create pandas dataframe by dimension and datatype, without copying

Question

I want to make a pandas dataframe that will occupy over 50% of my RAM. The normal creation of a frame (initializing numpy arrays and creating a dataframe out of that) will h