Random matrix theory (RMT) has been used to derive the asymptotic capacity of multiple-input-multiple-output (MIMO) channels by approximating the asymptotic eigenvalue distributions (AEDs) of the associated channel matrices. A novel methodology is introduced which enables the computation of the asymptotic capacity for a generalised system in which two relays cooperate to facilitate communication between two remote devices. It is computationally demanding to calculate this capacity using RMT when nodes are equipped with large-scale antenna arrays, and impossible in the case where asymmetry exists between channels within the system. This is because deriving the capacity across the combined channels from the relays to the receiver involves polynomials in large and non-commutative random matrix variables. This paper uses free probability theory (FPT) as an efficient alternative tool for analysis in these circumstances. The method described can be applied with no additional complexity for arbitrarily large antenna arrays. The minimum SNR required to achieve a given asymptotic capacity is computed and the simulation results verify the accuracy of the FPT approach.