Simply redesigning the architecture is no longer enough! You can harp on design flexibility all you want but we’re fast approaching the point that it’s becoming a diminishing factor and advancing the underlying technology is becoming the dominant concern.

While also ignoring how the ARM market works depends on fairly rapid end of life cycles of ARM based products. So anything that can slow that down is hardly insignificant to the ARM markets.

Add the increasing demand for performance, which ARM is only starting to be able to provide puts more emphasis on how far they can push the technology and only cutting edge then counts, which brings it back to my point.

As it stands, Intel stands to get over a year head start over everyone else in progressing to the next nm reduction and dealing with the issues of each reduction faster to get products out to market and all that adds to market momentum that will increasingly make it harder for ARM to keep up.

Keeping ahead of ARM in this way means they can counter the normal advantages of ARM’s simpler designs and normally lower costs.

Present schedule, assuming no further delays, for ARM is to go 28nm this year, but some companies are already stating they can’t do it yet because there is still too much power leakage to make it worth it. So the progress is already slowing!

Even if it wasn’t they won’t go 20nm for even low end applications till 2013, when Intel will already have it’s full lineup under 22nm, and 20nm for high end applications that would even compete with Intel’s idea of low end offerings won’t come out until 2014. But by then Intel will be going 14nm already for again it’s entire lineup.

AMD too is experiencing delays, so using similar facilities obviously doesn’t help, with it’s 28nm update and like many in the ARM camp won’t have much out till the later half of this year.

While some ARM solutions may not even come out till early next year. Especially those waiting on some of the next gen parts besides just the ARM cores.

Now while other components are progressing at different rates from the CPU core, that’s another reasons Intel has an advantage as they keep most everything in house and can progress those other parts faster as well.

So don’t underestimate them!

ARM Companies like Broadcom also has their own MIPS64-implementations.

Won’t be general purpose stuff but it’s quite a bit away from that happening on ARM too.

As most ARM-players use TSMC and their physical IP not much is actually 32nm which TSMC doesn’t produce at all. It’s mostly TSMC 40G. 45nm at some others. You can still make an awful a lot of design choices. Including configuration of FP/SIMD-units and everything not in the actual processor as always. Of course Intel will have some design and fabrication lead, if they choice to, but that don’t really mean Atom will be bleeding edge. 32nm on none-released stuff isn’t that impressive, for that matter Samsung does 32nm on their Exynos-series already that is coming out this year too, so we shouldn’t draw any big conclusions mobile SoCs hasn’t been the first applications to be shrunk at TSMC or other fabs. Neither is there much need. GF does 28nm on ARM too using the same equipment as AMD. TI’s latest designs will of course be 28nm too. It’s pretty bleeding edge in their fields.Intel won’t go 22nm until 2013 sometime here, so at most it’s proven 28nm designs against a new 22nm one. Most don’t have massive 28nm lines up and running trouble free yet, dosen’t mean it is not a good design target when products are rolling out in new consumer products. ARM Ltd themselves are already working towards 14nm but as said it’s not the only resource it that field. They have the rest of the players working for them, it’s not like Intel has all manufacturing capabilities.

However, none of that lets them deal directly with things like power leakage, which goes down to the individual components like transistors, and other similar issues as they continue to shrink the manufacturing process and have to overcome the issues of going that small causes.

Scaling the same architecture down only works so far before you have to make fundamental changes in what is being used in that architecture.

So the ARM business model doesn’t make it easy to make fundamental changes needed to make it easier to continue shrinking the manufacturing process without also making significant changes to the underlying technology being used.

What ARM is good at is allowing for flexible use of its technology, which allows for easily simplifying of designs, lowering costs, optimizing overall design efficiencies, scaling as needed, adding or removing components as needed, but that’s about it and they’re starting to reach the limits of that benefit.

While collaborating with ARM does help and manufacturers can make some changes on their own and together they’ll eventually deal with such issues as I’m pointing to, but my point was that Intel has a distinct advantage as they don’t have to waste time with any such back and forth when they have full control and can immediately make changes as needed and can even do all R&D in house.  So they can make those changes faster and it’s more in line with Intel’s business model than ARM’s to do so more consistently.

Examples of companies delaying going 28nm and still producing 32nm because of things like power leakage issues, which they specifically said was the issue, shows this is a growing problem for ARM.  Meanwhile Intel is heading towards 22nm and will go down to 14nm in 2014, which is more rapidly than either ARM or even AMD are making their own similar moves to smaller FABs.

Also don’t forget ARM is still 32bit, they may have started employing things like 64bit memory bus but they’re still years away from even sample full 64bit products and that’s another stumbling block they’ll have to deal with as they start to compete more directly with x86 products.

ARM Ltd doesn’t work directly with these companies but instead just licenses their technology out to them.  While many of these companies also don’t focus on just the manufacturing of ARM chips but also many other products and all of that means less flexibility on the Fabrication side of things.  Limiting them to what they are set to manufacture and what technology they can license.

This is where Intel can start getting an advantage because they have far more control over their fabrication and can more readily develop their own technology instead of licensing.  Factors like dealing with power leakage as the fabrication gets continuously scaled down is something Intel can deal with better than many of these others working with ARM.

The normal ARM strengths of being able to be more simply designed to reduce production cost and energy usage are starting to reach their limits and it’s the fabrication process that’s starting to become a more pertinent factor.

Some companies are already delaying the move to 28nm because of concerns over power leakage and production efficiency with ARM and is why we’re still seeing 32nm and larger FABs still being used.

Though on the other hand there are a lot of companies working together on ARM and that’s a significant factor as well but it remains to be seen whether the market fragmentation out weighs its strength in numbers.

22nm Silvermont will put all ATOM’s under SoC and Haswell will be the SoC version of the upcoming Ivy Bridge, both coming out 2013.  So hardly exclusive to Apple, never mind ARM has been making SoC’s for longer than Apple has joined them.

Apple just licenses and designs, leaving the actual manufacturing to companies like Samsung.  Just like most other companies that don’t have their own factories.

Regardless, what’s mainly differentiating one generation from the next is still the processor!  While SoC’s also still do not necessarily include everything like Storage drives, WiFi/3G/4G, etc that still varies per system.  So it’s fine to call them processors, when that’s the main part being referred to.

Apple develops what are conventionally referred to as “systems-on-a-chip”. The A5 is not a processor, it is a “system-on-a-chip”. Wikipedia!