I'm no expert on the field, but I believe the optimality of Grover's algorithm strongly suggests that the upper bound on the speedup from quantum computing is quadratic (and I apologize if I'm fudging something here). I have no idea about non-linear QM - really over my head - so I suppose I'll have to concede that point.
I agree with you that having conscious carbon but not conscious silicon is arbitrary and silly. Once we understand the brain better we'll surely be able to build one. Again, not something that should depend on location, nanomachines or not.