>Did you read and comprehend what David wrote? He's not disputing whether Sony
>counts loads/stores as FLOPs.

Which we know as is not the case, from APU figures.

> You are trying to argue they do, but so far the
> only theory you had was pretty much been debunked by the guy who oversaw the design of the chip.

As I understand, he is an architect, not the one who decides on the bits of a tiny subsystem.

> Right, and somehow one of the lead architects managed not to notice that?

Sure. This is DD2 we are talking about, a chip with its CPU swapped for something else. In fact, there is no requirement that the CPU core has to be PowerPC at all.

>Besides, there is no indication that Nintendo has any desire or need to use CELL.

No, more like IBM is recycling same FPU/Altivec macro design across three customers with very similar requirements. Why do the work three times when you can work once and cut and paste three times? IBM owns CELL IP and may do whatever it pleases with it, just like IBM owned Gecko IP and offered to sell it to 3rd parties beside Nintendo.

>Their console is much less powerful, because among other things, it won't be supporting
>HD output

Gecko FPU was "burrowed" from 7xx series; it was a 440 Integer + 7xx FPU hybrid.

> (at least not at the level of the PS3 and Xbox360). So why would they
>ever use CELL or a CELL-like chip, when they could use something cheaper?

I never said Nintendo was using CELL; IBM is simply recycling the same Altivec/FPU macro across three designs. It cuts cost and reduces time to market. Will Nintendo be penalized by extra transistor cost? Not really, when the transistor cost is so tiny nowadays.

> You're entire argument is pretty irrational.

It is very rational if you understand IBM's new product strategy; consolidate all CPU products around recycleble cores. That FPU/Altivec macro appear to be designed specifically for console customers, then it makes sense that it tries to fulfill the requirements of all three, including the support of Gecko FPU extensions.