Intel atom phone: Intel’s Last Atom in Smartphones: A 2018 Benchmark

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

• By Joel Hruska on December 3, 2020 at 3:03 pm

This site may earn affiliate commissions from the links on this page. Terms of use.

Update (11/25/2020): The article below may have been written in 2016, but it still stands up as a postmortem of what went wrong with Intel’s mobile efforts — with one very important omission. Back in 2016, we didn’t know Qualcomm had been ruthlessly enforcing licensing and purchasing terms that made it effectively impossible for manufacturers to offer Intel-based mobile devices. I remember wondering why Intel couldn’t find a single US company to produce a phone around its hardware platform for love or money when the original Xolo X900 compared well enough against a then-current iPhone.

Intel still made a number of mistakes with Atom, as this article discusses, but the fact that Qualcomm had a stranglehold on the market behind the scenes obviously had an impact on what kind of success Intel was ever going to achieve.

I genuinely liked the Xolo X900 device I tested all those years ago, and the Bay Trail tablets I had circa 2013 were great devices. Atom’s mobile efforts will always remain an enticing might-have-been. The US 3rd Circuit Court of Appeals has since thrown out the antitrust findings against Qualcomm. It should be noted that the FTC disagreed with that decision. Intel continues to manufacture 5G base stations, but the scope of its 5G business has shrunk significantly after selling its 5G modem to Apple.

Original story below:

In Part 1 of this two-part series, we discussed the difference between Intel and TSMC’s foundry models and how these differences made it extremely difficult for Intel to compete in mobile. In Part 2 we’ll explore the specific decisions Intel made, the rise and neglect of Atom, and why the company’s superior foundry technology wasn’t enough to conquer the market.

The common explanation for why Intel lost the mobile market is that its x86 mobile processors either drew too much power or weren’t powerful enough compared with their ARM counterparts. Intel’s decision to sell its ARM division and XScale processor line in 2006 has been widely derided as a critical error. It’s a simple, common-sense explanation with just one flaw: It mistakes symptoms for cause.

All of This Has Happened Before

Intel’s struggles in the mobile market didn’t begin with Medfield, Moorestown, or even the decision to sell its ARM business and XScale chip division ten years ago. As EETimes reported in 2006:

Intel (Santa Clara, Calif.) spent more than $10 billion to enter the communications business over the years, but the microprocessor giant lost its shirt — if not millions of dollars in the arena. The reported communications-chip sale is said to be part of Intel’s plan to overhaul the company. Intel is also set to include the layoff or redeployment of 16,000 employees, according to speculation from one Web site.

Change “communications” to mobile, adjust the number of fired employees, and that paragraph could’ve been written today. Intel’s problems in mobile aren’t new; Santa Clara has been struggling to enter new markets for nearly 20 years. Other articles from 2006 emphasize that XScale sales had been fairly low, as had revenue from Intel’s networking and communications division.

From Intel’s perspective, selling XScale made sense. Building a mobile processor business around ARM cores would have limited Intel’s ability to leverage its own IP and expertise in x86 manufacturing, while simultaneously cutting into its profits (Intel would have owed significant royalties to ARM if such a design ever became popular). Atom was already well into development in 2006 and Intel decided to bet on its own hardware expertise and software development skills.

Atom and the Rise of x86 Everywhere

Contrary to popular belief, Intel wasn’t caught completely off-guard by the rise of smartphones or the popularity of small, Internet-connected devices. Atom development began in 2004; the Silverthorne core that Intel debuted in 2008 had a TDP of just 2-3W at a time when most mobile Core 2 Duo processors were stuck in 35W territory.

Intel’s original Atom.

Most people remember Atom as the chip that launched a thousand netbooks, but that wasn’t Intel’s original plan. The company thought nettops and netbooks would be a niche market for Atom, not the chip’s primary platform. Atom and its successors were supposed to launch an armada of Mobile Internet Devices, known as MIDs.

Devices like the Gigabyte MID M528, shown above, look hopelessly quaint today, but Intel was clearly thinking about the future of mobile computing. The company envisioned an ecosystem of netbooks and MIDs driven by its own custom x86 architecture, a goal the press dubbed “x86 everywhere.”

Conflicting Priorities

There were clearly executives at Intel who understood how critical mobile would be to the company’s long-term future and pushed for aggressive positioning and product ramps. Unfortunately, those efforts were stymied by others who were concerned about the impact Atom and the low-cost devices it was supposed to enable would have on Intel’s primary business. MIDs and later netbooks were supposed to be bare-bones, low-cost devices, useful as secondary machines and for basic tasks, but no more.

Intel was working on Atom SoCs in 2008, but its first fully unified chip wouldn’t ship until 2012.

