How Intel Lost the Mobile Market, Part 2: The Rise and Neglect of Atom

how-intel-lost-the-mobile-market,-part-2:-the-rise-and-neglect-of-atom

This case might per chance per chance also do affiliate commissions from the links on this page. Terms of employ.

Update (6/1/2020): The article below might per chance per chance even have been written in 2016, but it surely mute stands up as a postmortem of what went rotten with Intel’s cell efforts — with one crucial omission. Aid in 2016, we didn’t know Qualcomm had been ruthlessly implementing licensing and buying phrases that made it successfully very now now not going for producers to present Intel-basically based completely cell gadgets. I be aware wondering why Intel couldn’t ranking a single US firm to build a cell phone around its hardware platform for love or money when the common Xolo X900 compared properly ample against a then-newest iPhone.

Intel mute made a wide range of errors with Atom, as this text discusses, but the reality that Qualcomm had a stranglehold within the marketplace gradual the scenes obviously had an impression on what roughly success Intel became once ever going to enact.

I in fact beloved the Xolo X900 machine I tested all these years within the past, and the Bay Path tablets I had circa 2013 were enormous gadgets. Atom’s cell efforts will the least bit times dwell a attention-grabbing might per chance per chance also-have-been.

Well-liked myth below.

In Share 1 of this two-half series, we mentioned the adaptation between Intel and TSMC’s foundry items and the plan these variations made it extremely sophisticated for Intel to compete in cell. In Share 2 we’ll uncover the squawk decisions Intel made, the upward push and neglect of Atom, and why the firm’s appropriate foundry technology wasn’t ample to conquer the market.

The same old clarification for why Intel lost the cell market is that its x86 cell processors either drew too powerful energy or weren’t great ample compared with their ARM counterparts. Intel’s decision to promote its ARM division and XScale processor line in 2006 has been broadly derided as a principal error. It’s a easy, same old-sense clarification with impartial correct one flaw: It errors symptoms for motive.

All of This Has Took situation Sooner than

Intel’s struggles within the cell market didn’t open with Medfield, Moorestown, or even the choice to promote its ARM enterprise and XScale chip division ten years within the past. As EETimes reported in 2006:

Intel (Santa Clara, Calif.) spent bigger than $10 billion to enter the communications enterprise over time, but the microprocessor huge lost its shirt — if now now not millions of bucks within the field. The reported communications-chip sale is speculated to be half of Intel’s idea to overtake the firm. Intel is also arena to consist of the layoff or redeployment of 16,000 workers, in conserving with hypothesis from one Internet situation.

Trade “communications” to cell, regulate the selection of fired workers, and that paragraph might per chance per chance’ve been written on the present time. Intel’s complications in cell aren’t original; Santa Clara has been struggling to enter original markets for nearly 20 years. Various articles from 2006 emphasize that XScale gross sales had been fairly low, as had revenue from Intel’s networking and communications division.

From Intel’s standpoint, promoting XScale made sense. Constructing a cell processor enterprise around ARM cores would have dinky Intel’s ability to leverage its own IP and experience in x86 manufacturing, while concurrently cutting into its profits (Intel would have owed principal royalties to ARM if the type of compose ever turned into standard). Atom became once already properly into fashion in 2006 and Intel determined to wager on its own hardware experience and power fashion abilities.

Atom and the Rise of x86 All over the save the save

Opposite to standard belief, Intel wasn’t caught utterly off-guard by the upward push of smartphones or the reputation of diminutive, Internet-linked gadgets. Atom fashion started in 2004; the Silverthorne core that Intel debuted in 2008 had a TDP of wonderful 2-3W at a time when most cell Core 2 Duo processors were stuck in 35W territory.

Intel-Penny

Intel’s common Atom.

Most other folks be aware Atom as the chip that launched a thousand netbooks, but that wasn’t Intel’s common idea. The firm concept nettops and netbooks might per chance per chance be a definite phase market for Atom, now now not the chip’s most principal platform. Atom and its successors were supposed to open an armada of Mobile Internet Devices, identified as MIDs.

Gigabyte-MID

Devices love the Gigabyte MID M528, confirmed above, uncover hopelessly quaint on the present time, but Intel became once clearly fascinated with the formula forward for cell computing. The firm envisioned an ecosystem of netbooks and MIDs pushed by its own personalized x86 structure, a plan the press dubbed “x86 in each save.”

Conflicting Priorities

There have been clearly executives at Intel who understood how serious cell might per chance per chance be to the firm’s long-time frame future and pushed for aggressive positioning and product ramps. Unfortunately, these efforts were stymied by others who were enthusiastic about the impression Atom and the low-mark gadgets it became once alleged to enable would have on Intel’s most principal enterprise. MIDs and later netbooks were alleged to be bare-bones, low-mark gadgets, precious as secondary machines and for same old projects, but no extra.

IntelSoC

Intel became once engaged on Atom SoCs in 2008, but its first fully unified chip wouldn’t ship unless 2012.

