Re: [AMBER] GPU benchmark - Amber18

From: David A Case <>
Date: Wed, 2 May 2018 08:18:39 -0400

On Wed, May 02, 2018, Hadházi Ádám wrote:
> - Can you tell me if there is any new paper regarding the performance of
> Amber18 on MD and TI/FEP?

Sigh...this information is indeed very hard to locate. Both the 2018
Reference Manual and the web page are quite out of date regarding
pmemd.cuda. I'm cc-ing this to people who put lots of effort into
adding free energy support and other updates into Amber18, in the hopes
of spurring some efforts to make this work visible to users.

Basically, the MD performance of Amber18 on GPUs is pretty much on a par
with the Amber16 benchmarks that are on the web. I know that Ross
Walker is in the process of creating an updated set of benchmarks, and
we hope/expect that a future update patch will further boost performance.

With regards to TI/FEP, the principal citations are in
$AMBERHOME/src/pmemd/src/cuda/cuda_info.F90, and are printed to the
output if you run a free energy calculation:

Daniel J. Mermelstein; Charles Lin; Gard Nelson; Rachael Kretsch; '
J. Andrew McCammon; Ross C. Walker "Fast and Flexible GPU Accelerated '
Binding Free Energy Calculations within the AMBER Molecular Dynamics '
Package" J. Comp. Chem., 2018, DOI: 10.1002/jcc.25187 '

Tai-Sung Lee; Yuan Hu; Brad Sherborne; Zhuyan Guo; Darrin M. York '
"Toward Fast and Accurate Binding Affinity Prediction with pmemdGTI: An '
Efficient Implementation of GPU-Accelerated Thermodynamic Integration" '
J. Chem. Theory Comput., 2017, 13 (7), 3077 '

...hope this helps....dac

AMBER mailing list
Received on Wed May 02 2018 - 05:30:02 PDT
Custom Search