ubuntu desktop 12.04 lts on nehalem workstation

[Liam Proven]

On 14 August 2012 17:05, Roger Miller <abq at lwasf.com> wrote:
>
> I recently installed ubuntu desktop 12.04 LTS on a workstation equipped
> with 12 GB of RAM and 2 Xeon 5600-series processors, each with 6 cores.
> This may not be the kind of equipment that the software was designed
> for, but I was in a pinch for time.

Sounds like a great box. I suppose you could run OpenSUSE to soak up
some more of those spare resources ;¬) but otherwise I see no problem.

> I am doing groundwater simulations using a widely used and very stable
> FORTRAN code (MODFLOW-2000).  The code compiles with or without MPI
> support (using either openMPI or mpich2) and reports no problems.  The
> simulations fail, reporting impossible problems that suggest to me that
> the memory is being accessed incorrectly.
>
> The same code compiled on an older computer runs on the workstation
> without problems as long as I am not using MPI.  When I try to run the
> simulation on more than one core simultaneously I get the same problems
> reported in the previous paragraph.
>
> I have not found any reports of problems with Linux (ubuntu or other)
> running on Nehalem workstations.  Are there known problems that I have
> overlooked?  Is there anything I can do to get a clearer picture of the
> problem?

I have to ask the silly, obvious question:

Are you using the 64-bit or 32-bit edition?

Both will recognise the memory, but I would anticipate problems with
the 32-bit system trying to manipulate large datasets. The 32-bit
version uses PAE to access > ~3½ gig of RAM and this works by paging
small chunks of the RAM above 4GB into address space under 3½GB. That
will cause problems for you, I think.


-- 
Liam Proven • Profile: http://lproven.livejournal.com/profile
Email: lproven at cix.co.uk • GMail/G+/Twitter/Flickr/Facebook: lproven
MSN: lproven at hotmail.com • Skype/AIM/Yahoo/LinkedIn: liamproven
Tel: +44 20-8685-0498 • Cell: +44 7939-087884