On Wed, Sep 22, 2010, colvin wrote:
>
> I am running MD on a system consisting 112550 atoms using pmemd on a
> quad-core workstation and it tooks about 17 hrs to complete 500ps.
>
> Then, i run the same system on 2 quad-core connected with Gigabit
> ethernet, and it took about 13 hrs.
This is expected, and there is not much you can do about it. Gigabit ethernet
is nearly useless as an interconnect for Amber MD simulations--it simply
cannot keep up with the communication demands our codes require. To go "off
node", you need an Infiniband-level interconnect. Alternatives are to gain
access to machines with more cores per node or to consider GPU's; note that
the latter are still "bleeding-edge", and many aspects of functionality are
not supported.
(Other codes, such as desmond, might scale better on low-end hardware, but I
don't have any experience with that.)
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Sep 22 2010 - 05:30:04 PDT
