One job meaning a single MPI process, or a single job with multiple MPI
processes ?
Depending on the system size, grid size and number of MPI processes
asked in 1 job, you may easily consume 64 GB .. At least that was the
case with Amber 12 and 14 ... I know that in Amber 16 a lot of
optimizations were done with the MMPBSA code, but we have limited
experience thus far with Amber 16 and MMPBSA.
All I can say is that we experienced this error often in this context
before ... But I also know this is a generic error that can be due to
multiple other reasons ... So, if you calculated your memory consumption
and you found is fine within your limits, you would have to search for
other potential issues ...
Sorry I cannot help more ...
Vlad
On 01/31/2017 10:27 AM, Albert wrote:
> But I am using amber 16.
>
> My machines contains 64 Gb RAM + 40 cores CPU. I only submit one job there.
>
> I going to recompile Amber to see if the error is still there.........
>
>
> On 01/31/2017 10:25 AM, Vlad Cojocaru wrote:
>> With earlier versions of Amber 12 (14) we have experienced this type of
>> error when the memory of the node was not enough to accommodate all MPI
>> threads requested ... Usually, we solved by reserving the node memory
>> and asking for less MPI threads than the node can accommodate ...
>>
>> But I am not sure if this is the same with the new MMPBSA in Amber 16.
>> To test, you could easily check the memory consumption for 1 frame (its
>> in MMPBSA output) and see how many such calculations your node memory
>> can accommodate simultaneously ... In this way, you will know if this
>> is likely to be your problem and how many threads you can accommodate.
>>
>> Having said this, the error is pretty generic (like a seg fault) and
>> there are different reasons for it .....
>>
>> Best
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Vlad Cojocaru, Ph.D., Project Group Leader
Department of Cell and Developmental Biology
Max Planck Institute for Molecular Biomedicine
Röntgenstrasse 20, 48149 Münster, Germany
Tel: +49-251-70365-324; Fax: +49-251-70365-399
Email: vlad.cojocaru[at]mpi-muenster.mpg.de
http://www.mpi-muenster.mpg.de/43241/cojocaru
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jan 31 2017 - 02:00:04 PST
