"As we've noted before, hardware is getting well ahead of software in the new multithreaded multicore world."

You may note this until the cows come home, but it's bollocks. The average PC has 1 or 2 cores and maybe as many as 4 threads. Writing software for that is piss easy and Windows typically has half a dozen threads in the "ready to run" state if you are doing anything more demanding than sitting idle. The average server machine might have more, but the average server load is embarrasingly parallel and would easily swallow an order of magnitude more cores than the likes of Intel are prepared to offer them. The software is *way* ahead of hardware here.

Having said that, there is this little thing called a memory wall that will make it pointless for either hard or soft ware to parallelise much further without a radical change in architecture. To cut a long story short, you can put as many cores and threads into the box as you like, but as long as you have only one "main memory", you're stuck in the 20th century.