When cooperating with others carries a direct fitness cost for the individual, natural selection should act against such behavior. However, cooperation is widespread across natural systems. Here we summarize and expand on a recently published work that highlights the importance a previously overlooked factor, the population shape. We used a combination of three in silico systems, Aevol (complex genomes, cooperation based on public good secretion), Aevol-lite (one-locus two-alleles, cooperation based on public good secretion) and CAevol (Prisoners dilemma with two pure strategies, cooperate and defect). In all three systems, higher level of cooperation evolved when populations lived in bulky worlds, a torus created by folding 100x100 quadrilateral grid, rather than slender ones, based on a 4x2500 grid. The result was not intuitive, and we able to discern the population dynamics only upon close examination of the full populations over time, in Aevol-lite. Populations are in dynamic equilibrium, constant emergence and extinction of cooperator patches. We show that slim populations effectively constrained the expansion and thus decreased the size these patches, leading to less cooperation on average. In the extended abstract we further discuss the implications of this work for natural systems, with populations embedded in different dimensions, populations as graphs, and also it with the work on infectious cooperation.