Haha
Well, we're still in the dark on what Nintendo are doing and that we'll probably not know the specifics until a teardown happens.
I doubt they'll reveal any meaningful specs during next weeks Switch presentation.
______
For those freaking out about 28nm, again TSMC have refined their process resulting in newer generation nodes which Thraktor highlighted 28nmHPC+ as being very close in comparison to 20nm with regards to power consumption although 28nm has a bigger die size although that's not a big deal regarding how big the Switch is.
Your only worst case scenario as someone else pointed out is that Nintendo didn't choose 16nmFF and put in 1-2 extra SMs to make it 50% to 100% more powerful which means they'd probably have to double the LPDDR4 RAM bandwidth to 50GB/s etc etc
Anyway, 9 days to go until the Switch presentation.
Honestly 16nm and 2sm is still possible. The 28nm answer is glaringly incomplete with fitting the puzzle of a better battery life and being more powerful than July devkits, thrakor is one of my favorite posters, but he thought Nintendo should use 12 A53 cores instead of higher single performance cores, arguing that parallel process would perform the same for less wattage, and I had to argue with him about higher single threaded performance being more important, and while he even said he didn't think they would ultimately go with what he was speculating, he thought it would be a good idea. It's fine for more parallel applications than gaming, but gaming is single threaded territory, yes there is parallel processing but taking Intel extreme 8 core CPUs, against higher clocked 4 core CPUs, the Intel 4 cores don't need double the cycles to crush the 8 core chips in gaming benchmarks.
Xb1s and ps4 pro just released 6 months before switch with 16nm chips that are far larger than what is in switch, still able to lower the price of manufacturing and volume hasn't been an issue, Switch is probably 16nm but what everyone here needs to get over is that doesn't mean anything when it comes to performance, it could end up being 32nm in bizarro world and it wouldn't change the end product's speculated performance of ~160gflops portable and ~400gflops docked. The people here arguing over what process they used, it's actually much cheaper to use 16nm than 20nm because the package can shrink from 11x11 to say 7x7 meaning far more chips per wafer, the 28nm version of this chip would be over 3times larger, even if the wafer was twice as expensive (it's not) it would still benefit Nintendo to go 16nm, not to mention that the production isn't a single run, meaning that the already mature process would only get more mature during production, meanwhile 28nm would get more expensive.
Finally, it really comes down to the October kits being more powerful and longer battery life. 28nm would literally drop the ball on at least 1 of those two, and would almost certainly actually be no change from the X1 kits, meanwhile both factors got better, it's probably 16nm and again it only academically matters, you're getting the same device either way.
P.S. Xb1s chips would be much more expensive to produce, yet xb1s is cheaper than xb1 to make, it is comparable in cost at worst to 28nm.