Doubt a Maxwell-based chip can cut the mustard in that regard. 

There are also other issues because Emily is saying those originally leaked specs are mostly true, one of the problems with that is Maxwell based Tegra has a memory bandwidth of 25GB/sec, this is also listed in the leaked specs for Switch. 

25GB/sec is going to be a big problem for PS4/XB1 parity, I think aboslutely 540p/720p are likely with reduced effects on top of that. 

900p-1080p I think is going to be in docked mode only and for games that Nintendo showed in the trailer -- Zelda: BotW, Mario Kart, Splatoon ... basically games that are only aiming for Wii U levels of visuals. 

He doesn't specifically say FP32 but we have also had ps4 pro numbers of 8 TF at FP16 which he didn't include either in his comparisons so I doubt he is mixing things up. Specially since he is writing an article about it which is based on his sources and the most common standard for measuring TF is FP32 so it would be very odd for his sources to give him FP16 measurements.

I'd rather put this rumor in the nonsense bin rather than give him this many legs to stand on considering the other stuff he says that is also pretty questionable that have been posted by other users.

Not saying it can't be Maxwell, just saying this article is nonsense

The sooner people accept that Switch is a handheld that moonlights as a home console the better it will be for everyone.

Sony doesn't really market FP16 numbers, whereas Nvidia does (sometimes misleadingly IMO) for the Tegra chips. Many sites say the Maxwell Tegra X1 is 1 TFLOP of performance, because Nvidia says so and while technically it may be true, it's also misleading. 

But he is not getting it from other sites or Nvidia... He is getting it from his sources which are supposed to give him accurate information...

Bolded is true and I had forgot about that. Xbox One is only 68GB/s for reference. 

https://en.wikipedia.org/wiki/Tegra#Tegra_X1

Considering that every AAA launched thus far can run on 2GB of maxwell architecture on PCs, I think it is not as dire as people think. Keeping it 720/900 with lower textures/AA/etc, I think, will be doable.

I'm just saying with several mistakes he already made, that mistaking the 1 TFLOP (FP16 number) is a common mistake to make. Even a lot of tech centric sites get that wrong. 

Again keep in mind too the guy is all over the place in his article. Switch is apparently only good for "cartoony" graphics for one, which I take it he means the system is good enough for minimally intensive games like Nintendo's "cartoony" games but not good enough for more realistic games.

If it's 1 TFLOP (FP32), it would have no problem with XB1/PS4 ports, so that whole reasoning by him doesn't make sense. 

Good. Very good. I like the name Maxwell much better than Pascal. Maxwell is a real 80s mans name.

My guess from the beginning was about 400 gflops performance (fp32) and all the information released more recently makes me more sure that Nintendo would have compromised performance slightly for commercial costing reasons. I have absolutely no problem with 400 gflops myself and it could be a bit lower or a bit more but I think thats realistic especially with the claimed longer battery life now.

The dev kit had 4 Arm A57 64bit processors and shared memory for graphics and cpu with a memory bandwidth of 25.6GB/s

In the Nintendo custom chip which could be using a later fabrication process as its a custom chip. It's going to need frame buffer memory in main chip with a few other caches.

This is not going to be a reference design with maximum performance it is going to be a costed version where price is going to be critical to Nintendo. They will want to use cheap memory and possibly multiple vendors of memory chips to keep prices competitive.

Lets not forget the end product will have it seems;

1. home console performance between last gen and current gen.
2. Portable performance of last gen plus a bit including a lot more memory
3. 1-2 player portable gaming out of the box of 5-8hrs life
4. Unlimited gaming in a vehicle using usb power
5. low cost virtual reality system capable of games somewhere between probably wii and wii u in performance
6. Easily transportable gaming engine (tablet) that you can move from dock to dock in the house so it can be used in the living room, bedroom, dining room televisons etc without having to disconnect wires or move power supplies.
7. Will get amazing Nintendo games from both their home console developers and portable game developers/studios.
8. Will get easy conversions of android and ios games with many enhancements and decent controls
9. Will get conversions of last gen games which can now be played on a portable and possibly in VR.
10. Will get VR versions of many classic Nintendo games.

I'm just trying to make the point the Switch offers huge gaming possibilities well beyond the small difference in performance it may or may not have. If Nintendo Switch really is lets say a 1 terraflop gpu we are going to get royally sc****d when it comes to battery life. Careful what you wish for.

But at this point, what we should be saying is that this article is nonsense instead of giving him benefit of doubt. Because what you are essentially saying is that his sources, for whatever reason, gave him FP16 numbers instead of FP32 which to me sounds like nonsense. Why would a source give the author marketing oriented numbers? I get the author doesn't understand it but the sources don't either?

If he has made this many mistakes, this article isn't reliable source of anything at all. 

Will it be Maxwell? It certainly could be. But is it due to this article being right? I would rather say he has no clue and this article is less valid than most other rumors we have already seen.