Re: [AMBER] CUDA problems in Amber18

From: Ross Walker <>
Date: Mon, 19 Nov 2018 21:14:40 -0500

Hi Julio,

First question did you run the test cases after building AMBER 18 and did those pass? If the simulation runs fine with AMBER 16 but gives incorrect results with AMBER 18 that suggests an issue with the AMBER 18 compilation. Can you potentially start again from a clean AMBER 18 untarred from the original tar files. Be sure to apply all the patches before compiling. Then after building the CUDA version of the code please run the test cases and see if you see any major differences.

All the best

> On Nov 18, 2018, at 22:33, Julio Dominguez <> wrote:
> We recently obtained our Amber18 license and while testing the new installation I ran into a few problems. Well, just one, really.
> First of all, my system configuration:
> Cuda compilation tools, release 9.0, V9.0.176
> Driver Version: 384.81
> Centos 6.4
> gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-23)
> Amber18/Ambertool18 up-to date
> When running a normal simulation, the temperature equilibration step has some weird increases in potential and total energy. These raises result in a
> distorted protein structure. Both energies return to "normal" (as compared to an identical simulation in Amber 16) during the pressure equilibration.
> Previous steeps, energy minimization with position restraints, seem normal. I have ran the same simulation in the same hardware and same compilers using CUDA but with Amber16 but no energy raise and with "normal" protein structures.
> Here is my input for equilibration, in dropbox:
> Can somebody see any mistake I made and cannot identify myself?
> Thank you and best regards.
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Mon Nov 19 2018 - 18:30:01 PST
Custom Search