The size of your data is fairly large at over 8GB. If you do not have the physical memory to match over all your processes, you may be swapping out to virtual memory from disk, and processes will be very slow. If you do have the memory, it can still be slow just because the data requirement is so large. Sometimes, data is oversized because it has been resampled to too fine a grid, often to match a high resolution anatomical dataset. You can resample with 3dresample to a coarser grid that matches your original EPI acquisition or better yet, change your processing pipeline to write out the coarser resolution in the first place (adwarp -dxyz nn or @auto_tlrc -dxyz nn).