Re: NAMD on Shared Memory (Altix UV 1000)

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Thu Dec 02 2010 - 21:24:44 CST

On Thu, Dec 2, 2010 at 7:11 PM, Alessandro Cembran
<a.cembran.ml_at_gmail.com> wrote:
> Hi,
>
> I was wondering if NAMD would take any significant advantage of a large
> shared memory system (SGI Altix UV 1000 with about 1,000 cores), or if the
> way that the job is sliced up and the messages passed would give such a
> machine little or no edge over a standard cluster with fast interconnects.

altix machines do make great message passing machines,
but the price tag is pretty high, too. i would expect that kind
of machine to outrun a standard cluster with fast interconnect,
but not be able to compete on the price/performance ratio.

in short, it depends on how much you want to pay for better
capability compared to more capacity. i would probably go
for the latter and consider spicing up the mix with a healthy
dose of GPUs.

cheers,
     axel.
>
> Thanks,
>
> Alessandro
>

-- 
Dr. Axel Kohlmeyer
akohlmey_at_gmail.com  http://goo.gl/1wk0
Institute for Computational Molecular Science
Temple University, Philadelphia PA, USA.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 05:23:27 CST