Re: [AMBER] Problem about one GPU & multiple CPUs

From: Daniel Roe <>
Date: Wed, 21 Dec 2016 09:56:39 -0500


pmemd.cuda is not currently written to take advantage of multiple CPU
cores with a single GPU (i.e. OpenMP + CUDA). Instead, all of the
computationally intensive work is done on the GPU, while a single CPU
thread handles things like IO etc. If you want to use the remaining
CPU threads you could run an analysis job at the same time with e.g.


On Wed, Dec 21, 2016 at 8:38 AM, 吴萌 <> wrote:
> Hi,
> When I used "pmemd.cuda" to response my work, I found that the work just run with one GPU and one CPU, but there are multiple CPUs available. I want to run under one GPU with multiple CPUs, because maybe it will run more efficient. I first tried "mpirun -np xx(CPU numbers) pmemd.cuda -...", the work failed, then I found "pmemd.cuda.MPI" in the manual, but it works with multiple GPUs. I originally consider that "pmemd.cuda" will automaticly call multiple CPUs, but the result is not. Maybe some solutions about this problem I have missed, this problem is still on at present.
> I am looking forward to receiving your reply. Thanks!
> Sincerely,
> Meng,Wu
> _______________________________________________
> AMBER mailing list

Daniel R. Roe
Laboratory of Computational Biology
National Institutes of Health, NHLBI
5635 Fishers Ln, Rm T900
Rockville MD, 20852
AMBER mailing list
Received on Wed Dec 21 2016 - 07:00:03 PST
Custom Search