As systems become more power-constrained, the challenge will be to maximize power-efficient performance across a broad operating range. Each core in a multi-core system, for example, may be operating at very low voltages when running highly-parallel workloads with sufficient work to utilize all cores. Seconds later, after other cores go idle, a single core may “turbo”, i.e., operate at a higher voltage and frequency, and use all the shared cache and bandwidth resources to improve single-thread performance. In systems with higher core counts, the high voltage/high frequency operating points will not be the operating points where performance is most critical. Instead, the highest performance operating point will be where the maximum number of cores is active. In a power constrained system, this operating point will be characterized by all cores operating at the lowest voltage to ensure the system meets power constraints.