Relations between various models for massively parallel computers are investigated. These are arrays of finite­state machines -- eventually augmented by pushdown storage -- operating synchronously. The architectures differ mainly in how the input is supplied and how the single nodes are interconnected. The comparisons are made in terms of their ... capabilities to time­construct and time­compute functions. That means given an constant input of length n a distinguished cell has to enter distinguished states after f(1); : : : ; f(n) respectively f(n) time steps.