It is a fact that gaming technology is a state-of-the-art tool for military training, not only in low level simulations, e.g. flight training simulations, but also for strategic and tactical training. It is also a fact that users of this kind of technologies require increasingly more realistic representations of the real world. This functional reality threatens both hardware and software capabilities, making almost impossible to keep up with the requirements. Many optimizations have been performed over simulation algorithms in order to fulfill these needs; no definitive solution, however, has yet been achieved. The question that arises naturally is, then: Does any generic global solution to the problem of the uneven growth of the computational power requirements with respect to the available capacities exist? This paper presents the problem by describing a real situation, analyzing the Batalla Virtual3 case of study and, in answering the motivating question, it proposes a potential software architecture employing a grid desktop computing (GDC) framework to empower constructive simulation systems, pulling off an adaptive hardware infrastructure. Additionally, as the constructive simulation scenarios do not fully adapt to the market-available GDC frameworks, the solution recommends suitable modifications to these software.