Intel’s post-launch attitude towards Atom is best summarized as benign neglect. While the chip went through several revisions to integrate components and reduce costs, Intel refused to commit the resources that would have made Atom a best-in-class player in the mobile market. From 2008 to 2013, Intel launched a cost-reduced version of its Nehalem architecture, the Westmere 32nm die shrink, a new architecture with integrated graphics (Sandy Bridge), a high-end enthusiast platform (Sandy Bridge-E), a new 22nm CPU with FinFET technology (Ivy Bridge), another architectural refresh (Haswell), and a second-generation enthusiast platform (Ivy Bridge-E). That’s two full tick-tock cadences for Intel’s big-core business, while Atom didn’t even make the jump to 32nm until 2012. Its single architectural refresh to date arrived in 2013, just after the launch of Ivy Bridge-E.

Despite being initially starved for resources, 32nm Atom chips were competitive in the midrange mobile market. With Medfield, Intel seemed to have turned a corner, but the company’s designs generally failed to find much traction in the market. Only Intel’s contra-revenue strategy won the company significant tablet market share, and those gains were only sustained through heavy financial losses.

Intel’s mobile and communications revenue and losses from Q1 2013 through Q2 2014.

Atom wasn’t the problem — Atom was the solution Intel didn’t have the guts to chase.

The Tough Decisions Intel Didn’t Make

Intel failed to gain traction in mobile because it wasn’t willing to risk upsetting the economic model that had transformed it into a titan of computing. The company’s fabs, manufacturing strategies, and resources were geared towards large, expensive processors, not churning out huge numbers of low-cost mobile cores. Prioritizing Atom over Core would’ve required the company to retool at least some of its fabs to emphasize throughput and lower costs in order to compete with the ARM processors built at Samsung and TSMC. It would’ve meant lower gross margins and less profit per unit sold.

Intel tried to speed things up eventually, but both its 14nm process and the next-generation Goldmont CPU core were delayed.

Intel did take steps to improve its competitive standing vis-a-vis ARM and ARM’s foundry partners, but it rarely took them quickly and often failed to follow through. Intel bought Infineon Wireless in 2011 for $1.4 billion, but to this day all of its publicly announced wireless products, including the XMM 7480 modem, are still built on 28nm at TSMC. Smartphones and tablets have always used SoCs, but Intel didn’t launch its first Atom-based SoC until 2012 — five years after the iPhone launched and four years after Atom’s own debut.

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

One thing we want to stress here is that Intel’s decision to protect its core (Core) business and product margins may have been wrong, but it wasn’t crazy. Refitting fabs, building expertise in SoC design, and porting modems from TSMC would have required large cash infusions and take significant amounts of time. If Intel had launched Atom with an aggressive plan to put the chip in smartphones by 2010, things might have played out very differently. By the time the company woke up to the threat it faced from ARM and merchant foundries, it was too late to make up the gap.

Why Intel’s Foundry Tech Couldn’t Save Its Mobile Business

Intel’s process technology leadership couldn’t save the company’s mobile division because it wasn’t designed to do so. Smartphone and tablet OEMs wanted devices with integrated LTE radios; Intel didn’t have them. Even the SoFIA partnership with TSMC never came to market, apparently because Intel couldn’t secure enough volume to kickstart production.

Intel’s 14nm problems delayed its next-generation tablet processors from 2014 to 2015. Its 10nm node, once expected to secure enormous economies of scale over TSMC, has been pushed to 2017 as well. I don’t think these delays played a huge role in Intel’s decision to leave the mobile market, but they may have influenced it. In 2012, Intel still expected to be on 10nm by 2016 with EUV (extreme ultraviolet lithography) ramping towards full production. The now-canceled 450mm wafers weren’t expected in-market quite this soon, but Chipzilla expected to recognize significant cost savings from moving to the larger wafers in the 2018 to 2020 timeframe — cost savings that could’ve further improved its standing against Samsung, TSMC, and GlobalFoundries.

If Intel had begun reorienting towards Atom when it launched the chip in 2008, it might’ve weathered these delays and cancellations without much trouble. Failing to do so left its beleaguered mobile business facing higher-than-expected costs and minimal revenue.

Intel didn’t lose the mobile market because Atom’s performance and power consumption didn’t compete with ARM; research and evaluation showed that Atom was capable of matching ARM performance in multiple market segments. It lost the mobile market because it didn’t make the changes that would have allowed it to compete on cost with products manufactured at TSMC and Samsung. The exacting rules and unique layouts that drove Intel to the top of one market could not be easily adapted to others, and Intel was unwilling to risk its position at the top of the conventional x86 market for a risky payoff in mobile. There’s no evidence that keeping XScale or developing ARM products would have changed that — if anything, the ARM division would’ve been under even more pressure to ensure it never became a threat to the x86 business.

Now read:

• Intel Core i9-10900K CPU Review: Comet Lake Paints a Target on AMD’s Matisse
• Intel May Have Reserved Its Top-End 28W Ice Lake CPUs Exclusively For Apple
• Intel, TSMC Reportedly in Talks to Build New US Foundries

Tagged In

This site may earn affiliate commissions from the links on this page. Terms of use.

