Feedback from massive stars dominates gas dynamics, and thus the gravitational potential, around forming stellar clusters. We report on numerical tests of two hypotheses here: that early star cluster dynamics will be strongly influenced by 1) primordial binaries or 2) early forming massive stars. We use the Torch software, which uses the Astrophysical MUltipurpose Software Environment (AMUSE) to couple together multiple codes every timestep. Codes used include: the Flash adaptive mesh refinement magnetohydrodynamics code, modified to include ray-tracing ionization heating, stellar feedback from massive stars, sink particles, and a star formation method; the ph4 N-body code; Multiples to follow tight binaries; and SeBa to follow stellar evolution of individual stars. We initialize spherical, turbulent, subvirial, magnetized clouds with <it>M = 10<sup>4</sup></it> M<sub>?</sub> and follow their gravitational collapse. When stars form, we 1) include primordial binaries following the field distribution or 2) force the first star to form to have a mass of 50, 70, or 100 M<sub>?</sub>. We find that primordial binaries are required to explain observed binary distributions, but they have only modest effects on the resulting structure of the cluster. That structure depends critically on the gas distribution. Subcluster formation and interaction is common, and results in strong morphological and non-monotonic mass evolution of subclusters. The time of formation of massive stars with strong feedback can dramatically change the mass and morphology of the resulting cluster. Lack of early massive star formation results in centrally concentrated clusters with high total star formation efficiency, while early massive star formation results in diffuse, inefficient subsequent star formation.