splitting into multiple processors• reduces horizontal waste• puts upper limit on peak throughput of each thread (IPC ≤ 2)