[gmx-users] 5 identical simulations --> 5 different results, why?

[Dmitri Dubov]

Hi, Justin.

You wrote:

> Ricardo O. S. Soares wrote:
>> Hello dear users,

>> in order to introduce Gromacs to a few students, I'm taking five 
>> computers with the same operational system (ubuntu), to run the same 
>> simulation independently. I prepared a box o water and ran all standard 
>> steps until the grompp. This is when I copied and pasted the folder to 
>> the other four computers. So I ran the same grompp command in each unit, 
>> and then the mdrun step. At the end we obtained 5 slightly different 
>> results for total energy, pressure, volume etc. Since MD is a 
>> deterministic process, one may expect to find the same results for the 
>> same simulations. So does anyone have a clue at which step is it 
>> possible that those little differences are being brought into the 
>> systems? I'm guessing the grompp step...

> This page may have some useful information for you:

> http://www.gromacs.org/Documentation/Terminology/Reproducibility

It says that

"As of GROMACS 4, mdrun -reprod will eliminate all sources of
non-reproducibility that it can, i.e. same executable + same shared
libraries + same run input file + same command line parameters will lead to reproducible results."

In addition:
The last condition should be applied not only to gmx parameters themselves. For instance, paralleling the same run between different numbers of my Intel Quad cores gives me slightly different energies too. For small systems about hundreds of atoms this is the case even with -reprod and under relatively long, 10 ns, simulations. 

-- 
Regards,
 Dmitri                          mailto:ddubov at ngs.ru
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20100617/55c4f04a/attachment.html>