Hi Tyler,
Yes it's tough to get really good energy conservation with Shake, including cases where the total energy actually decreases with time. It's the one thing that is still super sensitive to precision. However if you turn shake off and use a 0.5fs or better 0.1fs time step with a tight dsum_tol as you see you should be able to get energy conservation that is reasonable.
All the best
Ross
> On Apr 10, 2017, at 4:54 PM, Tyler Luchko (Lists) <tluchko.lists.gmail.com> wrote:
>
> Thanks George,
>
> I tried your suggestion and saw considerable improvement on the GPUs. So, over the weekend I tried a few variations. FWIW, smaller values of SHAKE tolerance had worse energy conservation. Fortunately, I was able to get acceptable energy conservation with
>
> cut=9
> dsum_tol=1e-7
> dt=0.0005
>
> Heating is about 0.1 K over 50 ns, at least for this test system.
>
> Thanks,
>
> Tyler
>
>> On Apr 4, 2017, at 4:36 PM, George M Giambasu <giambasu.gmail.com> wrote:
>>
>> Give this a try - it worked for me with both cpu & gpu versions of pmemd
>> with NVE simulations. Increasing the shake tolerance can help with
>> energy conservation.
>>
>>
>> MD NVE
>> &cntrl
>> imin = 0, irest = 1, ntx = 7,
>> cut = 9.0,
>> ntc = 2, ntf = 2, tol = 0.0000005
>> nstlim = 7500000, dt = 0.001,
>> ntt = 0, ntb = 0, ntb =1 ,
>> ntpr = 2000,
>> ntwx = 2000,
>> ntwr = 2000,
>> ioutfm = 1,
>> iwrap = 1
>> /
>> &ewald
>> dsum_tol = 0.000001
>> /
>>
>> -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> Dr. George M. Giambasu
>> Rutgers, The State University of New Jersey
>> 174 Frelinghuysen Road, Room 308E Piscataway,
>> NJ 08854, USA giambasu.rutgers.edu
>> www.rci.rutgers.edu/~giambasu
>> Office: 1-848-445-5223 Cell: 1-612-644-8285
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>
>> _______________________________________________
>> AMBER-Developers mailing list
>> AMBER-Developers.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Apr 10 2017 - 14:30:03 PDT
