Summary Opteron vs Nocona
- From: Luigi Cavallo <cavallo _(a)_ chemistry.unina.it>
- Subject: Summary Opteron vs Nocona
- Date: Tue, 17 May 2005 00:26:29 +0200 (MET DST)
Dear all,
this is the collection of all the very useful responses I got.
making a long story short, it seems that opterons are much more balanced than
noconas, and that they are the clear winner when memory bandwith (or memory in
general)
is of relevance. Noconas seems to be better when you only need floating point
speed.
However, there's one point about opterons. AMD just announced dual-core cpus,
which should perform far better than the single-core shipped now.
I contacted resellers here in italy, and they say that dual-core cpus will be
shipped after the summer. If you buy opetrons now they can sell you single-core.
Is this true ? Any out of there has been able to buy AMD with dual-core cpus as
the 275 ?
By,
Luigi
Hi,
we are going to invest some money for a few computers, and we have to
make a decision between the AMD-Opterons and the Intel-Nocona. What's
better ? We are experienced with the Opterons, but we have no idea about
the Noconas...
Major codes to run on them will be classical QM packages as ADF, G03, TM,
some AIMD as CPMD, and possibly some classical MD as gromacs.
Thanks,
Luigi
-= This is automatically added to each message by the mailing script =-
To send e-mail to subscribers of CCL put the string CCL: on your Subject: line
and send your message to: CHEMISTRY _(a)_ ccl.net
Send your subscription/unsubscription requests to: CHEMISTRY-REQUEST _(a)_
ccl.net
HOME Page: http://www.ccl.net
| Jobs Page: http://www.ccl.net/jobs
If your is mail bouncing from ccl.net domain due to spam filters, please
use the Web based form from CCL Home Page
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
===========================================================================
> From diep _(a)_ xs4all.nl Tue May 17 00:04:20 2005
Date: Sat, 07 May 2005 22:59:39 +0200
From: Vincent Diepeveen <diep _(a)_ xs4all.nl>
To: Eugen Leitl <eugen _(a)_ leitl.org>
Cc: cavallo _(a)_ chemistry.unina.it
Subject: CCL:Opteron or Nocona ? (fwd fromcavallo _(a)_ chemistry.unina.it)
hi Eugen,
Nocona has 4 cycle L1 cache ^at^ 1 slot and compares well to prescott core.
Very ugly bad for most software. Its IPC for
all my software is lower than the old Xeons.
benchmark of latest prescott dual core P4 versus dual core opterons:
http://www.sudhian.com/showdocs.cfm?aid=667&pid=2543
Please note Nocona is not dual core yet. Opteron is dual core.
So practical opteron like 3 to 4 times faster for most software. If
software in question uses SSE2, then just hack the code a bit, so that SSE2
works, that
usually speeds up incredible for Opteron, especially latest generations.
The few benchmarks where Nocona is faster than Opteron, without exception
they use software where the SSE2 has been enabled for Nocona and disabled
for opteron. Way to compare.
Real amateuristic compares.
You'll have to wait until Q1 2007 before there is a good new Xeon processor
i guess.
www.pricewatch.com for prices in shops of the different cpu's. Very cheap
to buy in USA usually, even when living in europe. At most 20% import tax.
Vincent
===========================================================================
> From stuart.midgley _(a)_ anu.edu.au Tue May 17 00:04:31 2005
Date: Sun, 08 May 2005 13:02:36 +1000
From: Stuart Midgley <stuart.midgley _(a)_ anu.edu.au>
To: Eugen Leitl <eugen _(a)_ leitl.org>
Cc: Beowulf _(a)_ beowulf.org, Luigi Cavallo <cavallo _(a)_
chemistry.unina.it>
Subject: Re: [Beowulf] CCL:Opteron or Nocona ? (fwd from
cavallo _(a)_ chemistry.unina.it)
Morning
It highly depends on your code and whether you are going to run more
than a single cpu in a box. If you are only going to purchase single
cpu boxes, I think they are more or less the same.
