Power consumption in digital systems, especially in portable devices, is a crucial design factor. Due to downscaling of technology, dynamic switching power is not the only relevant source of power consumption anymore as power dissipation caused by leakage currents increases. Even though power gating is a seemingly simple method for reducing the leakage power, the implications of introducing power gating to a design have to be analyzed in detail. We present an extensive analysis of the impact of fine-grained power gating on the overall power consumption. The presented results are based on the analysis of an actual implementation of power gating in the datapath of a very long instruction word (VLIW) processor. The extracted power consumption values clearly demonstrate that the overhead of power gating is, in contrary to the analysis found in previous publication, not determined by the energy required to switch a power domain on. Rather, it is determined by the energy consumption of additionally required modules. We show that, for the break-even point case, about 2/3 of the energy overhead is caused by the isolation cells, about 1/3 by the control modules, and only roughly 1% by the energy to switch a power domain on.
|Conference||20th symposium on Great lakes symposium on VLSI, Providence, Rhode Island, USA|
|Period||1/01/10 → …|