Hi Dave , Ross,
Thanks for the quick reply, will get back to you after I do the debugging
as suggested by you.
Best,
Senthil
On Thu, Jan 22, 2015 at 10:34 AM, Ross Walker <ross.rosswalker.co.uk> wrote:
> Hi Senthil,
>
> It could be either but likely it is a problem with your simulation input -
> bad starting structure etc. As Dave suggests I would do debugging here. Use
> just pmemd.cuda - see if that works - if it runs on one card and not the
> other that would tell you it is a hardware issue - if it fails on both
> cards then likely it is a problem with your input. Maybe you have atoms
> clashing with each other, a sudden decrease in box size, etc etc.
>
> Try running some short runs with ntpr=1 and ntwx=1 and see what happens -
> also try using the CPU code since this will give you more detailed error
> messages.
>
> All the best
> Ross
>
> > On Jan 21, 2015, at 9:27 AM, BERGY <nucleic81.gmail.com> wrote:
> >
> > Hi
> > I am running amber14 in a machine that came with preinstalled software of
> > cuda and amber14.
> > I get the following error during my runs.
> >
> > cudaMemcpy GpuBuffer::Download failed unspecified launch failure
> >
> > I have 2 x NVIDIA GeForce GTX 780cards
> > Is it a problem with the card or I a error in my script.
> >
> > I submit as.....
> > mpirun -np 2 pmemd.cuda.MPI -O -i prod.in
> >
> > Best,
> > Senthil
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jan 22 2015 - 10:00:04 PST