If, however, you are going to run more than a single cpu box, I think
it is fair to say, that the Opterons should give better performance.
Just look at the spec rate fp base and you will see that a dual
processor opteron system out perform the Nocona systems by a long way.
Each opteron chip has its own memory controller. So, if you have 2
cpu's, you have 2 memory controllers and your memory bandwidth scales
linearly. Nocona chips sit on a bus with a single memory
controller. If you have 2 cpu's you still only have 1 memory
controller, so each chip, on average, sees only 1/2 the memory
bandwidth.
The other issue is memory latency. Because the Opterons have the
memory controller on the actual cpu chip, their latency to memory is
very very low, which gives an effective increase in memory
bandwidth. The Nocona chips have the memory controller off chip, so
their latency is high, reducing effective memory bandwidth.
There are other minor difference, but the memory sub-system is the
one that gives the biggest performance difference.
Stu.
==================================================================================
> From chinet.pesch _(a)_ attglobal.net Tue May 17 00:04:41 2005
Date: Sun, 08 May 2005 07:32:30 -0400
From: "chinet.pesch _(a)_ attglobal.net" <chinet.pesch _(a)_
attglobal.net>
To: eugen _(a)_ leitl.org, beowulf _(a)_ beowulf.org
Subject: RE: [Beowulf] CCL:Opteron or Nocona ? (fwd
fromcavallo _(a)_ chemistry.unina.it)
Resent-Date: Sun, 08 May 2005 15:31:06 +0200
Resent-From: eugen _(a)_ leitl.org
Resent-To: cavallo _(a)_ chemistry.unina.it
[ The following text is in the "iso-8859-1" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
...to make a long story short: go for the opteron
Paul
Original Message:
-----------------
From: Eugen Leitl eugen _(a)_ leitl.org
Date: Sat, 7 May 2005 19:23:01 +0200
To: Beowulf _(a)_ beowulf.org
Subject: [Beowulf] CCL:Opteron or Nocona ? (fwd
fromcavallo _(a)_ chemistry.unina.it)
==================================================================================
> From wharman _(a)_ prism.net Tue May 17 00:04:54 2005
Date: Sun, 08 May 2005 13:16:29 -0600
From: William Harman <wharman _(a)_ prism.net>
To: 'Luigi Cavallo' <cavallo _(a)_ chemistry.unina.it>
Subject: RE: Opteron or Nocona ?
Luigi
Have you considered PowerPC processors form Apple or IBM? Check out
www.terrasoftsolutions.com
Bill Harman,
Salt Lake City office
P - (801) 572-9252 F - (801) 571-4927
wharman _(a)_ prism.net
billharman _(a)_ comcast.net
skype: harman8015729252
==================================================================================
> From m.somers _(a)_ chem.leidenuniv.nl Tue May 17 00:05:01 2005
Date: Mon, 09 May 2005 09:38:06 +0200
From: mark somers <m.somers _(a)_ chem.leidenuniv.nl>
To: Luigi Cavallo <cavallo _(a)_ chemistry.unina.it>
Cc: chemistry _(a)_ ccl.net
Subject: Re: CCL:Opteron or Nocona ?
[ The following text is in the "iso-8859-1" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
Dear Luigi Cavallo,
we have recently purchased and setup a new Beowulf cluster here in Leiden,
dedicated to run ADF and other DFT codes (Dacapo and Vasp) on it. We have
found the Nocona's, in combination with the Intel compilers, to be the best
combination especially for ADF and ADF-BAND.
Before we decided to go for the Nocona's, we did try with two test machines,
one being a dual-cpu AMD Opteron 246 (2.0 GHz) and the other being a dual-cpu
Nocona (3.0 GHz). Tests showed that the AMD has, in general, a slightly
better cache hierarchy and a better memory scalability, but also that the
lack of compilers able to tune for the Opterons is severe. Of course we took
the differences in cpu clocks into account.
