We present novel algorithms for efficient hierarchical collision detection and propose a hardware architecture for a single-chip accelerator. We use a hierarchy of bounding volumes defined by k-DOPs for maximum performance. A new hierarchy traversal algorithm and an optimized triangle-triangle intersection test reduce bandwidth and computation costs. The resulting hardware architecture can process two object hierarchies and identify intersecting triangles autonomously at high speed. Realtime collision detection of complex objects at rates required by force-feedback and physically-based simulations can be achieved.