Models of Computation

What determines the power and speed of a computational model?

Which features of a model change what it can compute, and which change the resources it needs to do so?