64 bit is just a doubling of the data bus, in other words for each clock cycle the CPU can move twice as much data than before, it also changes the address bus, however in the past theyve resorted to other techniques to increase addressable memory beyond what the 2^N power gives you.For example with an 8 bit data bus and 16bit address bus you have max values for data of 256 and max address ranges of 65536 which is how PC's first started.Now we have a data bus 64bits wide or 8 bytes, dunno what the address bus is these days but a pinout of a CPU will show it.As far as backward compatibility, the instruction sets a CPU has has remained backward comaptible for quite awhile, so in general old software still works, they just add new instructions, like MMX for example, its just some added instructions that help with thruput and math I think, when a program is compiled the compiler can tailor make an app to fit the system its going to run on and either take advanatge of the new instructions or not, this makes it possible to run old apps on new hardware.Some of it is software specific, but most things are backwards compatible.