Actually, I changed computers, told them I junked the other one. They wanted me a photograph of my system and the disc and my details, but that was about it. I don’t recall it being a hassle.
If the 32-bit operations were being done in some kind of ‘emulated’ mode because the CPU only had 64-bit ops available, then there would be performance concerns, but there isn’t with the x64 architecture. All ops run natively at full speed, just like 16-bit ops did when CPUs were 32-bit.
The only performance gains you’d get that are directly related to 64-bitness (rather than other general improvements like having more registers) are if the program is specifically manipulating large amounts of 64-bit numbers. In my experience, variables that are 64-bit specifically because they need to be (as opposed to being so just by being the default word size) are a rather small portion of all data and rarely the performance bottleneck.
And some people have argued that going to 64 bits actually causes a performance drop (again, separate from other architecture improvements), since now pointers are taking up more cache space, a larger proportion of the bus bandwidth, etc.