Multidimensional lattice constellations which present signal space diversity (SSD) have been extensively studied for single-antenna transmission over fading channels, with focus on their optimal design for achieving high diversity gain. In this two-part series of papers we present a novel combinatorial geometrical approach based on parallelotope geometry, for the performance evaluation of multidimensional finite lattice constellations with arbitrary structure, dimension and rank. In Part I, we present an analytical expression for the exact symbol error probability (SEP) of multidimensional signal sets, and two novel closed-form bounds, named Multiple Sphere Lower Bound (MLSB) and Multiple Sphere Upper Bound (MSUB). Part II extends the analysis to the transmission over fading channels, where multidimensional signal sets are commonly used to combat fading degradation. Numerical and simulation results show that the proposed geometrical approach leads to accurate and tight expressions, which can be efficiently used for the performance evaluation and the design of multidimensional lattice constellations, both in Additive White Gaussian Noise (AWGN) and fading channels.