[gmx-users] a question about energygrps

[Justin A. Lemkul]

On 4/25/12 10:49 AM, Albert wrote:
> hello Justin:
>
> thank you very much for your such kind explanations. It is quite helpful.
>
> I am wondering how many CPU did you usually use for the membrane simulations?
> and how many ns/day can you get the best if CPU is not a problem with Gromacs? I
> am very confused about this question since some one claimed that they can use
> optimized PMF/F paramters to achieve even 100ns/day for 20,000 atoms system. I
> also try to use g_tune_pme to get a better performance, but the results are not
> so satisfied...... Can you give me some advices for this?
>

I have never been able to get anywhere close to such performance for explicit 
solvent calculations using PME.  Maybe others who have can describe their 
experience.  I don't have an exact benchmark for the tutorial runs, but 24 CPU's 
gave somewhere in the ballpark of 10-15 ns/day on our (relatively old) hardware.

My rule of thumb is 1000 atoms/CPU, and then tweak from there if necessary.

-Justin

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================