[AMBER] constant pH and AMD integration

From: Anders Frederiksen via AMBER <amber.ambermd.org>
Date: Wed, 26 Nov 2025 12:05:01 +0000

Dear Amber developers and other clever people,


I have over the last months performed some replica exchange continuous constant pH simulations using all-atoms on a cluster with Nvidia GPUs with no problem.


Recently I wanted to swap to a cluster with AMD GPUs to perform some non-replica exchange simulations and thought this would pose no issue given the recent AMD GPU implementation of Amber, however upon doing a test simulation of a small system with a non-standard chromophore (used in the replica exchange simulations and confirmed to switch between protonation states prior to those simulations).


Unfortunately this chromophore does not switch protonation states on an AMD GPU architecture and I was thus wondering whether there is a bug/missed integration of this type of simulation with the AMD GPU architecture update, since it seems to behave exactly like as if the simulation was carried out on a CPU node (which is known to not work).


Hope to hear from you.

Yours sincerely,

Anders Frederiksen
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Nov 26 2025 - 04:30:02 PST
Custom Search