The processor itself may be able to run more efficiently when running with it's native 64-bit wordsize but that does not take into account that on Windows x86_64, 32-bit applications must run on the Windows On Windows framework. In other words, if not complied for 64-bit, 32-bit applications will run slower because of this abstraction layer between the OS and the CPU having to translate every call. A 32-bit application on a 32-bit OS will run faster than if run on a 64-bit OS, and hence 64-bit Windows OSes and the processors they run on can be thought of as being "slower" in this circumstance since the ability of the processor to perform operations on larger pieces of data at a time is not being utilized and there is extra overhead in running the application.
The Ithanium architecture was really nice actually. But no one wanted their x86 apps to run at 1/2 the speed on IA-64 processors. If compiled for IA-64, instead of x86, the applications ran much faster, far better than we get on x86_64 after normalizing for research gains. When AMD released their 64-bit processor that could run x86 apps at native speed, that was the death of IA-64. IA-64 was our gateway out of this miserable micro-instruction laden x86 architecture. AMD made millions. :- /