The dichotomy in the question is false. Part of the answer is model-independent (uncomputability), part is model-relative but empirically resolvable (physical bounds), and part is a category error if pushed too far (denying computation is a real feature of anything).