Hm. Could that explain the odd sizes of certain components?
Probably, in combination with other factors that I've mentioned in previous posts, such as adapting the Radeon components from TSMC (which they were specifically designed for originally) to Renesas. Also, "bc bloat" and different design targets could come into play.
Heard where? The guy from Chipworks said an advanced 40nm process.
40nm was Jim's educated guess at a glance of the chip. 40nm, 45nm, and even 55nm are apparently close enough that to tell for sure you need to precisely measure the transistor gates. Even if it is 45nm Renesas, he wasn't much off base. The nodes are very close and it appears to be using a similar CMOS bulk silicon process to TSMC's.
They're not confused about the CPU?
Weird. Although as we know Wii U isn't SoC, so would Nintendo go to expense of customising it. And of course the question has to be asked, why use 45nm above others?!
Nope, definitely talking about Latte. As for being a SoC, Latte kinda is. It's got what: an ARM cpu, gpu, AHB on-chip bus, DSP, eDRAM. Pretty much all the components a SoC has are also found in Latte.
Why? I speculate that it's a matter of pricing, reliability (as bg said, Nintendo like their main nodes, and 45nm is quite mature at this point), and Renesas' current position in the market. I'm not sure that Renesas have the capacity to mass produce 40nm SoCs at this point. Doing a quick search turns up many stories of them outsourcing 40nm production to TSMC. Meanwhile, another interesting article spoke of them putting priority on getting their 90nm and 45nm production lines back up after the Great Earthquake. These mature nodes account for a great deal of Renesas' current revenue.
Also interesting is that the 40nm eDRAM is still listed as "Under Development" on their website. Makes you wonder if they ever got there...
Doesn't the size of the eDram match Renesas numbers for 40nm? Or is this another size not matching up thing that we're just going to ignore?
Nobody has ignored any size discrepancies. The block sizes do seem to be somewhat of an outlier amongst all the data we have, though. When I did some measurements of the eDRAM a while back (which I can't seem to find now. ugh), it did seem pretty much in line with their 40nm cell sizes. That is if I measured correctly (I did it as accurately as I could, but there was some rounding involved). We should probably take into consideration that Renesas gave those numbers years ago. It seems possible that they got the cell size down a bit so that 45nm cells end up being the same/similar to their previous 40nm quotes.
As far as I'm aware 40nm and 45nm is essentially the same thing. Same node I mean. It's used interchangeably. Same thing with 32nm and 28nm.
Interesting. I'd like to hear more on this.