Is it? The last custom design from Nintendo in the console Market was GC 15 years ago, but both Wii and Wii U are rehashes of that architecture and portables have been designed with old/cheap technology forever.
People should not read into my words any kind of doom and gloom, or me trying to downplay Nintendo, I mean for me this is the most exciting product coming from Nintendo since GC and the second time I'm going to be there day one since N64.
But just following basic logic / ockham's razor with the info we really have (not speculation) the really close to 20nm Tegra X1 scenario seems a lot more likely than a super customized 16 nm SoC. I may be wrong and feel free to disregard my mumblings, but funny thing is I'm super pumped about playing portable 720p no slowdown BotW (and maybe -fingers crossed- 1080p on TV), so whatever happens it's already a big win-win for me.
Nintendo's handhelds may have used "old and cheap" technology, but that definitely doesn't equate to off-the-shelf chips. The GBA, DS and 3DS had custom SoCs designed in-house by Nintendo using ARM (and in the case of 3DS DMP) IP. They even designed a whole new chips for the DSi and n3DS as well.
On the console front, for their 3D consoles, we've seen:
N64 - CPU: Lightly customised NEC MIPS core - GPU: Heavily customised by Silicon Graphics (loosely based on an existing CPU/DSP)
GC - CPU: Existing IBM PPC core with significant customisations, including a large set of ISA extensions - GPU: Fully custom GPU designed for the ground up for Nintendo by ArtX.
Wii - CPU & GPU: Lightly modified versions of existing GC hardware.
Wii U - CPU: GC/Wii core modified with multicore support and new cache - GPU: Custom chip designed in-house by Nintendo using AMD IP.
You can argue that Wii and Wii U were "rehashes" of existing tech, but they sure as hell weren't off the shelf chips. Even going back to the NES and SNES Nintendo used custom hardware.
Now I'm not arguing that they're going to use a custom GPU designed from the ground up like back in the Gamecube days, that's clearly not practical for any console maker nowadays. They're going to use the Maxwell/Pascal GPU architecture and they'll almost certainly use stock ARM CPU cores, but their choice of how the GPU is configured (ie number of SMs, ROPs, cache configuration, etc) will be entirely unconstrained by what the TX1 happened to use. Similarly, the choice of CPU cores they use, their number, how they're clustered and what size cache to use is entirely open for them to choose as they wish without constraint. Ditto with memory interface and any other number of aspects of the SoC.
We also need to keep in mind how big a deal this chip is from the perspective of Nvidia's Tegra division. Even in the worst case scenario it will be the biggest selling Tegra powered device by at least an order of magnitude (possibly two) and will almost certainly comfortably outsell every other Tegra chip combined. They're not going to put piles of effort into the TX1 and then just phone this one in. We also know that Nintendo "contributed significantly" to Nvidia's big jump in earnings last quarter, which is about the time you may expect Nintendo to make final payments for completed R&D. They wouldn't be paying Nvidia that kind of money for a lightly modified TX1.
That's not to say I expect something vastly more powerful than TX1, as a 2 SM GPU would seem the most likely given Nintendo's cost and thermal constraints. However, any other similarities to TX1 will likely be little more than coincidence.