As it turns out, after having contacted the SCM people in Amsterdam, ADF can
effectively use the SSE3 registers and cpu intruction set and that makes it
run fast on the Nocona's. This, together with the prices being roughly the
same, made us decide to go for the Nocona's.
BTW, maybe this is known to you already, but Intel offers you their OpenMP
capable compilers for free for academic use.
The machine is up and running now in 64 bit mode and doing fine ;-).
Maybe the information can help you in your decision.
Best regards,
Mark Somers.
Dr. M. F. Somers
Theoretical Chemistry - Leiden Institute of Chemistry - Leiden University
Einsteinweg 55, P.B. 9502, 2300 RA Leiden, The Netherlands
tel: +31715274437
mail: m.somers _(a)_ chem.leidenuniv.nl
web: http://rulgla.leidenuniv.nl/Researchers/Somers.htm
room: HB322
==================================================================================
> From m.swart _(a)_ few.vu.nl Tue May 17 00:05:09 2005
Date: Mon, 09 May 2005 09:55:13 +0200
From: Marcel Swart <m.swart _(a)_ few.vu.nl>
To: Luigi Cavallo <cavallo _(a)_ chemistry.unina.it>
Subject: Re: CCL:Opteron or Nocona ?
[ The following text is in the "WINDOWS-1252" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
Dear Luigi,
probably someone at SCM would be willing to give you some data on how ADF
performs on both machines; all I remember from the bits and bats I've been
told over the past months, is that the Nocona seems to be doing very well
for ADF.
dr. Marcel Swart
Theoretische Chemie
Vrije Universiteit Amsterdam
Faculteit der Exacte Wetenschappen
De Boelelaan 1083
1081 HV Amsterdam
The Netherlands
Tel +31-(0)20-5987619
Fax +31-(0)20-5987629
E-mail m.swart _(a)_ few.vu.nl
Web http://www.few.vu.nl/~swart
^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V^V
==================================================================================
> From m.vanleeuwen _(a)_ clustervision.com Tue May 17 00:05:15 2005
Date: Mon, 09 May 2005 09:59:12 +0200
From: Matthijs van Leeuwen <m.vanleeuwen _(a)_ clustervision.com>
To: 'Gert Kruger' <KRUGER _(a)_ ukzn.ac.za>
Cc: cavallo _(a)_ chemistry.unina.it
Subject: RE: CCL:Opteron or Nocona ?
Dear Gert & Mr Cavallo,
Both the Intel Xeon EM64T (code-name "Nocona") and the AMD Opteron
processor are fine for number
crunching. The Nocona is slightly better at pure number crunching when memory
bandwidth is not so
important. When memory performance is important, the Opteron is the superior CPU
because it has a
superior memory bus.
Both CPUs can be clocked down by the BIOS if they overheat, but this should not
be possible when
they run in a properly cooled room, even when running at maximum load for a long
time.
This recent press release lists some of our installations in 2004:
http://www.clustervision.com/pr_2004_uk.html
These include both Intel Nocona and AMD Opteron CPUs and all these installations
are built do run
all the time and at maximum load.
ClusterVision is a hardware-independent cluster solution expert. We have no
preference for AMD or
Intel CPUs and always try to supply the CPU which is best for your applications.
The servers we use
are usually from Supermicro or Tyan, but we can also use other brands.
Hope this is helpful. Don't hesitate to contact me for further questions.
