“It is very hard to write software that actually uses multiple cores ...”
For some people ;-) ... I certainly see your point, but a lot of people are getting a lot better writing applications for parallel processing systems.
Parallel processing in general has gone from “solution looking for a problem” to “solution to performance issues since we can’t continue to increase clock frequencies on CPUs easily” in recent years. It’s not hype ... it’s the best way to design general purpose CPUs moving forward.
AMD’s latest and greatest desktop processor requires 210W to run at 5GHz with all cores blazing ... that is an insane amount of power (watts, not “performance power”) ... that requires some serious cooling.
Intel went the parallel processing route to kind of work around the insane power consumption required by high clock frequency CPUs. It’s not marketing nonsense ... it’s a bona fide solution to a big problem.
Of course, as you pointed out, not many programmers exploit parallelism well ... however, that is certainly changing. Some of the tools I use for FPGA development are doing a great job using multiple processors for a single FPGA build. A lot of gaming companies are FINALLY taking advantage of it ... it’s only a matter of time before something standardized will be used to exploit all of the cores we have sitting in our CPUs :-).
Yes those of us in RTL VHDL or Verilog land know how to do it. That's why we get the big bucs.
The problem is the average app coder is lazy and just wants it to work so they can get paid, they have no clue of what they are doing.
So here is an interesting question how many cores is your compiler using, be careful, how many did you pay to have unlocked?