A year and a half ago I wrote an entry on here regarding Intel in the mobile space. The argument was basically that Intel was finally getting their stuff together, and the market had gotten ready for Intel and x861 (as well as x86_64) to be a fully supported platform.
From Unity to the NDK to AVDs, Intel is now a first-class platform on Android.
But the industry runs at a very different cost and profit model from what Intel was accustomed. The highest-end ARM SoCs run from $30 – $70 per unit. Intel has long lived in a world where their solutions net hundreds to thousands of dollars per unit. But the market changes, and the ARM world isn’t going away if Intel just looks the other way.
Yet Intel seems to have just killed off their aspirations for the market. Their intentionally sabotaged Atom solutions are being bested by small competitors, and they can’t make the finances work.
Bizarre. I find it hard to believe, especially given that Intel has made significant noise about targeting the IoT market. I think the conclusions that people are drawing about Intel killing off the mobile Atom devices and a noncompetitive radio chipset — concluding that Intel is crawling back into their desktop and server processor shell, ceding defeat — highly unlikely.
More likely, I would guess that Intel is going to follow Nvidia’s lead, as there’s no way they’re simply giving up on mobile devices. Nvidia once had separate mobile and desktop engineering, with the duplicated costs that entailed, but with their Maxwell chipset the same designs, architectures and processes are used on both sides of the fold.
I expect Intel to pursue the same approach, simply scaling up and down their common contemporary core to all needs. There are Skylake processors available right now with a TDP of 7.5W (which is the going range for tablet SoCs). Core M processors with a TDP below 4W. The Atom processors didn’t serve a particular need beyond being sabotaged just enough that they didn’t threaten the more expensive markets. That approach doesn’t work anymore.
1 – As an aside, it’s impossible to discuss x86(_64) without someone confidently announcing that it’s a derelict bad design that deserves to die, etc, carrying on an argument from literally the late 1980s to the early 1990s. This betrays a general ignorance about the state of x86_64 vs ARM64, or the enormous complexities of modern ARM chips (with absolutely staggering transistor counts). They’re both great solutions.