This paper introduces 3DSSD, a lightweight and efficient point-based 3D single-stage object detector designed to balance accuracy and efficiency. Unlike existing point-based methods that rely on time-consuming feature propagation (FP) layers and refinement modules, 3DSSD eliminates these components to reduce inference time. The key innovation is a novel fusion sampling strategy that combines feature-distance-based (F-FPS) and distance-based (D-FPS) sampling to preserve more interior points of foreground instances, enhancing both localization and classification accuracy. The box prediction network includes a candidate generation layer, an anchor-free regression head, and a 3D center-ness assignment strategy to further improve performance. Evaluations on the KITTI and nuScenes datasets show that 3DSSD outperforms state-of-the-art voxel-based single-stage methods and matches the performance of two-stage point-based methods while achieving significantly faster inference speeds.This paper introduces 3DSSD, a lightweight and efficient point-based 3D single-stage object detector designed to balance accuracy and efficiency. Unlike existing point-based methods that rely on time-consuming feature propagation (FP) layers and refinement modules, 3DSSD eliminates these components to reduce inference time. The key innovation is a novel fusion sampling strategy that combines feature-distance-based (F-FPS) and distance-based (D-FPS) sampling to preserve more interior points of foreground instances, enhancing both localization and classification accuracy. The box prediction network includes a candidate generation layer, an anchor-free regression head, and a 3D center-ness assignment strategy to further improve performance. Evaluations on the KITTI and nuScenes datasets show that 3DSSD outperforms state-of-the-art voxel-based single-stage methods and matches the performance of two-stage point-based methods while achieving significantly faster inference speeds.