Recently, aggressive voltage scaling was shown as an important technique in achieving highly energy-efficient circuits. Specifically, scaling V dd to near or sub-threshold regions was proposed for energy-constrained sensor systems to enable long lifetime and small system volume . However, energy efficiency degrades below a certain voltage, Vmin, due to rapidly increasing leakage energy consumption, setting a fundamental limit on the achievable energy efficiency. In addition, voltage scaling degrades performance and heightens delay variability due to large I d sensitivity to PVT variations in the ultra-low voltage (ULV) regime. This paper uses circuit and architectural methods to further reduce the minimum energy point, or Emin, and establish a new lower limit on energy efficiency, while simultaneously improving performance and robustness. The approaches are demonstrated on an FFT core in 65nm CMOS.