Ya I actually completely agree with everything you said, my main complaint though is that the difference between my 16 MHz LC II in the early 90s and my blue 333 Mhz G3 iMac was almost two orders of magnitude, whereas my 2.3 GHz i5 Mac Mini only feels say 5 or 10 times faster than my iMac did. The question I ask myself is if Moore's law can no longer lead to faster computing, then what good is it?

For parallelism, I have lost hope that there will be much progress in the compiler automagically parallelizing things like for loops in C++. There's some cool stuff with SSE etc, but it all feels fairly ineffectual to me. Probably people will just move to a new language like Go or Haskell/Scala and just use C for glue code. I could maybe see something like Groovy for low level languages though, where it makes global variables illegal and does some other things to eliminate side effects and then compiles to, say, Go for the 10/100/1000x speedups we'd expect for things like graphics or matrix math.