Memory systems today are divided into multiple modules, each of which can be serving a separate request at a given time, in contrast to earlier architectures where there was a single interface to memory. What impact has such a memory architecture have on the number of processors that can be supported in a shared-memory system?