Gravity as long range force is the main reason of the difficulty of studying 
the evolution of astrophysical systems, from the solar system scale to the 
cosmological one. Supercomputing is thus required, and the common approach is 
resorting to the use of large parallel main frames. 
In this talk I will describe the, modern, approach by mean of Graphic 
Processing Units as powerful and cheap alternative to massive parallel 
computers.