I think a bigger bonus, although not visible of using NT x64 over NT x86 is the NT kernel itself. With each successive version of NT Microsoft has had to maintain compatibility with the older versions. This creates some pretty big issues when you want to make significant changes to the memory subsystem. When MS made the NT x64 kernel, starting with Windows Server 2003 x64, they radically changed large portions of the code. Going from 32 bit to 64 bit at the OS level requires you to rebuild most of the kernel, this includes drivers. During this rebuilding the MS developers threw out large portions of legacy code and revamped how the kernel handles memory access. The newer 64-bit kernel doesn't need to provide the same loopholes and hooks to 64-bit programs that the 32-bit kernel provided to 32-bit programs. New programming target (NT 64) means new rules on how to access OS level functions, and this resulted in them closing lots of older security holes and potential problems. To provide compatibility with 32-bit programs the NT 64-bit kernel uses a form of environmental emulation named Windows on Windows (WoW). It creates a virtual memory address space, and a virtual registry for any 32-bit program and locks that program inside that space. Thus a 32-bit program can never directly access memory inside a 64-bit program, including the kernel itself.
Needless to say, the NT64 kernel is light years ahead of the NT x86 kernel both in stability and security. The only way I've ever been able to bring down a NT 64 system was by using badly written HW drivers. And honestly there isn't much you can do about that, HW drivers operate in kernel mode and can't really be protected against (although Windows does try).