"VASP" is not a method, it is a code. If you are asking for the difference between the Car-Parrinello molecular dynamics (as implemented for example in the code "CPMD") and Bohn-Oppenheimer molecular dynamics (implemented in "VASP", "CPMD", "CP2K", ... most DFT codes), there are several books and reviews available. What do you want to know exactly?

Just some general comments: Car-Parrinello dynamics is stable, ie. the constant of motion _is_ a constant, whereas in BO-dynamics there is always a drift, which can be made smaller and smaller by increasing the accuracy of the forces (and possibly decreasing the time step). The time step in Car-Parrinello dynamics is shorter than in BO-dynamics, but the evaluation is faster on the computer. Generally it is regarded difficult to calculate metals with Car-Parrinello dynamics, at least without any "tricks" (Blöchl-Parrinello thermostat on the electrons), and I don't know of any implementation of the original Car-Parrinello dynamics with many k points and fractional occupation numbers: Might be impossible.

Thank you soo much

But what i was trying to as is How many atoms we can simulate in CPMD and VASP.

My college have Both VASP as well as CPMD,

i wanna to study DNA binding study of a Epoxide metabolite which should i prefer

Thanks again

The bigger your computer, the more atoms one can study. I would roughly estimate that on, say, 32-64 modern processor cores one can go up to some hundreds of atoms, closer to 1000 if you can wait longer. If you need a long MD trajectory, it might take indeed a while.

Thank You so much,

my college have 192 cores cluster on Linux, and 64 cores on Windows

and i wanna to simulate B-DNA (Representative sub-structure) and my drug.

so i can proceed..

Thank you again

thnak you for this information , i would like to know the best specification of my PC to get best data of semiemiprical calculation for large molecules with hyperchem program