CCL:motherboard for AMD processor (summary) (fwd)

Eugene Leitl Eugene.Leitl at lrz.uni-muenchen.de
Wed Jul 25 09:34:21 PDT 2001


-- Eugen* Leitl <a href="http://www.lrz.de/~ui22204/">leitl</a>
______________________________________________________________
ICBMTO  : N48 10'07'' E011 33'53'' http://www.lrz.de/~ui22204
57F9CFD3: ED90 0433 EB74 E4A9 537F CFF5 86E7 629B 57F9 CFD3

---------- Forwarded message ----------
Date: Wed, 25 Jul 2001 11:30:33 -0400 (EDT)
From: Borislav Kovacevic <boris at spider.irb.hr>
To: chemistry at ccl.net
Subject: CCL:motherboard for AMD processor (summary)



Dear CCL'ers,
I would like to say thanks to all people who helped me to make
decision which computer configuration to buy.


Here is the summary.

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

From: David Konerding <dek at cgl.ucsf.EDU>

I have the A7A266 and it works great.  I am running Windows 2000 on it
but also spent some time running linux on it.  I have the DDR and th emachine
is a real screamer.  DDR gets you maybe 15% better memory bandwidth, which
would probably be seen while running gaussian.  I would try the machine with
the DDR RAM and then swap it out for SDRAM (that's the nice thing about
the board- both are supported) and see how your timings differ.

Yes, portland's compiler will run on the AMD.

Dave

############################################################################
From: Scott Anderson <anderson at chem.utah.edu>

I have an 1GHz thunderbird on an asus A7V socket 7 motherboard with PC 133
ram.  It works find with gaussian and with the PGI compiler


Prof. Scott L. Anderson
Department of Chemistry
University of Utah
315 S. 1400 E. Rm 1216
Salt Lake City, UT  84112
(801)585-7289
FAX(801)581-8433

##########################################################################
From: Niclas Andersson <nican at nsc.liu.se>

For what it's worth...

I have no experience of DDR SDRAM yet. We have used ASUS Motherboards for a
long time and are very pleased with the stability.

I would go for the A7M266. The ALi Magik chipset on the A7A266 haven't shown
that good performance. The VIA chipset for DDR (ASUS has no MB with that) also
has problems. I am hoping VIA soon will fix their chipset. Their eariler
versions for AMDs and SDRAM har performed very well. I assume if VIA can get a
chipset working good for AMD and DDR, that will become the "default" chipset
since AMD "... really don't want to produce chipsets". The best chipset for
Thunderbird and DDR is now AMDs own which is mounted on the A7M266.

With todays prices and derivatives of prices. DDR is the way to go. Unless you
can prove you don't need the extra memory bandwidth. But with Gaussian and
Games I reckon you do.

The only nagging problem is the larger and larger non-ECC memory ...

I can't see any problems using Portland Compilers in this setup. As I said,
we've been using AMD in several clusters and Portland is the compiler to use
for Fortran! It's stable, it is forgiving (easy to port), and the performace is
good. It can generate 3DNow instructions but that's not that interesting unless
you use singe precision. For C and C++, GCC is an alterative not because of
performance but because it is the native compiler.

Don't forget extra cooling fan in the box. 1.4GHz Thunderbirds run hot!


Our current clusters: <http://www.nsc.liu.se/beowulf>

Regards,
	Niclas
-- 
Niclas Andersson                                E-mail: nican at nsc.liu.se
National Supercomputer Centre                       Phone: +46 13 281464
Linkoping University, S-581 83 Linkoping, Sweden      Fax: +46 13 282535

############################################################################
From: Antoine <antoine.fortune at ujf-grenoble.fr>


My PC is a T-bird 1.2GHz + 256 Mo DDR on a MSI K7T266Pro mobo and it works fine
! MSI mobo is linux certified and all reviews I saw present that mobo more
reliable than A7A266 but a bit slower so they worth.
It is fine for overclocking, Asus is better.
It is based on a VIA chipset.
If you look for pure performences you migh wait for the SIS735 chipset based
mobo that should be released in a couple of monthes and should be better than
AMD  or VIA chipsets.
DDR is the good choise if you want to upgrade your PC later, DDR is becoming the
new RAM standard even if the speed gain is just about 15% (vs DRAM133). It is
going cheaper and cheaper.
AMD proc run Portland compiler without problems, they are i586 complients.

