Hi Mayank,
yes, I'm doing that with a Sun Grid Engine scheduler. The only
difference is in the script that you submit to the nodes - it has to
contain a reference to pmemd.cuda executable and pmemd.cuda has to be
visible to all nodes (placed on an shared hard drive). Also, each GPU
node needs to have the appropriate Nvidia driver installed and to have
access to the CUDA runtime (preferably, also located on a shared drive).
Finally, you need to add a queue at the SGE master host that includes
the nodes with GPUs and specify 1 as the number of slots on each node.
You can submit jobs individually or as an array job (for a number of
similar simulations).
Hope it helps
Sasha
Mayank Daga wrote:
> Hi,
>
> I would like to know whether it is possible to run AMBER on a cluster where
> each node has one GPU. If yes, what are the things that I have to
> differently?
>
> Thanks,
> ~mayank
>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Aug 24 2010 - 10:30:05 PDT
