A destructor has nothing directly to do with releasing memory - instead it is a "hook" to allow custom code to be run when the object is eligible for reclamation. That is, it's the opposite of the constructor - the constructor does not allocate the memory (as that is done by the GC prior to the constructor being invoked) and thus the destructor does not release the memory (as that will be done by the GC afterwards).
While the GC can manage native resources (e.g. other objects and object graphs) just fine, external resources such as file handles must still be "manually disposed". For instance, imagine a MyFile class, where the destructor would ensure the file, if open, would be closed - while it is arguably "better" to make it a requirement to invoke a Close/Dispose operation upon the object, the destructor can be used as a fall-back mechanism in this case.
I would argue against the general use of destructors in languages with a GC. There are a number of subtle issues they can introduce such as apparent non-determinism and the ability to accidentally keep objects alive - even in languages like PHP that uses reference-counting. (The Java/JVM and .NET models use finalizers which are even more finicky.)
Happy coding.