[gmx-users] optimize cpu with gpu node for gromacs

[rahul dhakne]

Dear all Gromacs User,
           I am using one GPU node (NVIDIA Tesla C2050, 2.5 Gb ,6 core )
for simulation purpose on Intel(R) core I7 (3.0 *GHz, 16gb*) system. I did
tested on this workstation with Gromacs 4.6. It seems from these that
performance of GPU is imbalanced with CPU. It was showing shorter cut_off
but I am using least of it.

Using 1 MPI thread
Using 8 OpenMP threads

1 GPU detected:
  #0: NVIDIA Tesla C2050, compute cap.: 2.0, ECC: yes, stat: compatible

1 GPU auto-selected for this run.
Mapping of GPU to the 1 PP rank in this node: #0

starting mdrun 'Protein '
500000 steps,   1000.0 ps.

Writing final coordinates.


NOTE: The GPU has >20% more load than the CPU. This imbalance causes
      performance loss, consider using a shorter cut-off and a finer PME
grid.

               Core t (s)   Wall t (s)        (%)
       Time:    26585.222     3827.438      694.6
                         1h03:47
                 (ns/day)    (hour/ns)
Performance:       22.574        1.063

How to optimize this mdrun to get maximum use of GPU and accelerate the
performance?. How many thread to be given? What to be optimized so that I
can use for simulation purpose? Please let me know this is enough
information.



-- with regard
    Rahul
-