I noticed something interesting in an article about the PowerVR architecture. The PowerVR 6 series has twice as many 16-bit float units than 32-bit float units, while the K1 has only 32-bit units. Now, if you write shaders for PowerVR you will notice that using lowp variables in your shaders makes it go a lot faster. The reason is, of coarse, because lowp (and mediump) use 16-bit floats. But speed isn't the only thing - the 16-bit float units also use less power. So, this power saving ability is simply not available at all on the Tegra K1.

Looking at a few things...

First, most mobile developers don't focus their life on one specific chip. They get to target all mobile devices. So you don't get to target the latest high-end chip, but instead set your minimum spec to something that was released two or three years ago.

Second, the article points out that the chip they describe won't be on the market for another 18 months or so, while the Tegra k1 you compared it against is in use in a bunch of stuff right now. Declaring that 2015 hardware might be better than 2013 hardware, .... well, it isn't really exciting news.

Third, there is so much very valuable missing information that is missing from the release, that it is nearly useless. Sure, they reveal that one part of the chip will have some attractive parts, but then limit the announcement only to the attractive bits without revealing all the important bits. That is common marketing hype.

Fourth, if you look at the "theoretical limits" comparison on the last page, you'll see that they list two PowerVR chips, one a mainstream chip and the other a high performance chip that compares almost identically to the K1. That's pretty good because it means competition in the marketplace, but also not so great if you look at the mainstream chip details. They're a whole lot of "meh". Those won't be groundbreaking devices, and since those mainstream chips are also coming out in 18 months or so, we'll need to support that kind of thing until around 2018 or later.

It appears the chip will be competitive as a high end mobile chip, but I don't see anything revolutionary about it. A few mobile device programmers may get to write code to take advantage of it, most people will just continue to write code targeting the average consumer device.

Some mobile users may not target a specific chip, but nearly all of them use lowp and mediump specifiers.

Also, you got it the wrong way around, Tegra K1 is not currently in any devices, while PowerVR is. (I think the 16-bit float units are present in 5 series as well)

"In which case Series6XT equipped SoCs would start appearing in 2015, likely in the latter half."

That is about 18 months.

In contrast, K1 reference devices are circulating today.

Don't get me wrong, competition between the hardware makers is a good thing for the industry. It's just that while PowerVR is a great brand and everything, the chip is both in the distant future and also not something developers can take direct advantage of any time soon. But declaring an announced chip with an estimated release date in over a year to be superior to today's chips, well, yeah, that's really to be expected.