Kind regards,
Matthijs
__________________________________________________
Dr ir Matthijs van Leeuwen
ClusterVision
Tel NL: +31 23 5677510 / Tel UK: +44 870 080 1980
Mob NL: +31 6 41941992 / Mob UK: +44 7966 012215
Fax NL: +31 23 5677519 / Fax UK: +44 870 132 4332
Email: m.vanleeuwen _(a)_ clustervision.com
Web: http://www.clustervision.com
==================================================================================
> From alex.brown _(a)_ ualberta.ca Tue May 17 00:05:20 2005
Date: Mon, 09 May 2005 12:01:16 -0600
From: Alex Brown <alex.brown _(a)_ ualberta.ca>
To: mark somers <m.somers _(a)_ chem.leidenuniv.nl>
Cc: chemistry _(a)_ ccl.net, Luigi Cavallo <cavallo _(a)_
chemistry.unina.it>
Subject: Re: Intel compilers: was Opteron or Nocona ?
Dear Mark, Luigi, and others,
BTW, maybe this is known to you already, but Intel offers you their OpenMP
capable compilers for free for academic use.
I have seen this stated several times on the list. While I believe that this
used to be the case, Intel has clarified their position on academic use of their
non-commercial compilers. As taken from their FAQ page (http://www.intel.com/software/products/noncom/faq.htm)
Q. I am engaged in academic research. Can I use the non-commercial product?
A. If you, as an individual, are receiving any form of compensation for the
research project (i.e., you receive a salary, or funding, etc.), you do not
qualify for a non-commercial use license. However, you may qualify for the
academic license (Click to academic page)
Note that I do not have any affiliation with Intel nor does it bother me what
individuals choose to do with their copies of the non-commercial compilers.
However, I wanted to clear up what I think is a common misconception about their
compilers.
Cheers,
Alex
Alex Brown
Assistant Professor
Department of Chemistry
University of Alberta
Edmonton, AB
T6G 2G2
http://www.chem.ualberta.ca/~abrown/
==================================================================================
> From mwill _(a)_ penguincomputing.com Tue May 17 00:05:25 2005
Date: Mon, 09 May 2005 13:49:09 -0700
From: Michael Will <mwill _(a)_ penguincomputing.com>
To: cavallo _(a)_ chemistry.unina.it
Subject: compilers for em64t vs. opteron
[ The following text is in the "ISO-8859-1" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
There are actually quite good compilers optimizing for the opteron
out there. One from PGI and one from pathscale, both are commercial
compilers for c++ and fortran90.
We expect the opterons to outperform the em64t xeons across the board,
but offer both to our customers.
Michael
--
Michael Will
Penguin Computing Corp.
Sales Engineer
415-954-2887
415-954-2899 fx
mwill _(a)_ penguincomputing.com
Visit us at the following Linux Shows!
Bio-IT World Conference and Expo '05
Hynes Convention Center, Boston, MA
May 17th-19th, 2005
Booth 201
==================================================================================
> From alex.brown _(a)_ ualberta.ca Tue May 17 00:05:29 2005
Date: Mon, 09 May 2005 12:01:16 -0600
From: Alex Brown <alex.brown _(a)_ ualberta.ca>
To: mark somers <m.somers _(a)_ chem.leidenuniv.nl>
Cc: chemistry _(a)_ ccl.net, Luigi Cavallo <cavallo _(a)_
chemistry.unina.it>
Subject: CCL:Intel compilers: was Opteron or Nocona ?
Dear Mark, Luigi, and others,
BTW, maybe this is known to you already, but Intel offers you their OpenMP
capable compilers for free for academic use.
I have seen this stated several times on the list. While I believe that this
used to be the case, Intel has clarified their position on academic use of their
non-commercial compilers. As taken from their FAQ page (http://www.intel.com/software/products/noncom/faq.htm)
Q. I am engaged in academic research. Can I use the non-commercial product?
A. If you, as an individual, are receiving any form of compensation for the
research project (i.e., you receive a salary, or funding, etc.), you do not
qualify for a non-commercial use license. However, you may qualify for the
academic license (Click to academic page)
Note that I do not have any affiliation with Intel nor does it bother me what
individuals choose to do with their copies of the non-commercial compilers.
