I reckon cell sounds likes a really good idea, but it looks like sony has just moved hardware complexity to the burden of software developing. This might be all well and good for companies with close ties to sony and it's ps3 who will have a better chance of harnessing the power of cell, but what about companies that write once for the least common denominator and then port many times? Is cell a good 'base' platform to develop for?
There's also the arguments about a compatible ISA. As we know the x86 is so strong because of it's backward compatibility - which is what's kept other ISA's out of the market so far. Intel's future multi-core 32nm manufacturing process sounds a lot like cell at a high level so it looks like they're seeing somethign alot more technical than what cell is at the moment, not to mention Intel already have a massive market share, so this might actually be the deciding factor for cell and it's uphill battle
I reckon Cell could excel in games because those developers are said to not mind the burden of deathmarching
, but outside this im not too excited about Cell (even though cell like processing power of offloading instructions to other cell processors is a really neat idea). But hey, i dont think i know enough to know
With this said im wondering if Intel's future dual and multi-core platforms with bring enough parallism and performance for more realistic physics and AI, or will AGEIA's PPU bring cell like architecture to the pc platform. Time will tell i suppose.
Also, the TI's mvp 320c8x is a multi-core cpu that sounds a lot like cell architecture as well, so maybe it's nothing new (i dont know) ->
http://focus.ti.com/docs/military/ca...ata/cm/miligen
