[gmx-users] GPU in Ubuntu

Szilárd Páll pall.szilard at gmail.com
Sun May 11 02:03:14 CEST 2014


Hi,

First of all, the GeForce 610 is a rather basic card, in fact the most
basic among the low-end ones. Unless the computer it is connected to
is *very* slow (e.g. something like a dual-core CPU from 5+ years
ago), my guess is that using it is not going to improve performance of
a GROMACS simulation.

If you still want to try to get it to work, I suggest cleaning out all
NVIDIA driver installations and doing a fresh installation. After
that, even without CUDA, the GPU should work; e.g. the nvidia-smi
command line tool should list it, it should also be useable to drive a
display, etc. If that all works, there is a good chance that
computational codes will work on it too.


Secondly, I am not aware of any free computational resources that can
be used with GROMACS and I have a feeling that you may not find any
offers for substantial amount of free resources. You could consider
cloud-based resources, e.g. you can get a GPU equipped node from
amazon for $0.0656/h which is about $1.6 per day.

--
Szilárd


On Sat, May 10, 2014 at 10:06 PM, lloyd riggs <lloyd.riggs at gmx.ch> wrote:
>
> Dear All,
>
> I recently ran into a problem with a small geforce 610, and aside from
> woundering if it was worth the effort, found in Ubuntu 13.10 gromacs can not
> find the GPU.  The Nvidia site found this problem and states to install
> libcudart.5.5.so or higher, i tried this and the 6.0 as well, with no luck.
> The OS itself also has a hard time finding the GPU, however it is found with
> any comand line tool rather easy as a PCI device.  Any fixes, thoughts or
> suggestions as this is simply for set up for systems (EM, initial system
> setup).
>
> A secound question, does anyone know of server based gromacs, as I am at the
> moment hitting a wall with the universities resources (lack of access for
> wierd reasons) and wanted to persue simulations of a few things.  I am aware
> of Groningens site, but have not found much for gromacs any place else,
> although there are several others with different formating, resulting in
> users are only able to use their (the sites) online analytical tools which
> are usually very limited in scope?  With the Groningen site, the only
> complaint is there is no fine tuning of varied parmeters, which means the
> submission works or does not (ie no setting of mdrun parameters such as
> -rcon, -dd, -mpi , only a .tpr and or pdb  file are submitted to the system)
> It is better than nothing, but just a critique, as it worked for me for
> standard systems, but crashed for me with wierd ligands when I used it
> breifly.
>
> Sincerely,
>
> Stephan Watkins, PhD
>
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a
> mail to gmx-users-request at gromacs.org.
>


More information about the gromacs.org_gmx-users mailing list