If there's some uncomputable physics that would allow someone to build such a device, we ought to redefine what we mean by computable to include whatever the device outputs. After all, said device falsifies the Church-Turing thesis, which forms the basis for our definition of "computable".