The physical activity is different in each system
The concept of physical activity includes a concept of potential activity. So when you have different counterfactual possibilities, you have different physical activity. So the physical activity in each of the imagined machines is different.
Argument anticipated by Tim Maudlin (1989).
The argument as anticipated by Maudlin
In Maudlin's words:
"Since a computationalist cannot give up the sufficiency condition, and since it would tear the guts out of the notion that mental states arise from a complex computational structure to give up the necessity condition, our central focus must be the plausibility and force of the supervenience thesis. Here two avenues of approach are possible. Either the computationalist can argue that the thesis is fine but that, contrary to appearances, the physical activities with and without the machinery hooked up are not the same, or else he can try to renounce the thesis altogether. Let us consider these possibilities in turn.
The supervenience thesis depends upon some independent notion of physical activity, and, in our application of it, upon the notion of sameness of physical activity. It would be beyond our needs to try to formulated an exact definition of physical activity here, but one line of argument deserves our attention. Perhaps one might urge that the counterfactuals supported in the first case differ from those supported in the second case, and ipso facto the physical processes occurring in the two instances must differ" (T. Maudlin, 1989, p. 423-4).
References
Maudlin, Tim. 1989. Computation and Consciousness. The Journal of Philosophy, vol. LXXXVI, no. 8, pp. 407-432.