oberon, i was pretty close then with the 6 years
and pipelined cpus have been around since at least 1962.
while its true that clock speed isn't the global benchmark in PC's it isn't as simple as it appears, and its well beyond the laymans abilities to understand it, which is why it makes such a good marketing ploy. With multiple pipelines, out of order execution, prediction and caching its a phenomenally complex model.
You can sort of compare clock speeds if you normalize them into the same space first, which is approximately what AMD did with their ratings, the advertised clock speed is the normalized rating against intels, they did this because they knew the consumer and marketing depts. would look at clock speeds as a comparison, so they did what they could to show the benefits of their more bang for the buck with a normalized ghz/mhz rating
A 5 GHZ p4 will outrun a 3 GZ P4 , even if the only thing that changes is the clock speed, so ergo the clock speed is important, but as with any statistic it is important to make sure that the scale used is correct.
Anyway with a 64 bit CPU the biggest gain is generally the width of the data transferred, as well as domain size, however alot of systems already transfer at 64/128 bits.
Theres no guarantees a 64 bit system will be faster than a 32 bit one, it may even be slower, it can just be setup to give you the larger domain for a speed penalty.
Its all very complex, and even generic benchmarking really only tells you about that generic benchmark, its easy to make your chip look better since it is such a complex system, and theres a lot of snake oil.
Simplest thing is just to get real world benchmark information using the applications and data you would actually use the machine for as opposed to some tabloid newspaper style benchmark such as 3dmark etc, only that will tell you if its better or not.