In recent years, increasing power consumption and rising electricity costs associated with supercomputers have become critical concerns. The A64FX processor, developed for the Fugaku supercomputer, includes “Power Knob” capabilities that reduce power usage by adjusting specific hardware functions, such as clock frequency, the number of active floating-point units, and core states. This study investigates the correlation between these power knobs and application energy consumption using performance monitoring unit (PMU) counter values on the A64FX processor. By measuring the energy consumption of the eight microbenchmarks, we demonstrate that the optimal power knob configuration can reduce energy usage by up to 53.8%. Additionally, we collected all PMU event counter values and identified events that exhibited remarkable changes in response to power knob adjustments. Based on these observations and the characteristics of each application, we selected representative events most closely linked to application behavior. We then derived summary metrics to determine the optimal power knob settings. Using these metrics, we classified applications according to their power characteristics, demonstrating that optimal power knob configuration can be selected based on application-specific tendencies.