For the purpose of this white paper, I consider the following Petaflops class architectures in a time frame of about 2007. We assume that individual memory chips will have a capacity of two Gigabytes.
