You need a good MD program and a shell script to modify the parameters to
files like ${AMBERHOME}/dat/leap/parm/frcmod.tip5p and
${AMBERHOME}/dat/leap/lib/solvents.lib. Then, apply whatever parameter
scanning strategy you wish. Use bigger water boxes to test the properties,
if they converge better with larger system sizes. You will not be able to
run a 2,000 atom system much faster than a 20,000 atom system, but if you
want to compromise and do 12,000 atom systems you can turn on the CUDA
Multi-process service and get an extra 80% throughput out of the newer GPUs
with 2-4 simulations running per card (as opposed to just one). With very
small systems you can get up to 140% additional throughput via MPS, but
again if more water in your simulation will help statistics as much as
longer timescales, then go with a bigger system size.
To start MPS:
#!/bin/bash
set -e
export CUDA_MPS_PIPE_DIRECTORY=/tmp/nvidia-mps
export CUDA_MPS_LOG_DIRECTORY=/tmp/nvidia-log
nvidia-cuda-mps-control -d
To stop MPS:
#!/bin/bash
echo quit | nvidia-cuda-mps-control
Dave
On Mon, Apr 8, 2019 at 2:56 PM Gustavo Seabra <gustavo.seabra.gmail.com>
wrote:
> Hi all,
>
> I want to create a TIP5P water molecule from scratch. Does anyone know of
> a good resource I can use, like a step-by-step procedure, tutorial, etc?
>
> Thanks,
> Gustavo.
>
> Sent from Mail for Windows 10
>
> _______________________________________________
> 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 Mon Apr 08 2019 - 13:00:02 PDT
