Image-Space Sheet-Buffered Splatting is a popular hi gh quality volume-rendering technique specially suitable for zoomed views of the data and point-based surface models, that projects the voxels in slabs perpendicular to the viewing direction. Recently, a GPU design of this method has been proposed that considerably accelerates the render ing stage. However, the bottleneck of the method is the computation of the buckets, i.e the structure handling the voxels to be rendered in each slab. This stage of the method is done on the CPU. In this paper, we propose a ne w design of the method that creates and manages the buckets on the GPU. The proposed method is more than three times faster than the previous ones.