We have developed a PC cluster connecting two or more PCs that enable high-speedcalculation processing. The layout problem is an important problem generate d in many fields,such as placement of shops and the machines in a factory. The number of combinations ofplacement increases explosively with the increase in the number of placing elements, i.e., theproblem is NP-hard. In this paper, we consider application of genetic algorithms to the layo utproblem placing shops with variable aspect ratios. Then, we try to evaluate the two types ofparallel GA, i.e., distributed GA and master slave type GA, based on numerical e xperimentation.Finally, it is shown that distributed GA is more efficient than master slave type GA.