[MITgcm-support] new cluster advice
Klymak Jody
jklymak at uvic.ca
Wed Mar 4 18:51:24 EST 2009
Hi,
On Feb 15, 2009, at 23:52 PM, Martin Losch wrote:
> did you get an answer to your question? I am no specialist, but I
> have had very bad experience with quad-core proocessors, because of
> memory access bandwidth (see a thread in October 2008). Apparently,
> the MITgcm (as most other CFD/GFD codes) have a low "computational
> efficiency" (FLOPS per memory access), so that the memory access of
> the cores must be fast, and in quad cores the core appear to compete
> for memory access causing a drastic performance drop. So my advice
> is to advoid quad-cores (or at least not to expect 4times the
> performance with quadcores).
Apple just released their new machines using Intel's "Nehalem" quad
cores, with DDR3 memory, they claim 2.4x the memory bandwidth of their
previous dual 4-core machines. Will that help the bandwidth issue?
They also allow hyperthreading, which allows two threads to run
simultaneously on one core. Does that mean each core can be set up
with 4 processors, 2 threads per processor in the MITgcm?
I need to set up a small cluster (~$20k), and am somewhat ignorant
about how to translate the vendor literature into how it will affect
my workflow with the MITgcm.
Thanks, Jody
--
Jody Klymak
http://web.uvic.ca/~jklymak/
More information about the MITgcm-support
mailing list