Design and operation of a processing device is configurable to optimize wake-up time and peak power cost during restoration of a machine state from non-volatile storage. The processing device includes a plurality of non-volatile logic element arrays configured to store a machine state represented by a plurality of volatile storage elements of the processing device. A stored machine state is read out from the plurality of non-volatile logic element arrays to the plurality of volatile storage elements. During manufacturing, a number of rows and a number of bits per row in non-volatile logic element arrays are based on a target wake up time and a peak power cost. In another approach, writing data to or reading data of the plurality of non-volatile arrays can be done in parallel, sequentially, or in any combination to optimize operation characteristics.