. Besler" , chemistry@ccl.net Subject: CCL:Re: CCL:A General Statement about Non-Intel x86 Processors In-Reply-To: References: <9803271550.AA07421@ouchem.chem.oakland.edu> X-Mailer: VM 6.43 under 20.4 "Emerald" XEmacs Lucid Message-ID: <13611.27441.926717.360810@liposome.genebee.msu.su> Michael Hodgson writes: > I would however agree here, but for different reasons. > > 1) Compilers, kernals etc, can often be optomised for a generation of > processor. This means intel, as they largely control the standard. > (Based largely on linux observations) Definitely true. I have consulted the passage from an pdf document on AMD K6 optimization from AMD's site in regards to x86-deviating specifics. It is a very long list, and afaik no C compiler is supporting this currently. Imo, there is definitely lots of room in code optimization for the AMD K6 family of processors. The same also applies to the Alpha AXP line, though now there is egcs. FYI, Alpha will be going 1 GHz sometime soon. > 2) Upgradability or "Future proofing". Give your machine the longest life > you can. If you buy pentium II the chances are your memory, HDD MB etc > will last for a number of years, and with intel promising 450+ cpu speeds They may last a decade, or more, but they will be distinctly outperformed by then. We now have 100 MHz SDRAM modules and motherboards. We are likely to get RamBus, or similiar memory next. If you talk SCSI, then _which_ SCSI? Fibre channel is in the pipeline. If you think EIDE, there is UltraDMA, which is probably not the last thing to come. I don't think it is wise to purchase top-notch hardware and expect to run it until it rusts. Planning for incremental upgrades is imo the way to go in current rapidly changing markets. > by the end of the year there is still a lot of expansion left in them. If But at which cost? Both $$$'s, and the damage to the marketplace. A monopoly's maxime is profit maximization, not optimal technology for the customer. Haven't you heard the developments for Merced planned to be NDA, thus effectively shutting out free source OSsses, mostly free Unices? Do you really want to support a company with such a policy? > you still need more power you can go for a duel processor machine, but > multiprocessors again means intel (they also have to be the same speed). There is an alternative specification for SMP AMD complies to, but there are not motherboards supporting it. FYI, AMD & Cyrix very recently have joined forces, having adopted a common MMX standard (AMD's). (Don't laugh, SIMD in a register is great for certain codes). > Going for a K6 however and you have to ask how much more can they get out > of socket 7 ? Maybe 300Mhz, but that's gonna be one hot cpu :) This is also wrong. AMD K6/266, which is a very cool CPU, should be now available in quantities. AMD K6 3D, soon to come out, will feature 300 MHz. Of course the concurrence labours against the handicap of DECs high-frequency technology having been bought by Intel. We should not forget that DEC has been forced to this unpopular decision by Intel's monopolist market position. > 3) At the moment the bottom end of the pentium II market is really very > cheap indeed and you would not save much in buying AMD. If you are > looking at upgrading a current system, then it's a different matter, but > buying new I think the PII wins every time on price/performance. I think it depends on your problem. For MD a few-node switchless (multiple NICs/node) AMD K6/233 (266?) Beowulf would seem ideal. Investing the $$$ difference into a 100 MHz motherboard, 1 MByte 2nd level cache and 2 UltraDMA-capable HDs in RAID 0 (disk striping) should compensate somewhat for its slight float weakness (but do a benchmark! The difference may be less than you expected). FYI, Mesa, the free OpenGL clone now supports Voodoo chipsets, which