In this paper, we propose a bandwidth-effective volume rendering scheme which subdivides the volume into the sub-volumes and transmits them to the texture units in visibility order. Each sub-volume is rendered in the same manner as the original volume on the graphics hardware and the corresponding sub-image is blended in the alpha blending unit. The sub-volume oriented processing improves the cache efficiency and allows empty space skipping. Moreover, it is capable of rendering volume datasets that do not fit into the texture memory. Simulations show that the proposed scheme is effective for 3D texture-based volume rendering on commodity graphics hardware by reducing memory bandwidth up to 30 times when compared with the traditional method.