Yes, it is very low margin. And it is also very demanding. But is it high volume. Look at the companies that walked away from this business, Freescale and TI. Intel has a lot of cash, but this may be more than they have the stomach to invest in.
This is trench warfare ( attrition ), Intel has to get into smartphones and tablets not to make same level of profit as for CPUs but to deny upstarts ARM and TSMC future viability. Even if it takes $ 20 billion and 5 years to get there it would be well worth it for the future of Intel. They have the money in the bank for such a strategy.
Technologies like finFET and low power dissipation transistors and architectures are the foundation for it. Once Intel brings out Medfield at 20 nm later this year, it will be very hard for system builders now tied to fabless processor vendors ( who get their chips made by TSMC, now 2 nodes behind Intel ) or even integrated vendors ( Samsung ), to ignore it.
The consumer gets to choose who's the winner. If Intel doesn't fail to deliver a comparable power saving product to any ARM based product, Intel might have a chance. Otherwise, how many people are going to buy it? Another factor would be the availability of apps and, in particular, free apps. I am happy to see more competition. At the end, consumers win. ;)
It's a lower margin business. It made no sense to use advanced technologies for the purpose of lower power consumption. 22 nm nodes' will be at least 40% more expensive than 32nm and 14 nm would be even much more when EUV litho kicks in. They should run the PC/server CPUs on new technology nodes and use the older generation technology at depreciated fabs to run lower margin products. Why can't Intel push the design architecture to get lower power same as ARM?
All 4 core APU for smartphone will start shipping using 28nm 2012. it still can't handle the full-HD 3D graphic. Smart phone application is still pushing for better & faster chips even in low power application. Your cost analysis is only right for foundries' node migration. It doesn't necessary right to compare Intel 22nm vs foundry's 32nm. Intel delayed two generation immersion litho (compared to foundries) introduction just to save cost. Intel does follow Moore's law in perfomance/cost/schedule so far. and multi-patterning is required in too many critical layers from 28/22nm in foundries, you don't need to be in 14nm.....
There are two monopoly words used in this article aimed at Intel. While ARM monopoly in the wireless sector is rewarded with words such as "maintain leading position". This aritcle is not meant to inform readers, but rather, to pre-empt Intel in order to pave the way for future lawsuits from ARM and AMD against Intel. I buy and support Intel products because of the jobs issue. The jobs created in the US from ARM and AMD is a drop in a bucket compare to jobs that Intel created. If you want to support jobs growth in the US, buy tech. products that contain Intel chips.
The article is talking about Intel getting into the wireless market, but not wireless technology; Intel is going to be in the handhelds and tablets that traditionally have been ARM based. I can't say what the power/performance tradeoff numbers are between Atom and the new/future generation of ARM cores, but I have an Atom based laptop that is not very impressive (as a laptop mind you). I think if the OS and system design delivers speed and long battery life then Intel will have a winner, if not then they will not achieve the success they aim for.
Blog Doing Math in FPGAs Tom Burke 23 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...