It amazes me how people don't understand that number of cores and all this crap isn't what defines how good a system is.
-These systems don't run your home copy of Windows. They don't have that overhead slowing them down. That's a huge factor
-The hardware is customized. The OS is completely custom. Everything about how they develop the system is aimed at running games first and foremost. When you write an OS to be as absolutely light as possible for a single task, you are way way way more efficient with less hardware. That's why PS3 and 360 can do what they do with what amount to ****ty computers today.
-Whoever it was that complained about going to less cores on the next system for PS4 - CELL is a highly specialized chip aimed at floating point calculations - super computing. It wasn't designed for PS3, it was a joint venture by IBM, Sony and Toshiba that produced a chip that broke current paradigms. This meant that is was more generalized to super computing and other design goals. Thus, it isn't the best suited for running games due to its nature. Also, four cores can be easier to manage than eight (and PS3 has seven by the way - eight to start, one disabled on purpose during production to account for production errors). Multi-core is about what it's good for, and there is some debate how many cores you need in gaming, given how specialized your code has to be aimed at dividing up tasks a certain way - you're going to get slowed down by the slowest task, and the more you divide up the tasks, the more potential for waiting around. In supercomputers highly designed to run numerous threads independent of one another (i.e. not needing each others data right away to compute), many many cores work. In a game where calculations are so interdependent, you end up losing a lot of efficiency trying to code so all the parts arrive in time.
Oh, and a slower clocked chip doesn't mean it's worse. If you like, I've got a P4 3.2Ghz single core system sitting around. I'll trade you it for basically any quad core system currently on the market from AMD or Intel, even at half that clock speed. Even considering most applications aren't using more than a single core, the modern implementation of instruction sets in those chips allows them to be faster.
The fact is that the Cell was optimized for very specific hardware instructions to certain tasks. To accomplish things that aren't specifically in the hardware instruction set, you need more clock cycles than a chip which has an instruction set that has more robust options. I.E. on the CELL, when they can't do something directly that other chipsets can, they are forced (the compilers that is, mainly, when translating it to machine code) to waste a ton of time simulating that function that isn't in the hardware via many clock cycles doing that through calculations on the fly.
If you want a real world example today, the Snapdragon S4 processor from Qualcomm has only two cores, and clocks at 1.5Ghz. The Tegra 3 from Nvidia has four cores running at 1.4Ghz (with a fifth low power core). The Dual Core S4, at only slightly higher clock speed, is being shown to readily destroy the Quad core Tegra 3. Note that these are all the same ARM based designs. It has to do with specific implementation, design, interconnects, and efficiency. The upcoming dual core Samsung Exynos 5250 is leaps and bounds ahead of both (though this is more due to more implementation of the newer ARM A15 chipset)
TL;DR - You can't tell a damn thing about modern processors, even ones based on the same architecture, based on clock speed and number of cores alone. You need a lot, lot more knowledge than that. Add to that the fact that gaming systems don't run a modern, bulky OS, and that's why these figures really mean nothing till we see the results. If they are even the right figures.
Last edited by Pawsed; 04-06-2012 at 01:04 PM.