Accurate segmentation of teeth and root pulp canals from cone-beam computed tomography (CBCT) images is essential for clinical applications such as treatment planning, root canal therapy, and prosthetics. Manual segmentation is time-consuming, subjective, and impractical for routine use, motivating the need for automated approaches. In this work, we propose a solution based on nnU-Net for multi-class dental structure segmentation. Our pipeline incorporates customized preprocessing, efficient training, and lightweight post-processing. Furthermore, we introduce inference acceleration strategies, including the removal of redundant augmentations and optimized interpolation, which reduce inference time by nearly fourfold with only marginal performance degradation. Experimental results on the MICCAI STSR 2025 Challenge Task 1 demonstrate that our approach achieves competitive segmentation accuracy across multiple metrics, achieving a top-three ranking in the competition. These findings highlight the effectiveness of nnU-Net and our acceleration strategies in achieving a favorable balance between accuracy and efficiency, underscoring the potential of our method for clinical deployment. Our codes are available at: https://github.com/duola-wa/MICCAI-2025-STSR-Task-1 .