This paper presents a SIMD-accelerated software rendering pipeline for 3D graphics processing with multi-core architecture. The multi-core architecture is based on ARMv5 ISA which employs a SIMD unit developed by this work. We also propose a window search bounding box algorithm that can achieve zero failure in pixel tests so that speedup traversal stage is about 20 times faster than the traditional method. Finally, we use an early culling strategy to decrease unnecessary lighting operations.