You wouldn't need to do gradient descent because what it does is search the global minimum (which you already have). But calculating a global minimum is is most cases impossible, that's why we use GD to approximate it.