You guys realize that a huge number of X360 and X360/PS3 games basically rely upon one HW thread to do the vast majority of game functionality, right?
Other threads do side work, like audio, streaming, OS stuff, but they're far from "busy", relative to the main thread. The only times when all cores are decently busy are during portions of the frame where parallel work is obvious.
As I mentioned above, doing tasks in parallel, which have the same memory access pattern, yields horrendous cache thrashing. (The Cell concept gets around this by not sharing quick-access memory with other cores/threads)
Adding more general purpose cores, with a shared cache (unless it was GIANT), would be... stupid.
The "next" Xenon would be perfectly fine as, for example, a quad core, with a larger cache, in the next gen. The only reason the Cell is "better" than the Xenon, as a CPU, is because it can basically absorb all the animation work (which must be done on the CPU) and all but the most basic vertex processing operations (like... projection), thus freeing the GPU to burn most of its cycles on pixel work.
The "Nextbox" would be plenty more powerful if it merely added more flexible pipelines to its GPU, to support more vertex processing work, thus freeing more for pixel work -- it suffers at the moment, because although it has more pipelines than the RSX, it needs to devote a large number of them to vertex processing (mesh skinning is good example, of something expensive the Xenos usually does on the X360, and the SPUs usually do on the PS3) during the frame, whereas the RSX gets everything spoon-fed to it, via the SPUs.
The device described in the original article basically describes a marginally more powerful CPU (I don't even think they need to go that far... 4 cores maybe.. not 6, if they are all sharing a L2 cache), but it could be describing a massively parallel GPU, which would basically embody everything the "next" generation of games would need.
