Intel's smartphone platforms

Theoretically K1 can be downclocked for a smartphone platform; the trick question is then how much exactly and then you can define the difference. At similar GPU frequencies the difference isn't obviously "far behind"; power consumption being of course the bigger and most important unknown for both platforms.

 

ROFL if true. If it ends up in a =/< mainstream Samsung smartphone clocked at maximum 1.6-1.7GHz as the author states then it's most likely for real.

 

If anything, it'd be the CPU that'd need the down clocking. Four of those powerhouse cores is overboard. A two core CPU config like Merrifield plus Moorefield's G6430 would be good, or Moorefield with the above-mentioned reduced clocks and full speed GPU could make for a winning combo.

 

First that would have to worry is Mediatek; its tablet MT8135 flagship and smartphone MT6595 flagship SoC contain each a G6200@600MHz. If all the above is true and Samsung should leave the G6430 in Moorefield at its peak 533MHz, it'll end up with a roughly twice as fast GPU than the two MTK SoCs.

 

The H2 2014 seems more promising than ever for anyone looking at midrange phones.

With Mediatek offering a 4*Cortex A17 + 4*Cortex A7 big.LITTLE combo and Intel 2/4 Silvermont cores with LTE, and many chinese manufacturers also starting to use the same Exmor sensors from Sony, it'll be increasingly hard to justify the price difference for a flagship with a Snapdragon 801/805.

Moto G has already shown how a pure AOSP can perform using just 1.2GHz Cortex A7 cores and an Adreno 305. The Xiaomi Hongmi with the "old" MT6589 is also an excellent performer with MIUI.

Give this range of phones a 2GHz Cortex A15 or Silvermont and they'll fly. No wonder nVidia got off the midrange smartphone business. This is all pointing to a bloodbath for Qualcomm.

 

Samsung makes a lot of phones using a lot of SoCs, an Atom-based one was pretty much inevitable. Especially since they already made a tablet with CloverTrail+.

I'd go so far as to say that any single Samsung phone outside of a flagship like Galaxy S or Note isn't a huge deal by itself. If it shows up in a broad range of phones, that's a huge deal.

 

What is Intel's baseband story?

May be hard to win business away from Snapdragon without being able to offer the whole package.

If they're going to offer SOCs at cost, they might as well rent out their fabs to Apple or some other big SOC manufacturer, as has been suggested. That would be more profitable than giving away SOCs at cost.

 

I think I would want something with fast single-threaded performance, being interested in Firefox OS. So a dual core (at most!) Cortex A12, A15 or A17 or Atom or even MIPS, don't care about the GPU besides it or its driver not crashing, while being as cheap and low power as you can get away with.

 

I agree that a 2+2 big.LITTLE approach would've made the most sense, but if the cores can completely power down then it's not like it hurts to have them in there.

Regardless, Moorestown should be your choice.

But Firefox OS? Who wants that?!

 

That'd be off-topic but basically, I don't have Android devices anyway and I'd be worried about the security (there's even a recent announcement that Android will give broader permissions to applications)


Incidentally there's one phone where you can choose between Android and FF OS, that uses an older gen Atom 1.6GHz (Clover Trail)
http://www.pcmag.com/article2/0,2817,2454008,00.asp

 

Makes perfect sense, given their increasing efforts on their iGPUs.

But my opinion was pretty much discarded in this forum when I suggested this over a year ago..

 

In 2012 there weren't any rumors either that Intel might eventually consider to abandon the smartphone market entirely in the future if the tide doesn't change in their favour.

 

1 - So there was a chance that Intel would forever keep spending R&D and fab resources in smartphone chips if they never got any market traction ever?


2 - What does that have to do with selling IMG stock and starting to use their own GPUs in smartphone SoCs?

 

Their outlook and expectations were obviously quite a bit higher in early 2012, than what they gained in the meantime.

Nothing directly; they didn't see so far any land whether with or without 3rd party GPU IP, which would suggest that the faults were probably elsewhere. You know strategies have that nasty tendency to change according to fluctuating market tendencies which firms usually adjust to.

Intel's ULP SoC GPUs have always been too little too late and I hope they'll change it in the longrun irrelevant of what kind of GPUs they'll use. Neither Moorefield nor Baytrail (Android tablet variants and such) turn any heads for the GPU compared to competing solutions and Baytrail ironically even less.