[gmx-users] diff behaviour from same .tpr file under v3.3 /3.3.2

[Mark Abraham]

Paul Rowntree wrote:
> Greetings All;
> I have reinstalled my Rocks cluster which went from CentOs4 to CentOs5 
> .  I downloaded the Gromacs 3.3.2 RPMs, and although the examples work 
> fine, my own simulations (surface tethered species, OPLS-AA) are blowing 
> up, even though they are minimized as usual.  I see no differences in 
> the topologies, but the blow-up clearly starts near the tether points.  
> I went back to a tpr file that worked well previously with v3.3, but 
> even it diverges with the LINCs constraint failures after 10-50 ps, 
> depending on the run.
> 
> If all of the run information is in the tpr file, does this mean that 
> the mdrun codes are significantly different?

It looks like it. You seem to be implying that your new 3.3.2 mdrun 
blows up run input files that worked with 3.3 mdrun. I've no idea what 
might have changed in the code to cause this, if this really is what is 
happening.

> Should I expect to have similar levels of stability under different 
> versions of gromacs?

Yes, unless you're using something that was improved or bug-fixed or 
(heaven forbid) bug-introduced. There's revision histories on the 
GROMACS website you might like to check against the details of your 
simulations.

> Are Gromacs v4 RPMs available?  I did not see them on the website ...

Yes, but they're not provided by the GROMACS team. Google finds them.

Mark