Moore's law is only about transistor density, the clock speed race of the 90's mirrored Moore's law but was only indirectly related (in that the smaller you make a circuit the less time a signal needs to propagate through it).
The improvement in GPUs is all in line with Moore's law.
>On top of that, there's been failure to move to distributed and multicore processing outside of GPUs
Few ordinary tasks besides graphics processing benefit significant from multi-threading, and the resulting program complexity means developers don't bother with it for minor improvements.
There are other benefits from the power of modern processors, such as running programs in sandboxes to improve security, and being able to run multiple CPU intense programs on separate cores.
> but when the single fastest increase in my computing in a decade came from installing a 512 GB SSD drive, it means that something went terribly wrong.
No it doesn't, it is just that SSDs cut disk access speeds more in absolute terms than going from 1Ghz to 4Ghz improves program speed. A modern 4Ghz CPU is more than 4 times faster than a 1Ghz early 2000's CPU, but in absolute terms the speed up is small (ie 500 ms to 50 ms is more noticeable than 50 ms to 5 ms).
Try running a CPU bound task (the ordinary user doesn't have these) on a early 2000's CPU compared to a modern CPU.