As long as the simulation follows consistent rules, there's no difference between living a simulation and a computable universe. So if you believe in a computable universe, the question is an idle one (except in the context of anthropic arguments, but then it becomes about whether you sample once from every class or once from every instance).
If you believe in a noncomputable universe, the question becomes, what are the observable differences? Do you think there is a true random source at the center of the cosmos?