On Mon, 28 Sep 2009, Sean O'Riordain wrote:
Good morning Keith,
Have a look at
http://cran.r-project.org/bin/windows/base/rw-FAQ.html#There-seems-to-be-a-limit-on-the-memory-it-uses_0021
The short answer is that "it depends"...
a) memory is limited under windows
Yes, but 64-bit builds can be used on Windows -- that needs commercial
compilers and there are commercial vendors of such builds.
Even with the CRAN binary, a 64-bit version of Windows offers double
the memory over a (vanilla) 32-bit version.
b) R is essentially a serial program - HOWEVER it depends what you're
actually doing - if you're working with large matrices then there are
parallel versions of BLAS that can be used... On a multi-core windows
machine with lots of memory you can of course run up multiple copies of R
and run each independently
There are several packages that parallelize their computations with
MPI etc, and others that help with parallelization (papply, foreach,
gputools, ....). And apart from Rmpi/rpvm/snow there is also
'multicore', but not on Windows. See the R-sig-hpc list for follow up
on such issues.
As for Vista vs Windows 7, this is not the right list but Windows 7
behaves just like a version of Vista as far as we have explored it
(and the current rw-FAQ includes it and Server 2008 in the Vista
section).
Many of us have bought dual quad-core servers in the last year or so:
that includes Uwe Ligges' winbuilder machine. I suspect most of the
usage is separate R jobs running simultaneously: certainly that is the
case in my dept (where there are at least 6 8-core servers running R
jobs).
Kind regards,
Sean
On Mon, Sep 28, 2009 at 4:40 AM, Keith Satterley <ke...@wehi.edu.au> wrote:
I've read some postings back in 2002/2006 about running R on multiple core
CPUs. The answer was basically separate processes work fine, but
parallelization needs to be implemented using snow/rmpi. Are the answers
still the same?
I ask because we are about to order a laptop running Windows for a new
staff member. Some advice on the following would be helpful.
It will be ordered with Vista, with a free upgrade to Windows 7. It will
have 8GB of memory
A quad core CPU costs about AUD$1100 more than the fastest (Intel T9900-6M
Cache, 3.06 GHz) dual core CPU.
I'm wondering if there is value in ordering the quad core. We are looking
at a time frame of 3-4 years.
Is anyone aware of near future plans to implement some form or
parallelization that would more or less be hidden from the normal user?
It is anticipated that analysis of Next Gen sequence data will be
important.
I've read the Windows FAQ about running R under Vista. We will probably
start with Vista. I've read some posts in R-devel indicating people are
running R under Windows 7. Is it safe to assume that R will run under
Windows 7 after it is released?
We are hoping to make use the 8GB of memory. Am I right in assuming that
when the 64 bit version of Windows 7 is available, it will allow R users to
make good use of the 8GB of memory. Does this happen under the current
higher end versions of 64 bit Vista?
cheers,
Keith
========================
Keith Satterley
Bioinformatics Division
The Walter and Eliza Hall Institute of Medical Research
Parkville, Melbourne,
Victoria, Australia
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
[[alternative HTML version deleted]]
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Brian D. Ripley, rip...@stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
