No offense, Deneidez, and I know I play dumb on the forums here alot for the sake of fun, but the beer truth (that's as truthful as it gets in my book) is that the 360's CPU is not as powerful as the Cell, in any way, shape, or form, when it comes to game programming, which is often very amenable to parallelism, even outside of the rendering stages.
All the reports you can dig up on "teh interwebs" are basically hogwash cooked up by people who don't know what a mutex or a ring bus is, or what they mean to parallel algorithms, or professional game programming.
The GPU of the 360 is what gives it an advantage over the PS3. Not the memory architecture, and not the CPU. The GPU of the 360 is better, without question. The end result varies, depending on how the game is written -- game engines which start from a single process basis, like PC engines, are going naturally be better on the 360. Game engines written from the ground up for a specific console are going to end up better on the PS3. Its pretty cut-and-dried, and its played out that way for a few years now on the market -- its pretty evident from the PS3 exclusives being better, and the 360 cross platform games being better, and it makes a lot of sense... to me, anyway.
Newell doesn't like the PS3 because it doesn't fit nicely with Valve's PC engine architecture. He knows that parallelism is the future, but he doesn't like the SPU concept -- he prefers that each core had full access to main memory. The SPU idea reduces the amount of raw materials (transitors) needed to make large parallel CPUs possible, in exchange for a lot of ease-of-use functionality. Frankly, I understand his thoughts completely -- and I dislike the Cell idea as well, in that it is crazy inconvenient... but I know it has darn good reasons to exist as it does. I also dislike that I don't live in Hawai'i. Dang it!
What a pain it would (will) be if all parallel systems were (are going to be) as indirect as the Cell architecture. That, and Newell's, opinion doesn't stop the Cell from being a faster CPU than the Xenon though, and it doesn't make multi-core architectures less prone to heat issues, power consumption, and production costs, unless they go the way of the Cell.
We have 8 and 16 core processors (the full-scale kind) on the horizon. When the 16-core mega CPUs are outperformed by 4 core, 32-SPU versions of the Cell for a huge number of applications though... what is there to say? Lets face it, you don't need more than a couple cores in a business environment, outside of some number crunching apps which not every office suite user even cares about. Performance processing (games included) is about performance, not convenience. If it costs the same, or even less, to make, by nature of low transistor count per SPU core... the Cell idea may very well hit the bigtime in the future -- as soon as people accept it, that is.
But.. I just like to procrastinate, drink beer, and make joke posts on teh interwebz. I'm not actually joking with this post, but... it will be entertaining if you want to believe that I am, so that's cool too.
That's a lot of words for Procrastinato. Back to normal for me now. Beer rules.
