Will Smith (will.smith.insights@gmail.com) on 1/12/11 wrote:
---------------------------

[...]

>All this points in one direction: An selected number of very able companies (NVIDIA
>confirmed, Apple rumored, there might be others in far east with quite some probability)
>are in the process of 'massaging' the most modern ARM cores to optimally fit netbooks,
>tablets and mobile devices - plus they add graphics capabilities to the very same
>chip. The term 'fit' in this context means low power, very low cost at just the right performance and functionality.
>
>In addition, the corresponding ecosystem is rapidly expanding, adding OS support
>to the already wide range of on-chip functional building blocks.
>
>Add all of that up and you get a game changing disruption - it will feel like an
>earthquake IMHO. Why? This will open up the field considerably towards a new competition
>for everything between a mobile phone and a mid-range PC. It will give system integrators
>a way to differentiate, and it will give truly fabless companies a way to enter
>the CPU/GPU markets. Wow, no longer an Intel play with a few Intel-tolerated players! Exciting future!

Obviously, ARM is the architecture of choice with devices having the right mix of features, performance, efficiency, and licensing at the right time... It benefits from its long legacy in the phone market; But I think ARM is riding the smartphone/tablet wave, and not the other way around.

I contend that ARM, having the right chips at the right time, made it possible for the smartphone ecosystem to be born, but now it's the ecosystem that's driving ARM up, and not the other way around... And there's room for another player besides it and x86.

Consider this question: Why did MS decide to port Windows to ARM?

While Apple popularized both the smartphone and the tablet in their contemporary forms, they are closed hardware platforms and are not likely to run Windows or any other software platform besides what Apple dictates. Ever. That they are ARM-based is incidental: Microsoft did not decide to port Windows to ARM because of Apple's use of it, and the companies thrusting ARM development ahead are not vying for Apple, but rather for A) their own in-house products, or B) to sell as parts to MID OEMs.

There is no roll-your-own with Apple, so if it were JUST Apple, and Apple gained sufficient traction to scare MS because there were no x86 designs that could touch ARM, then MS would have had to tap/bless an alternative architecture of their own as a countermeasure, and either develop the HW on their own, or wait for OEMs to do so. ARM? Maybe. Maybe probably.

But as it happens...

Android is a software platform that drops the barrier of entry into the MID market to not much more than the cost of developing the hardware. And ARM currently takes a huge bite out of the cost of hardware development, with A) licensing for those with the ability to roll their own value-added chips, and B) with the SoCs available and in development for those who just want to go the COTS route.

MS knows these factors will result in the creation of devices that encroach upon MS's Wintel bread&butter. MS feels they cannot afford to be excluded from this new, energetic market of sexy products; thus, Windows on ARM, because MS (rightfully) sees ARM as the alternative HW platform of choice at the moment.

But MS is merely reacting to a situation that's out of its control... The real threat is that the barrier to entry into Wintel territory has been lowered for both hardware AND software. And I'd posit that the software part is more important than the hardware part. The de facto standard platform supporting the now-solid base of Android software forced MS's hand in the choice of MID architecture... ARM had gained its momentum before MS panicked.

But Android is still compatible with multiple hw architectures... It already has ARM and X86... And MIPS.

Android provides a level playing field for ARM, x86, and MIPS to compete on. I don't know what the fraction of apps that are java-only vs. java+native, but Android already has to deal with that issue wrt ARM and x86... Creating infrastructure to deal with multi-arch issues only needs to be done once.

And MIPs is better-prepared than ARM to move (return, actually) upstream... It has been to the top of the mountain, so to speak; supercomputers, workstations, PCs, embedded... While a little late to the smartphone/tablet market compared to incumbent ARM, products are in the pipe and will have Android support right out of the gate. (Google "mips smartphone tablet")

Would you care if your Android tablet had a Tegra-like MIPS SoC, as long as it A) ran the apps B) with equivalent performance?

It's going to be a very fun year...

fpg