However, I wanted to clear up what I think is a common misconception about their
compilers.
Cheers,
Alex
Alex Brown
Assistant Professor
Department of Chemistry
University of Alberta
Edmonton, AB
T6G 2G2
http://www.chem.ualberta.ca/~abrown/
=======================================================================================
> From konstantin_kudin _(a)_ yahoo.com Tue May 17 00:05:32 2005
Date: Mon, 09 May 2005 19:01:04 -0700 (PDT)
From: Konstantin Kudin <konstantin_kudin _(a)_ yahoo.com>
To: mark somers <m.somers _(a)_ chem.leidenuniv.nl>,
Luigi Cavallo <cavallo _(a)_ chemistry.unina.it>
Cc: chemistry _(a)_ ccl.net
Subject: Re: CCL:Opteron or Nocona ?
Actually, highly optimizing compilers for Opteron do exist (such as
Pathscale). Even Intel Fortran 8.1 (either 32 or 64 bits) produces
decent code for Opteron since Opteron is not as fragile as Pentium 4 in
terms of which code it likes and which it does not.
There is one specific situation where Nocona could come as a clear
winner, but such cases are not too common. Basically, codes that run
BLAS routines such as DGEMM for the majority of their runtime gain a
lot on Nocona when the correct libraries are used. This is because both
Opteron and Nocona run BLAS proportionally to the clock speed, so the
much higher clocked Nocona wins (~3 vs ~2 Ghz). On the other hand,
multi-processor Noconas can choke because of limited memory bandwidth,
whereas Opterons under such circumstances do just fine.
Last, here are some last year's Gaussian03 benchmarks, where both
Opterons and Noconas were comparable:
http://www.princeton.edu/~kkudin/g03_b5_tests_1.txt
YMMV ...
Kostya
=======================================================================================
> From reuti _(a)_ staff.uni-marburg.de Tue May 17 00:05:36 2005
Date: Mon, 09 May 2005 23:11:44 +0200
From: Reuti <reuti _(a)_ staff.uni-marburg.de>
To: Eugen Leitl <eugen _(a)_ leitl.org>
Cc: Beowulf _(a)_ beowulf.org
Subject: Re: [Beowulf] CCL:Opteron or Nocona ? (fwd from
m.somers _(a)_ chem.leidenuniv.nl)
Resent-Date: Tue, 10 May 2005 10:29:10 +0200
Resent-From: eugen _(a)_ leitl.org
Resent-To: cavallo _(a)_ chemistry.unina.it
[ The following text is in the "ISO-8859-1" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
Hi,
Quoting Eugen Leitl <eugen _(a)_ leitl.org>:
> ----- Forwarded message from mark somers <m.somers _(a)_
chem.leidenuniv.nl>
> -----
<snip>
> BTW, maybe this is known to you already, but Intel offers you their OpenMP
> capable compilers for free for academic use.
I understand the license terms of Intel a little bit different: a student may
use the compilers for his/her homework for free and also OpenSource
programmers. But as soon as the person using the compiler or the compiled
binary get any payment for their work, you have to buy (an academic) license.
Maybe the license terms changed again, but can please anyone clarify.
=======================================================================================TIA
- Reuti
> From Christian.Boehme _(a)_ gwdg.de Tue May 17 00:05:44 2005
Date: Tue, 10 May 2005 11:41:46 +0200
From: Christian Boehme <Christian.Boehme _(a)_ gwdg.de>
To: Luigi Cavallo <cavallo _(a)_ chemistry.unina.it>
Cc: chemistry _(a)_ ccl.net
Subject: Re: CCL:Opteron or Nocona ?
