| Aprisaiden said: Here is what SONY needs for PS4: - CELL 2.0 -> 2x PPE, 16x SPE (2x PPE and 12x SPE avalible to devs.) @ 5.4Ghz |
Ok, just a simple question: Who would develop this processor?
Let's face it the last cell was based on the interests of 3 companies, but now they have their architectures.
IBM doesn't need a high power PPU, for their workflow the original PPU would probably be sufficient for 12 SPE, but such a system wouldn't make the Playstation 4 any faster. In fact for them the memoray performance and the local memory sizes of the SPEs are much more critical issues, than the PPU, which is used in most cases as a simple dispatcher, while SPEs represent the number of nodes the chip offers.
Toshiba doesn't even need a general CPU for their workflow, instead they use the SPUs as a kind of SuperPipeline. Both companies use the archicture as it was intended.
The Big idea behind the Cell were the SPUs, but game developers have serious problems to use them in a very effective manner. Instead the next generation of GPUs offer parts of their shaders as a kind of low scale SPUs which would make the SPUs redundant! It might sound a bit exagerated but apple already uses their GPUs as coprocessors and Microsoft goes this way too. The reason is simple. SPUs are stream based processors as each of the shaders in a GPU. But in many cases the application doesn't need all these shaders. The big difference: you do not need a third processor architecture, your graphics guys can do this. You do not need additional hardware, you only use otherwise not needed reserves. For Toshiba and IBM this track is not really interesting, because their systems do not need/use high end GPUs.
As I See it Sony only has the options to stay on the Cell track, pay for the necessary changes and developments and anger the developers that can't integrate the SPUs in their general models, or they will have to switch to a normal multi core and GPU architecture, which is much cheaper due to the fact that these systems will be designed anyway, but it will be nearly impossible to sustain exclussive titles (competitors use more or less the same architecture).
The first way would probably be doomed, due to the higher costs and a lower effective power for game developers, while the second way would mean that the PS4 would no longer be compatible to the PS-3 but it would prabaly be much cheaper and even faster.
