[gmx-users] Free energy calculation - parallelization

[Andreas Schöppach]

Hi,

I have a question concerning the parallelization capabilities of 
Gromacs  (>=v5.0.5) for free energy calculations.
My computer specifications are 2xIntel® Core™ i7-4790K (2x4 cores) and 
2x NVIDIA Tesla k20.
Reading part of the manual, I know that gromacs will offload the 
nonbonded forces (VdW, Coulomb) to the GPU for a normal MD simulation 
automatically.
But how will Gromacs parallelize a TI? In the same way as a normal MD 
for each lambda step? Are different parallelization schemes for a TI 
possible regarding different lambda steps?
Thanks for your replies!