Re: NAMD Simulation in a laptop

From: Vermaas, Joshua (Joshua.Vermaas_at_nrel.gov)
Date: Mon Mar 26 2018 - 13:42:06 CDT

Yeah. Once the system is solvated, you are probably looking at something in the neighborhood of 50,000 atoms. If you only have 4 cores (the AMD GPU won't help you), this is going to run really slowly. Typically, you'll want 1 core to handle about 500-1000 atoms for a reasonable tradeoff between efficiency and performance. If you absolutely MUST run this on your laptop, I'd go the implicit solvent route, since most of your atoms right now are water. I've seen moving towards implicit solvent cause protein conformational changes, so harmonic restraints might be required, but you'll get much better performance if you reduce your atomcount (or add an CUDA-capable GPU).

-Josh

On 03/25/2018 10:45 PM, Richard Overstreet wrote:

Denish,

I would submit a proposal to a supercomputer center for an allocation of CPU/GPU walltime. In some cases an initial allocation is fairly easy to get. Laptops aren't really suited to running heavy computational jobs. As axel said there is always a choice.

Regards,

Richard

On 03/25/2018 10:37 PM, Denish Poudyal wrote:

On Mon, Mar 26, 2018 at 7:52 AM, Axel Kohlmeyer <akohlmey_at_gmail.com<mailto:akohlmey_at_gmail.com>> wrote:
On Sun, Mar 25, 2018 at 9:56 PM, Denish Poudyal <qrystal45_at_gmail.com<mailto:qrystal45_at_gmail.com>> wrote:
> i thought I could run a 100 ns NAMD simulation in my laptop for a
> protein-hormone system containing around 2000 atoms in its pdb
> file(4lnx.pdb). But even to run a simulation for four hundred fifty thousand
> steps, it has taken almost 4700 minutes. is this normal or its low?

nobody can give meaningful advice on such vague information. you're
not saying whether you are using implicit or explicit solvent, what
force field, what MD settings, how you run NAMD, which binary and
version you are using and so on.

NAMD prints out some benchmark information during the run, which can
help to make an estimate.

> the device has 4 cores with 16 gigs of ram and 4 gigs of gpu(amd). Or any
> suggestions to run it faster???And I still am aware that using laptop for
> such work isnt advisable but I dont have a choice.

there *always* is a choice.

axel.

>
> Denish Poudyal
> CDPTU, Nepal
>
>

--
Dr. Axel Kohlmeyer  akohlmey_at_gmail.com<mailto:akohlmey_at_gmail.com>  http://goo.gl/1wk06m1QDaNFxLN79FIRimawiMZd0E%3D&reserved=0>
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.
Water is used as solvent explicitly with NaCl of 0.15 mol/l used to ionize the system. And the simulation is done at 2 fs/step with spherical boundary conditions. CHARMM Force fields are being used. As I'm very much new to this method, please suggest how can I conduct this work smoothly.
--
Denish Poudyal
CDPTU, Nepal

This archive was generated by hypermail 2.1.6 : Wed Dec 04 2019 - 23:19:44 CST