Two Valleys, One Global Minimum

$f$ is convex, so gradient descent converges to the same point, $x = 2$, from every initialisation.

$f$ is not convex: $x = -1$ and $x = 2$ are both local minima, only $x = 2$ is global, and $f(2) = -27$.

$f$ is not convex: $x = -1$ and $x = 2$ are both local minima, and $x = -1$ is the global one because $f(-1) = 0$ is below $f(0) = 5$.

$f$ is not convex, but $x = 0$ is its only critical point, so gradient descent always ends up at $x = 0$, where $f = 5$.