Dear Amber users,
We are having a problem while running one specific set of simulations on
cuda7 GPUs.
I have three identical runs of a protein that are 100ns MD each. one of
them is completed with no problem but the other two have been canceled
several times.
With no error from Amber. But we do get the following error in the job
submission log file:
Static FFT Slab Distribution:
FFT slabs assigned to 1 tasks
Maximum of 80 xy slabs per task
Maximum of 108 zx slabs per task
Count of FFT xy slabs assigned to each task:
80
Count of FFT xz slabs assigned to each task:
108
I have shared this error with our IT department and they have tried to
increase the RAM while running these jobs. From their point of view, there
shouldn't be any issue anymore but those jobs are still getting canceled.
Can you give me a hint on how this might be sorted?
What is surprising to me is that one of these replicates is completed with
no issue. Is it possible that applying the random number generator during
the simulation, leads to a conformation that then leads to this issue
because of unfavourable clashes?
Although visualization of the trajectory and the energetic terms are
comparable with the one that shows no issue.
Thanks in advance.
*Maryam Azimzadeh Irani *
*PhD in Computational Structural Biology *
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri May 24 2019 - 23:30:02 PDT