[ The following text is in the "ISO-8859-1" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
Dear Luigi,
We recently purchased a 32 dual-processor Opteron (2GHz) cluster. We
used CPMD and several artificial benchmarks on Opteron and Xeon (3GHz)
machines to come to this decision. With CPMD we found that the advantage
of the Opteron over the Xeon was very significant for 32-bit memory
intensive applications running on both processors of a dual processor
box (Intel compiler on both architectures). When using just one
processor, the Xeon scored much better. As Intel has not changed the
memory interface of the Nocona, I do not think this trend would be
different for 64-bit applications. Regarding Fortran compilers, from our
experience both the Intel and the Portland can now be used to make
production quality 64-bit binaries for the Opterons. Best wishes
Christian
--
Dr. Christian Boehme
GWDG Private:
Am Fassberg Wilhelm-Raabe-Str. 15
37077 Gvttingen 37083 Gvttingen
email: Christian.Boehme _(a)_ gwdg.de ChristianBoehme _(a)_ web.de
phone: +49 (0)551 201-1839 +49 (0)551 3077000
fax: +49 (0)551 201-2150 +49 (0)551 3077077
=======================================================================================
> From maurice _(a)_ harddata.com Tue May 17 00:05:48 2005
Date: Tue, 10 May 2005 01:44:32 -0600
From: Maurice Hilarius <maurice _(a)_ harddata.com>
To: beowulf _(a)_ beowulf.org
Subject: [Beowulf] Re: CCL:Opteron or Nocona ?
Resent-Date: Tue, 10 May 2005 18:00:15 +0200
Resent-From: eugen _(a)_ leitl.org
Resent-To: cavallo _(a)_ chemistry.unina.it
[ The following text is in the "ISO-8859-1" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
>
>From: mark somers <m.somers _(a)_ chem.leidenuniv.nl>
>Date: Mon, 9 May 2005 09:38:06 +0200
>To: Luigi Cavallo <cavallo _(a)_ chemistry.unina.it>
>Cc: chemistry _(a)_ ccl.net
>Subject: CCL:Opteron or Nocona ?
>
>Dear Luigi Cavallo,
>
>we have recently purchased and setup a new Beowulf cluster here in Leiden,
>dedicated to run ADF and other DFT codes (Dacapo and Vasp) on it. We have
>found the Nocona's, in combination with the Intel compilers, to be the best
>combination especially for ADF and ADF-BAND.
>
>Before we decided to go for the Nocona's, we did try with two test machines,
>one being a dual-cpu AMD Opteron 246 (2.0 GHz) and the other being a
dual-cpu
>Nocona (3.0 GHz). Tests showed that the AMD has, in general, a slightly
>better cache hierarchy and a better memory scalability, but also that the
>lack of compilers able to tune for the Opterons is severe. Of course we took
>the differences in cpu clocks into account.
>
>
>
LACK OF COMPILERS IS SEVERE?
You apparently did next to no research on the subject.
A simple Google search on the terms: "Opteron compiler" provides a
wealth of information.
GCC, Intel, Absoft, and especially Pathscale compilers are excellent for
Opteron.
New Revision E and dual core Opterons have SSE3.
As for your comment about "taking the differences in CPU clocks into
account" I am quite puzzled.
Even Intel is now pushing "performance ratings".
In current CPUs a comparison based on clock speed is nearly meaningless.
>As it turns out, after having contacted the SCM people in Amsterdam, ADF can
>effectively use the SSE3 registers and cpu intruction set and that makes it
>run fast on the Nocona's. This, together with the prices being roughly the
>same, made us decide to go for the Nocona's.
>
>BTW, maybe this is known to you already, but Intel offers you their OpenMP
>capable compilers for free for academic use.
>
>
No, they do not!
They are only "free" for essentially hobbyist use.
If you use them in a university or research cluster that violates the
license.
I think you had a nice relationship with the Intel reseller, and "drank
the Kool-Aid" he gave you.
Sorry to be harsh, but the statements you make above are so one-sided
and uninformed, I had to speak out.
For HPC the numbers are so overwhelmingly in favour of Opteron as to be
virtually insurmountable.
