On Sun, Jan 03, 2021, Matthew Guberman-Pfeffer wrote:
>
>However, when running the NPT
>equilibration on GPUs, I got an error message that the cutoff was too
>small, that using GPUs was unsafe, and that this issue would be corrected
>in a future release. I tried increasing the cutoff from 10 to 20 Å,
There is no direct fix to this problem, and it is unlikely (in spite of the
error message) that anything will change soon. You have a couple of
options:
1. You could try to slightly *decrease* the cutoff (say to 9 or even 8 Å).
It is the ratio of the cell size to the cutoff that is important here.
The default cutoff is 8 Å, and that value is OK for many protein crystal
simulations.
[Aside: could you quote the exact error message, so that I could look for it
in the code? I don't think(?) it should be saying that the cutoff is too
small, but maybe I don't know exactly what error message you actually
triggered.]
2. If you are (understanably) leery about ignoring some Lennard-Jones
interactions beyond a smaller cutoff, you need to simulate a bigger system.
The usual procedure is to create a supercell consisting of several unit
cells. The advantage here is that you are now simulating N proteins
simultaneously, and so get better statistics and more sampling. The
physical realism is also greater: on average, the structure of the protein
is the same in each unit cell, but at any given point in time, each unit
cell will have its own conformation.
[Substitute the word "protein" above as appropriate.]
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jan 04 2021 - 05:30:02 PST
