Multi-core Processors

Ars Technica has an article on why multi-core processors are going to lose out in the long run. This type of parallel programming problem was talked about in some of my senior level classes at Virginia Tech. The real problem is that there's only so many problems that fit the parallel programming model. After you start stretching the model, you're basically into the realm of "solution looking for a problem" and you can't get the performance benefits you'd naively expect. So it sounds like all the fun research will be in memory buses for the next few years.