How can values of the function f(x) = √(x+2) - √x be computed accurately when x is large?

This is from numerical methods. The answer I thought I logically guessed was that if x is large, it won't make a significant difference on f(x) whether it's rationalized or not. What do you think?