Actually I was the author of the other article and was well aware this one was the same story. However I couldn't be arsed to change it around to say its a rumor. But I made sure Tidus did, the question mark indicates the rumor

Last I checked, engineers don't make the calls on what they can do. Artists and sound techs don't make that call either. Saying "they" hate it is a bit off, you can see in the writings the people are pretty much happy about anything they get to play with and they study the hardware down to the finer points, although I doubt they'd really disclose the whole thing for the sake of retaining some trade secrets.

I'm pretty sure anyone could do better, it seems that EA just wrapped the code and then added some bare options to it for the Orange Box. Given they do have a standing publishing agreement it's not like they can go back to Vivendi and say "hey you know how we told you to jump in the lake... yeah... can you port and publish this violating our other agreement with EA?" I'm sure there was something there for EA to capitalize on through all this since they have no IP rights to anything Valve makes.

L4D plays well enough to generate hype, the only question is if it can sustain that hype after people start playing it in a mass audience, constantly doing 4 player missions with random people who have the option to revive you if they want or just leave you to sit until they absolutely need a second hand and then unlock the door. The adaptive zombie behavior is always good but I guess the question is where the AI peaks because that will be the marker when things start to slide off.

The IBM SPE setup isn't planned for use in anything in the next 2-3 years, the focus for business computing is still large cache based processing. Such reason is why you see Intel chips sporting 12MB of cache or more given the amount of data the pipeline is trying to shove through. Right now the focus is on producing more cores per die while retaining equal performance across the silicon.

Their venture after that is bringing the GPU on die with the CPU for Hybrid / Fusion processing. On top of that NVIDIA and ATI are working on open computing based on their GPU technology instead of reinventing the wheel and trying to use something like an array of SPE's since the performance figures from their GPU line has shown to be stunning anyway.