Hope this help even if I don't really answer your question ...

Antoine

Antoine FORTUNE

Ingenieur Modelisation Moleculaire
Dept. Pharmacologie Moleculaire - UMR 5063
Universite Joseph Fourier

Faculte de Pharmacie
38706 LA TRONCHE  - FRANCE

Tel: 04 76 63 71 40   Fax: 04 76 51 86 67
e-mail: Antoine.Fortune at ujf-grenoble.fr

#############################################################################
From: a3arzi <Arturas.Ziemys at vaidila.vdu.lt>


Im my mind, it is very good choice.

I have no detail knowledge about your M/B, but take a look if it has
VIA chipset. I have heard that VIA have made its chipsets well optimised
for AMD CPU. RAM is extensively used in ab initio calculations. So DDRAM
should increase the performance.

Insure that your RAM and M/B both run 266 MHz FSB.

Best regards
Arturas Z.

############################################################################
From: Bruno Sopko <b_sopko at yahoo.com>

we have used the configuration, you have mentioned,
and encountered quite a few problems. However, the
similar price is for two processor system (Intel PIII
750MHz) with normal 133 SDRAM 1Gb, which worked fine,
in some cases even faster , without any problems,
patches and so on.
That's more or less all. If you have any other
questions, do not hesitate to contact me

Bruno Sopko
Dept. of Biochemistry,
Charles University - Faculty of Science,
Albertov 2030
128 40 Praha 2, Czech Republic,
Phone: +420-2-21952173
       +420-607-836499
FAX:   +420-2-21952331

##########################################################################
From: David van der Spoel <spoel at xray.bmc.uu.se>

I've just installed a MicroStar motherboard w 1.4 GHz Athlon. I did
however get 1.5 Gb of SDRAM with it for Gaussian calculations, memory is
cheap, 512 Mb < 100 US$. The amout of memory will have a large impact on
Gaussian calculations, much more than the type of memory. I only regret I
couldn't install more memory on the motherboard. As for Protland compiler,
you have to have it because Gaussian will not compile without it.

-- 
Groeten, David.
________________________________________________________________________
Dr. David van der Spoel, 	Biomedical center, Dept. of Biochemistry
Husargatan 3, Box 576,  	75123 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://zorn.bmc.uu.se/~spoel
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

##########################################################################

From: Dr. Peter Burger <chburger at aci.unizh.ch>

yes, just skip it and do for the A7M266 or go for any other AMD760
chipset. SSR RAM is really rather nice. Stream bandwidth test on the
A7M266 gives ca. 1 GB/sec and just 700 MB/sec on the A7A266. That really
shows up in the quntum chemistry programs.

> motherboard (for example A7M266)? Then, is it better to use DDR RAM or
> SDRAM? I'm also interested to find out may I use Portland compiler on this

DDR is ca. 10-15 faster. Pgroups compiler works nicely (including
automatized prefetch for the Athlon).

Hope thi helps.

Peter
>........................................
Peter Burger
Anorg.-chem. Institut
Universitaet Zuerich

###########################################################################
From: jmmckel at attglobal.net

There has been a benchmark posted on CCL about Gaussian and various cpu's and
memory.  I believe that it showed that DDR memory was better than SDR.  At least
in the US the price differential is small.

John McKelvey




 ===========================================================================

  Boris Kovacevic			Phone: (385) (1) 4561-117
  Quantum chemistry group		Fax:   (385) (1) 4680-197
  "Rudjer Boskovic" Institute		e-mail: borislav.kovacevic at irb.hr
  Bijenicka 54, 10000 Zagreb
  Croatia

 ============================================================================



-= This is automatically added to each message by mailing script =-
CHEMISTRY at ccl.net -- To Everybody  | CHEMISTRY-REQUEST at ccl.net -- To Admins
MAILSERV at ccl.net -- HELP CHEMISTRY or HELP SEARCH
CHEMISTRY-SEARCH at ccl.net -- archive search    |    Gopher: gopher.ccl.net 70
Ftp: ftp.ccl.net  |  WWW: http://www.ccl.net/chemistry/   | Jan: jkl at osc.edu










More information about the Beowulf mailing list