Magnetic reconnection is a fundamental plasma process in which magnetic field lines change topology and rapidly convert magnetic energy into thermal energy, which is often directly radiated and thus astrophysically observable. However, the rate at which this process occurs in the classical picture is orders of magnitude too slow to explain solar flares. The recent identification of the plasmoid instability, a super-Alfvenic, high wavenumber instability has fundamentally altered our understanding of reconnection theory by providing a mechanism to greatly speed up reconnection. However, the majority of the work done to date has focused on 2D reconnection layers, assuming symmetry in the plane of the current sheet itself. The plasmoid instability is inherently multi-scale, with a large separation between the global scale of the reconnection layer and the resistive length where the instability grows, making 3D simulations impractical before now. We have begun to use the 3D adaptive mesh refinement code Enzo to resolve the reconnection layer. We show the growth of a secondary instability in the plane of the current sheet that drives a huge increase in the rate of reconnection. Understanding how the saturation of this instability controls the global, 3D structure of reconnection regions is required to predict the observable properties of flares, the mass loading of coronal mass ejections, and the acceleration of charged particles in the corona. This research was partly supported by NSF grant AST10-09802, and used computational resources provided under XSEDE grant TG-AST120045.