The difference is that if Riemann Zeta is incorrect, many theorems that are based on the hypothesis being correct will also be invalidated. Whereas assuming P != NP is limited to a relatively small area of complexity theory. I think a much more equivalent conjecture in complexity theory would be the unique games conjecture. Because the unique games conjecture already provides certain problems to be completely characterized (see: https://en.wikipedia.org/wiki/Unique_games_conjecture#Releva...). Thus, proving this to be true kind of closes the chapter on many problems. Whereas proving P != NP true still leaves many gaps open.