A Simple Recipe for Competitive Low-compute Self supervised Vision Models