Typically, architectural state is replicated, but the execution resources (ALUs, caches, etc.) are not.

Thread Level Parallelism