Before I get flamed, yes, I agree there are some codes that are better
suited to the Intel CPUs.
Just not the vast majority.
Further, now that dual core Opterons are shipping, the whole game just
got reset.
Any advice you can give based on a purchasing decision made even 3
months ago is now so obsolete as to be virtually irrelevant.
Here is a recent and fairly comprehensive comparison on the AMD versus
the Intel single and dual core CPUs performance.
http://techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=1
The power consumption and heat production is especially telling.
Add to that the fact that with dual cores the whole architecture just
changed drastically and I think you will see that we now start over with
"new rules".
With our best regards,
Maurice W. Hilarius
Hard Data Ltd
email: maurice _(a)_ harddata.com
http://www.harddata.com/
====================================================================================
> From eugen _(a)_ leitl.org Tue May 17 00:05:55 2005
Date: Tue, 10 May 2005 13:02:46 +0200
From: Eugen Leitl <eugen _(a)_ leitl.org>
To: chemistry _(a)_ ccl.net
Subject: CCL:[Beowulf] Opteron or Nocona ? (fwd from
m.somers _(a)_ chem.leidenuniv.nl) (fwd from lindahl _(a)_ pathscale.com)
----- Forwarded message from Greg Lindahl <lindahl _(a)_ pathscale.com>
-----
From: Greg Lindahl <lindahl _(a)_ pathscale.com>
Date: Tue, 10 May 2005 00:03:11 -0700
To: Beowulf _(a)_ beowulf.org
Cc:
Subject: Re: [Beowulf] CCL:Opteron or Nocona ? (fwd from
m.somers _(a)_ chem.leidenuniv.nl)
User-Agent: Mutt/1.4.1i
On Mon, May 09, 2005 at 08:05:28PM -0400, Mark Hahn wrote:
> I would very much like to see a good comparison of Intel compilers
> versus the alternatives. I know good things about Pathscale, for instance,
> and gcc 4.0 and 4.1 seem pretty impressive.
In addition to the usual benchmarks like SPECfp and Polyhedron, there
is a computational chemistry comparison done by Daresbury Labs in the
UK. Opteron+PathScale does quite well.
http://www.polyhedron.co.uk/compare/linux/f77bench_AMD.html
http://www.polyhedron.co.uk/compare/linux/f90bench_AMD.html
http://www.cse.clrc.ac.uk/disco/index.shtml, especially
http://www.cse.clrc.ac.uk/disco/Benchmarks/IntelProj.Serial.update1.pdf
We're looking forward to Daresbury testing with our newest compilers,
as we've made some substantial performance improvements since the
version they used (1.2).
-- greg
_______________________________________________
Beowulf mailing list, Beowulf _(a)_ beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
----- End forwarded message -----
--
Eugen* Leitl <a href="http://leitl.org";>leitl</a>
______________________________________________________________
ICBM: 48.07078, 11.61144 http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
http://moleculardevices.org http://nanomachines.net
-= This is automatically added to each message by the mailing script =-
To send e-mail to subscribers of CCL put the string CCL: on your Subject: line
and send your message to: CHEMISTRY _(a)_ ccl.net
Send your subscription/unsubscription requests to: CHEMISTRY-REQUEST _(a)_
ccl.net
HOME Page: http://www.ccl.net
| Jobs Page: http://www.ccl.net/jobs
If your is mail bouncing from ccl.net domain due to spam filters, please
use the Web based form from CCL Home Page
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
==================================================================================
> From konstantin_kudin _(a)_ yahoo.com Tue May 17 00:06:00 2005
Date: Mon, 09 May 2005 19:01:04 -0700 (PDT)
From: Konstantin Kudin <konstantin_kudin _(a)_ yahoo.com>
To: mark somers <m.somers _(a)_ chem.leidenuniv.nl>,
Luigi Cavallo <cavallo _(a)_ chemistry.unina.it>
Cc: chemistry _(a)_ ccl.net
Subject: CCL:Opteron or Nocona ?
