Retrasado said:
actually, they leave the last SPU unused so that if one of the SPUs is non-functional after fabrication they don't have to scrap the entire chip. Also, the PS3 and the Xbox 360 both have 512MB of main RAM (I have read a few places that the 360 has 10-20MB more on the graphics chip, but I am not sure). The difference is the 360 has 512MB of 700MHz DDR3 memory that is shared across both the CPU and GPU. The PS3 has 256MB of 700MHz DDR3 memory and 256 MB of 3.2GHz XDR RAM (a specialized type of DRAM developed for the PS3 and afaik never really used anywhere else because of the awesomness that is GDDR5). The CPU can access the entire memory but the GPU can only access a portion of the memory. Originally this was the 256MB XDR memory but they have been gradually reducing the OS footprint and allowing the GPU to access more of the memory because htey don't have to worry about it interfering with the OS. I think the idea with the "partitioning" of the RAM was to make it so the CPU would always be able to access some memory and the OS would too. Although this makes it a little more complicated to program for because it is a departure from the more common shared memory architecture, it actually is not a very bad idea and actually foreshadows the current trend toward embedding a large amount of fast RAM on the GPU die. However, in retrospect, it would have worked waaay better if they had included 512 MB (or more) XDR memory for the GPU and keep the 256MB for the CPU (since the CPU can access the GPU's memory if it needs it). Because they did not, it has it made it very challenging to migrate code originally written for the 360's 512MB of RAM to the PS3's (effectively) 300-400MB. While the processor transition would not be too difficullt because basically all you have to do is move the code for the physics and some of the polygon drawing from the GPU to the Cell's SPUs, managing the memory differences would require a major rewrite of the code. This is most likely the main reason the major developers switched to the PS3 for their lead dev platform. It would be quite ironic, but it is possible that this (and the other) improvement to the OS memory footprint will actually make it easier for a game company to go multiplat on the HD consoles. Regardless, its definitely a good thing for the devs.
|
Reading comprehension FTW.
, which is why there will prob. be no one to ever get those security KEY's, that 8th spe cannot be accessed. which is why it's completely disabled to developer's and hacker's alike.
