There would be differences between the complex conscious computers
Because the computers needed to instantiate a consciousness program would in reality be so large and complex, there would have to be differences in physical activity between the two systems.
The supervenience thesis wouldn't be violated then, because the physical activity underpinning the conscious system and the nonconscious system would be different. So computationalism isn't self-contradictory.
Argument anticipated by Tim Maudlin (1989).
The argument as anticipated by Maudlin
Maudlin writes:
"To avoid getting bogged down in general claims, let us look more closely at the details of Olympia, and at the sorts of changes which are sufficient to defeat the support of counterfactuals provided by the blocked machinery. Above we considered the cases of Olympia running with and without the idle machinery present. Because of the immense quantity of machinery involved, one might misgive that its removal would necessitate some considerable change in the physical happenings associated with the machine." p.425
References
Maudlin, Tim. 1989. Computation and Consciousness. The Journal of Philosophy, vol. LXXXVI, no. 8. p. 407-432.