No, I don't buy this.
If the final target was e.g. a Tegra X2 using only the Cortex A57 cores, there would be nothing stopping Nintendo from sending devkits with Tegra X1 and the only practical difference would be memory bandwidth and memory amount. This is hardly a showstopper for developing 1st-gen games, especially if the final hardware is coming with better specs.
You might not buy it, but out there in the real world it's what Nintendo do and it's what they did this time.
Nintendo had basically final prototypes of the Switch back in June last year at latest (even down to the cooler and battery), possibly had them earlier, and they had final silicon for many months before that. Dev kits were shipping in their tens of thousand in 1H 2016. Switch isn't a generic phone or tablet, it's a distinctive product and Nintendo will have been working on it for some time.
You might think Nintendo should work very differently - and are perfectly entitled to do so - but Nintendo will have had reasons to develop Switch the way they did and over the timescales involved.
I think it's very likely that TX2 was never even a contender for Switch.
That amount of people who buy Nintendo hardware for those reasons has been counted.
It's 13.5 million sales, and it was considered an enormous failure.
If Nintendo had lifetime projections of 13.5 million sales for the Switch, they wouldn't have launched the console at all.
You're forgetting the tens of millions of Nintendo fans who game on mobile. It's also a mistake IMO to assume that every customer that
might buy a platform for Nintendo software would buy
every Nintendo platform.
Anecdotally, I know people that didn't buy WiiU that were day one for Switch. Funnily enough, all of them bought Zelda - a game also available for the WiiU.
As soon as the reality of its performance capabilities settles in, Nintendo will be selling this as a handheld first and foremost. The post-3DS crowd will be their target, not Nintendo enthusiasts.
Of course, charging $300 for a 3DS replacement will also wear down the sales eventually.
I'm not sure you get to define who is a Nintendo enthusiast or not. There are plenty of 3DS owners who seem to like Nintendo plenty. And plenty of retro hipsters who wear Nintendo t-shirts and hoovered up the NES mini. And all that Pokemon go and Mario run stuff went down like a house on fire.
Nintendo have currency far beyond the people who bought the WiiU.
Tegra X2 has been finalized at least since late 2015 and it has been in production at least since mid-2016.
Full assembly for the Switch's production model reportedly started in November, but the Tegra X2 had been going into Tesla cars one month earlier. And you can bet automobile QC is a lot slower and much more demanding than it is for handheld consoles.
And yet TX2 has
only just come to Nvidia's embedded systems kit. And I don't agree that the power figures prove it was a sensible choice for Switch - board TDP vs actual platform power drawn from the mains leaves lots of room for unknowns.
Tesla's cars have bigger batteries than the Switch, and won't have to run in "No Denver" mode which disables much of the TX2 platform. Teslas are also not low BOM devices that sell millions per quarter.
Even with lower clocks and Denver disabled to try and squeeze down into the Switch, I'm not sure that spending the power on a double width memory bus would have been a great choice for Nintendo. Even Snapdragon 820 isn't using a 128-bit bus.
Keeping BOM under control is critical for Nintendo.