nicely complement a Matrox Millenium II with its raw xstones. A dual CPU Pentium II machine with SMP Linux would also seem a choice. > -Michael Regards, Eugene From nmills@eefus.colorado.edu Thu Apr 9 14:05:27 1998 Received: from eefus.colorado.edu for nmills@eefus.colorado.edu by www.ccl.net (8.8.3/950822.1) id NAA25428; Thu, 9 Apr 1998 13:33:28 -0400 (EDT) Received: (from nmills@localhost) by eefus.colorado.edu (8.6.9/8.6.9/CNS-3.5) id LAA19749 for chemistry@www.ccl.net; Thu, 9 Apr 1998 11:17:06 -0600 From: Nancy Mills Message-Id: <199804091717.LAA19749@eefus.colorado.edu> Subject: IGLO running on SGI? To: chemistry@www.ccl.net Date: Thu, 9 Apr 1998 11:17:06 -0600 (MDT) X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit I would like to run IGLO on a SGI O2 10K. I understand that IGLO was originally written to run on IBM RS-6000 computers. Has anyone ported it to the SGI. If so, I would appreciate some information about problems you encountered. Thanks much. Professor Nancy Mills, Department of Chemistry Trinity University, San Antonio, TX 78212-7200 nmills@trinity.edu on leave at the University of Colorado at Boulder, 8/97-7/98 303-492-0998 email will be forwarded From elewars@alchemy.chem.utoronto.ca Thu Apr 9 16:23:34 1998 Received: from alchemy.chem.utoronto.ca for elewars@alchemy.chem.utoronto.ca by www.ccl.net (8.8.3/950822.1) id PAA25984; Thu, 9 Apr 1998 15:39:10 -0400 (EDT) Received: (from elewars@localhost) by alchemy.chem.utoronto.ca (8.7.4/8.7.3) id PAA13994 for chemistry@www.ccl.net; Thu, 9 Apr 1998 15:39:08 -0400 (EDT) Date: Thu, 9 Apr 1998 15:39:08 -0400 (EDT) From: "E. Lewars" Message-Id: <199804091939.PAA13994@alchemy.chem.utoronto.ca> To: chemistry@www.ccl.net Subject: HERE IS A REF FOR ZPE'S 1998 May 9 ZPE reference Regarding the recent question about the ZPE correction to total energy: the best source of ZPE corrections, for various correlation methods and basis sets, is A. P. Scott and L. Radom, JPC, 1996, 100, 16502. E. Lewars ========================= From shenkin@still3.chem.columbia.edu Thu Apr 9 17:23:41 1998 Received: from mailrelay1.cc.columbia.edu for shenkin@still3.chem.columbia.edu by www.ccl.net (8.8.3/950822.1) id RAA26467; Thu, 9 Apr 1998 17:01:04 -0400 (EDT) Received: from still3.chem.columbia.edu (still3.chem.columbia.edu [128.59.112.36]) by mailrelay1.cc.columbia.edu (8.8.5/8.8.5) with SMTP id RAA19373; Thu, 9 Apr 1998 17:00:58 -0400 (EDT) Received: by still3.chem.columbia.edu (950413.SGI.8.6.12/930416.SGI.AUTO) id RAA03834; Thu, 9 Apr 1998 17:00:57 -0400 From: "Peter Shenkin" Message-Id: <9804091700.ZM3832@still3.chem.columbia.edu> Date: Thu, 9 Apr 1998 17:00:56 -0400 In-Reply-To: Nancy Mills "CCL:IGLO running on SGI?" (Apr 9, 11:17am) References: <199804091717.LAA19749@eefus.colorado.edu> X-Mailer: Z-Mail (3.2.3 08feb96 MediaMail) To: Nancy Mills , chemistry@www.ccl.net Subject: Re: CCL:IGLO running on SGI? Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii On Apr 9, 11:17am, Nancy Mills wrote: > Subject: CCL:IGLO running on SGI? > I would like to run IGLO on a SGI O2 10K. I understand that IGLO was > originally written to run on IBM RS-6000 computers. Has anyone > ported it to the SGI. If so, I would appreciate some information > about problems you encountered. ... If you mean Iris GL on OpenGL (the intermediate layer that allows OpenGL-based graphics adapters to run applications using the older IrisGL graphics libraries), then no porting is necessary; all SGI's come with IGLO. I think it even comes installed by default. -P. -- ************** In memoriam, Grandpa Jones, 1913-1998, R.I.P. ************** * Peter S. Shenkin; Chemistry, Columbia U.; 3000 Broadway, Mail Code 3153 * ** NY, NY 10027; shenkin@columbia.edu; (212)854-5143; FAX: 678-9039 *** *MacroModel WWW page: http://www.columbia.edu/cu/chemistry/mmod/mmod.html *