Core parking
A power-saving technique where the operating system "parks" (makes idle and puts into a deep sleep mode) some CPU cores under low load, and concentrates the work on fewer awake cores. It saves energy and can let the active cores boost higher.
Mechanic: under light load not all cores are needed. The OS scheduler consolidates the threads onto a few cores and parks the rest — they're put into a deep C-state where power gating turns off almost all power to them. Wins: (1) energy (parked cores leak almost no power), and (2) often higher boost for the awake cores (more of the power/heat budget becomes available when the others are off). When the load rises, parked cores are woken (with a small latency). Trade-off: overly aggressive parking can cause sluggishness (latency to wake cores when a sudden load arrives) → the scheduler balances it. Especially important in heterogeneous chips (park P-cores and run on E-cores under light load). It builds on power gating. Related to power gating and heterogeneous cores.