Absolutely no way they could make a 32GB cartridge for the price of a bluray disc. I've seen bluray films (poor one's) designed to retail at about £1 and must cost in the region of 20p or so to make. If flash memory or any type of random access memory was as cheap as that it would be a revolution in hardware prices. Why the hell is the Switch fitted with so little storage memory. The maximum size of launch Switch cartridges appears to be 8GB (64Gb) and 16GB (128Gb) but most games appear to be on much smaller cartridges.

Do you have a link for this 32GB equals bluray manufacturing cost. How much is a writable bluray disc because they normally cost more than pressed blurays and still pretty cheap. 

My guess for a bluray disc duplicating cost would be about 30c/25p and for a cartridge by Macronix lets say 8GB and 16GB would be $3 and $4.50.  However most Switch games seem to be on very small capacity cartridges so probably somewhere around $1-2.

Third parties seem to be highly motivated to bring small games to Switch. Hence ports of old small games or simple games like Bomberman. It seems third parties are highly motivated to keep their game sizes very small to lower their cartridge costs. There is absolutely no indication of low cartridge costs.

xbox one is like 80watts? or so to reach 1.4 teraflops.

Switch is like 20? watts.

Its just not possible.

The thing is, the document says 1024 Flops. So I can't really assume at this point. I kinda doubt switch would be .5TF, we've heard .75 from digital foundry. The document is confusing. It really doesn't specify if it's a Megaflop, Gigaflop or at FP16 or 32 but you are most likely right. 

It could be 1024Gflops and be 1TF FP16, or 1024Gflops FP32, aka 512TF. But do hardware manufacturers default to FP16 when noting specs?

I think the idea is to optimize software for FP16 so that you really do have 1TF worth of compute power.

If you look at spec sheets provided by nVidia for Tegra, they list both FP16 and FP32 tflops. We're used to seeing FP16 / 2 = FP32 flops, but that isn't always the case. The Tegra K1 only had ~350 gflops in either FP16 or FP32 modes because the Kepler architecture couldn't even leverage FP16. Starting only with Tegra X1 did they do anything to take advantage of FP16.

People have been reeeaaally quick to write off FP16, but seldom give you valid, scientific reasons why.

I think the idea is to optimize software for FP16 so that you really do have 1TF worth of compute power.

If you look at spec sheets provided by nVidia for Tegra, they list both FP16 and FP32 tflops. We're used to seeing FP16 / 2 = FP32 flops, but that isn't always the case. The Tegra K1 only had ~350 gflops in either FP16 or FP32 modes because the Kepler architecture couldn't even leverage FP16. Starting only with Tegra X1 did they do anything to take advantage of FP16.

People have been reeeaaally quick to write off FP16, but seldom give you valid, scientific reasons why.

Not expert on subject, but few things to consider:

"To get an idea of what a difference in precision 16 bits can make, FP16 can represent 1024 values for each power of 2 between 2^-14 and 2¹⁵ (its exponent range). That’s 30,720 values. Contrast this to FP32, which can represent about 8 million values for each power of 2 between 2^-126 and 2¹²⁷. That’s about 2 billion values—a big difference."

https://devblogs.nvidia.com/parallelforall/mixed-precision-programming-cuda-8/ 

Current performance king in gaming GPUs, Titan X:

FP32: 10,157 GFLOPS

FP16: 159 GFLOPS

This is of course nVidia's way to prevent gaming cards being bought instead of Teslas for tasks that actually benefit from FP16...but just shows how little FP16 performance is important in games.

Honestly, not an expert on the subject, but last time I recall any talk about FP16 in gaming was some 15 or so years ago. Sure, mobiles have it, but degradation in quality seems to be quite noticeable.

Because people don't make the distinction between rendered resolution and output resolution.