I have used GROMACS to simulate coarse-grained lipid membranes, as it is a code designed for (mainly) simulating bio-molecular systems. There are also an extensive set of analysis tools which are straightforward to use, and the documentation is very good. I have never used NAMD, so am not saying it wouldn't be just as good, but I know from colleagues that it is good for systems with large numbers of particles and it has the ability to integrate with VMD.

NAMD and GROMACS both are good for membrane proteins, Gromacs is fast but NAMD is slow on single node but if you use in parallel it have better performance and accurate.