Clock speeds were always more about marketing than actual performance.
There are a few power-user applications where a 10X to 1000X speed bump would be very welcome - mostly video, audio, and AI.
For example - if you generate/process audio at 32X or 64X oversampling you can eliminate all of the usual DSP and conversion artefacts, even for difficult processes like non-linear distortion.
But for most applications, most users have more cycles than they need.
Today's CPUs are literally like 10-50x faster than CPUs of that era depending on how many cores you're using at once. Storage is an order of magnitude faster as well. Mission accomplished!
I've got plenty of complaints about how the industry has changed since then, but raw hardware performance certainly isn't one.
I think a lot of those predictions came from a world unconcerned with cores and threading.
The jump from sub ghz to 1 ghz (p2 to p3) came about suddenly, the jump from 1ghz to 3ish ghz came about suddenly (p3 to p4); and then things moved to prioritizing chip real estate for cores/threading/parallelism.
The chips we have now are great, but it would be damn nice to have a 10ghz core somewhere.
( I eventually upgraded my RAM before ditching the Presario, too, but I think it's because at the time I wanted to play Dark Age of Camelot without lagging to death in team fights.)