I think himacs is asking a more fundamental question.
If the system is a 64 bit one, in general (but not always), it is referring to the address register. And, in that scope, in general, it means that the data registers are at least the same size. Also, in general, it means that that bus sizes are the same as well, although they may/may not be.
In most cases where everything is matched, its going to be on high end processors where the data registers, address registers, the address bus, the data bus, etc. are going to be pretty much the same, in this case 64 bit. In some cases, a piece or two of the above may be larger, though I can't think of one off hand.
In cases where the address register is larger than the data bus, etc. the reason is often a case where a product is being introduced as a method to provide alternate cost (and therefore market) positions.
The cleanest way to consider a system 64 bit would be to think along the lines that you've got 64 bit addressable ram, 64 bit data representation internally in the cpu, and a nice complement of 64 bit hardware to address ram, etc. external to the chip. Having a single element of the above not being true, may not necessarily disqualify the chip or whole architecture from being a 64 bit architecture system.
We are the people our parents warned us about --Jimmy Buffett