Actually, highly optimizing compilers for Opteron do exist (such as
Pathscale). Even Intel Fortran 8.1 (either 32 or 64 bits) produces
decent code for Opteron since Opteron is not as fragile as Pentium 4 in
terms of which code it likes and which it does not.
There is one specific situation where Nocona could come as a clear
winner, but such cases are not too common. Basically, codes that run
BLAS routines such as DGEMM for the majority of their runtime gain a
lot on Nocona when the correct libraries are used. This is because both
Opteron and Nocona run BLAS proportionally to the clock speed, so the
much higher clocked Nocona wins (~3 vs ~2 Ghz). On the other hand,
multi-processor Noconas can choke because of limited memory bandwidth,
whereas Opterons under such circumstances do just fine.
Last, here are some last year's Gaussian03 benchmarks, where both
Opterons and Noconas were comparable:
http://www.princeton.edu/~kkudin/g03_b5_tests_1.txt
YMMV ...
Kostya
==================================================================================
> From gustavo.seabra _(a)_ gmail.com Tue May 17 00:06:07 2005
Date: Wed, 11 May 2005 14:31:06 -0500
From: Gustavo Seabra <gustavo.seabra _(a)_ gmail.com>
To: chemistry _(a)_ ccl.net
Subject: CCL:Opteron or Nocona ?
[ The following text is in the "ISO-8859-1" character set. ]
[ Your display is set for the "US-ASCII" character set. ]
[ Some characters may be displayed incorrectly. ]
Notice that AMD has recently launched the Dual Core Opteron
processors. There you can basically get the equivalent of twice as
much processors, for about the same price. Check:
http://www.infoworld.com/infoworld/article/05/04/21/17TCamd_1.html
http://www.d-silence.com/story.php?headline_id=21036&comment=1
--
---------------------------------------------------------------------------------------------
Gustavo Seabra Graduate Student
Chemistry Dept. Kansas State University
Registered Linux user number 381680
Say NO! to software patents: http://www.nosoftwarepatents.com
---------------------------------------------------------------------------------------------
==================================================================================
> From eugen _(a)_ leitl.org Tue May 17 00:06:15 2005
Date: Thu, 12 May 2005 10:41:00 +0200
From: Eugen Leitl <eugen _(a)_ leitl.org>
To: chemistry _(a)_ ccl.net
Subject: CCL:[Beowulf] Opteron or Nocona ? (fwd from
m.somers _(a)_ chem.leidenuniv.nl) (fwd from kus _(a)_ free.net)
Sorry for big delay w/answer: today is 1st working day after May's
holidays here in Russia ;-)
I've no doubts that in average Opteron is today better, in particular
for computational chemistry codes. The reasons is both performance
(for
"top frequency" chips) and price/performance ratio. We stop to use
Intel x86 for computational chemistry and began to use AMD from
"Athlon
times" - because of price/performance reason. But now Opteron is in
average absolutely better due to
a) single CPU performance (See, for example, SPECfp2000 which includes
also some molecular dynamics codes)
b) power dissipation (which is lower for AMD)
c) dual (and higher also:-)) CPU system architecture: 2 Opteron's
have 2 independed pathes to main memory instead of common system bus
in the case of Nocona. As a result, Opteron's have much more good
memory throughput scalability.
Pls take into account that many modern quantum chemistry codes (like
G03) are "restricted" by memory throughput !
What is about compilers, it was answered in a set of previous messages
here.
IMHO, the only HPC area where Nocona has clear advantage over Opteron,
is floating point applications reaching performance close to peak
value (like Linpack). In that case it should be take into account that
Peak_Performance = 2*frequency for both processors, and Nocona
has therefore higher values. But it's not typically for computational
chemistry.
Yours
Mikhail Kuzminsky
Zelinsky Institute of Organic Chemistry
Moscow
==================================================================================