Intel’s put up-open perspective against Atom is most attention-grabbing summarized as benign neglect. Whereas the chip went thru plenty of revisions to combine ingredients and minimize costs, Intel refused to commit the resources that might per chance per chance per chance have made Atom a most attention-grabbing-in-class participant within the cell market. From 2008 to 2013, Intel launched a mark-lowered model of its Nehalem structure, the Westmere 32nm die shrink, a brand original structure with integrated graphics (Sandy Bridge), a high-cease fanatic platform (Sandy Bridge-E), a brand original 22nm CPU with FinFET technology (Ivy Bridge), every other architectural refresh (Haswell), and a 2nd-generation fanatic platform (Ivy Bridge-E). That’s two paunchy tick-tock cadences for Intel’s gigantic-core enterprise, while Atom didn’t even compose the jump to 32nm unless 2012. Its single architectural refresh to this level arrived in 2013, impartial correct after the open of Ivy Bridge-E.

No subject being at the start starved for resources, 32nm Atom chips were aggressive within the midrange cell market. With Medfield, Intel seemed as if it might per chance most likely per chance per chance have turned into a corner, but the firm’s designs on the entire did now not ranking powerful traction available within the market. Handiest Intel’s contra-revenue approach won the firm principal tablet market share, and these gains were most attention-grabbing sustained thru heavy monetary losses.

intel-mobile-chart

Intel’s cell and communications revenue and losses from Q1 2013 thru Q2 2014.

Atom wasn’t the jam — Atom became once the solution Intel didn’t have the guts to chase.

The Critical Decisions Intel Didn’t Abolish

Intel did now not compose traction in cell since it wasn’t bright to trouble upsetting the economic model that had transformed it into a titan of computing. The firm’s fabs, manufacturing strategies, and resources were geared against big, costly processors, now now not churning out astronomical numbers of low-mark cell cores. Prioritizing Atom over Core would’ve required the firm to retool as a minimum some of its fabs to stress throughput and lower costs in notify to compete with the ARM processors built at Samsung and TSMC. It might most likely per chance per chance per chance’ve meant lower noxious margins and much less revenue per unit sold.

Intel's Atom Roadmap

Intel tried to inch issues up within the kill, but each its 14nm job and the next-generation Goldmont CPU core were delayed.

Intel did snatch steps to make stronger its aggressive standing vis-a-vis ARM and ARM’s foundry companions, but it surely infrequently took them fleet and continually did now not take a study thru. Intel sold Infineon Wi-fi in 2011 for $1.4 billion, but to this day all of its publicly launched wireless products, in conjunction with the XMM 7480 modem, are mute built on 28nm at TSMC. Smartphones and tablets have the least bit times mature SoCs, but Intel didn’t open its first Atom-basically based completely SoC unless 2012 — five years after the iPhone launched and four years after Atom’s own debut.

SoFIA integrated Atom SoC, details

The SoFIA partnership with TSMC raised eyebrows, but now now not revenue.

One ingredient we want to stress right here is that Intel’s decision to present protection to its core (Core) enterprise and product margins might per chance per chance even have been rotten, but it surely wasn’t crazy. Refitting fabs, building experience in SoC compose, and porting modems from TSMC would have required big cash infusions and snatch principal amounts of time. If Intel had launched Atom with an aggressive idea to position the chip in smartphones by 2010, issues might per chance per chance even have conducted out very in every other plan. By the level the firm wakened to the threat it confronted from ARM and merchant foundries, it became once too unhurried to compose up the gap.

Why Intel’s Foundry Tech Couldn’t Establish Its Mobile Trade

Intel’s job technology leadership couldn’t do the firm’s cell division since it wasn’t designed to enact so. Smartphone and tablet OEMs crucial gadgets with integrated LTE radios; Intel didn’t have them. Even the SoFIA partnership with TSMC never came to market, it sounds as if because Intel couldn’t loyal ample quantity to kickstart manufacturing.

Intel’s 14nm complications delayed its next-generation tablet processors from 2014 to 2015. Its 10nm node, once expected to loyal gigantic economies of scale over TSMC, has been pushed to 2017 as properly. I don’t ponder these delays conducted an gigantic role in Intel’s decision to leave the cell market, but they might per chance per chance per chance even have influenced it. In 2012, Intel mute expected to be on 10nm by 2016 with EUV (crude ultraviolet lithography) ramping against paunchy manufacturing. The now-canceled 450mm wafers weren’t expected in-market fairly this soon, but Chipzilla expected to ogle principal mark savings from involving to the elevated wafers within the 2018 to 2020 timeframe — mark savings that might per chance per chance per chance’ve extra improved its standing against Samsung, TSMC, and GlobalFoundries.

If Intel had begun reorienting against Atom when it launched the chip in 2008, it might per chance most likely per chance per chance also’ve weathered these delays and cancellations with out powerful anguish. Failing to enact so left its beleaguered cell enterprise going thru higher-than-expected costs and minimal revenue.

Intel didn’t lose the cell market because Atom’s efficiency and energy consumption didn’t compete with ARM; examine and evaluation showed that Atom became once able to matching ARM efficiency in a couple of market segments. It lost the cell market since it didn’t compose the changes that might per chance per chance per chance have allowed it to compete on mark with products manufactured at TSMC and Samsung. The exacting principles and uncommon layouts that drove Intel to the give up of one market might per chance per chance now not be with out complications tailored to others, and Intel became once unwilling to trouble its situation on the give up of the mature x86 market for a unhealthy payoff in cell. There’s no proof that conserving XScale or rising ARM products would have modified that — if anything else, the ARM division would’ve been below powerful extra tension to compose definite that it never turned into a threat to the x86 enterprise.

Now learn:


Leave a comment

Your email address will not be published.


*