ExtremeTech Newsletter

Subscribe Today to get the latest ExtremeTech news delivered right to your inbox.

This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our
Terms of Use and
Privacy Policy. You may unsubscribe from the newsletter at any time.

More articles

Smartphones with Intel Atom Z3560 processor

• Add to compare

  Lenovo K8 K80m

  China · 3GB · 32GB

• Add to compare

  Asus ZenFone 2 ZE550ML WW

  Global · 2GB · 16GB

• Add to compare

  Asus ZenFone 2 ZE550ML WW

  Global · 4GB · 32GB

• Add to compare

  Asus ZenFone 2 ZE550ML WW

  Global · 4GB · 16GB · Deluxe

• Add to compare

  Asus ZenFone 2 ZE550ML WW

  Global · 4GB · 16GB

• Add to compare

  Lenovo K8 K80m

  China · 4GB · 64GB

• Add to compare

  Asus ZenFone 2 ZE550ML CN/IN

  Global · 4GB · 32GB · Deluxe

• Add to compare

  Asus ZenFone 2 ZE550ML CN/IN

  Global · 4GB · 16GB · Deluxe

• Add to compare

  Asus ZenFone 2 ZE550ML CN/IN

  Global · 2GB · 16GB · Deluxe

• Add to compare

  Asus ZenFone 2 ZE551ML CN/IN

  Global · 2GB · 16GB

• Add to compare

  Asus ZenFone 2 ZE550ML CN/IN

  Global · 4GB · 16GB

• Add to compare

  Asus ZenFone 2 ZE550ML CN/IN

  Global · 4GB · 64GB

• Add to compare

  Lenovo K8 K80m

  China · 2GB · 32GB

• Add to compare

  Asus ZenFone 2 ZE550ML CN/IN

  Global · 2GB · 32GB

• Add to compare

  Asus ZenFone 2 ZE550ML CN/IN

  Global · 4GB · 32GB

• Global · 2GB · 16GB

  From $ 289

Tech Specs | ASUS Global

Performance for Today, and Beyond

ZenFone 2 is powered by a 64-bit 2.3GHz Super Quad-Core Intel® Atom™ Z3580 processor, and is the world’s first smartphone with 4GB of dual-channel DDR3 RAM. Featuring LTE Category 4+ for download speeds up to 250Mbit/s, and a 60ms touch response time, ZenFone 2 gives you power beyond your imagination.

Try it







64-bit Intel Atom 2.3G

CPU







6430 PowerVR 533MHz

GPU







64-bit Dual-Channel 4GB

RAM







Cat4+ 250Mbps

LTE W/ CA







Android L

OS







60ms

Touch Responsiveness







5-mode Positioning

Navigation







5 LDS

Antenna







DSDA

SIM Card







802. 11ac 433Mbps

WiFi







5.5in FHD

Screen

The Newest 64-bit Intel® Atom™ Processor Z3580

The 64-bit Intel® Atom™ Super Quad-Core 22nm processor inside ZenFone 2 carefully balances brilliant performance and superb power efficiency, empowered by the sophisticated Silvermont architecture.

Intel® Exclusive 3D

Tri-Gate Transistor

First introduced by Intel® in its desktop-class Ivy Bridge processors, ZenFone 2 employs the very same 22nm technology in its 64-bit processor. The Tri-Gate transistors have almost three times (3X) more surface area as compared to planar transistors, enabling more electrons to flow when it in the ‘on’ state. It also improves the transistor switching performance which increases the computing performance and decrease the power leakage.


Performance
Increase
37%


Power
save
50 %

Dual Channel Data Transmission

With 64-bit CPU, now the processor can fully unleash the power of the Intel® dual channel RAM architecture, by delivering 2 pipes (32-bit each pipe) of data from CPU to RAM at the same time.

Android L

OS designed for 64-bit Intel CPUs

Unprecedented Performance in Every Scenarios

WebXPRT and MobileXPRT are two of the most sophisticated benchmarks that measure the real-world performance of web- and Android-based usage scenarios. ZenFone 2’s Intel Atom Z3580 easily outperforms both mid-end even high-end competitors.

System Performance

MobileXPRT

Web-Browsing Performance

WebXPRT

System

Performance
(MobileXPRT)

Web-Browsing Performance

(WebXPRT)

Incredible Graphics Performance

ZenFone 2 PowerVR 6430 graphics-processing unit (GPU) packs a serious punch, for computing experiences that feel like using your desktop computer. Demanding, graphics-intensive games and videos run buttery smooth and look splendid on ZenFone 2’s Full HD display.

GPU Performance

GFXbench, T-Rex Offscreen

GPU Performance

GFXbench, T-Rex
Offscreen

GPU Performance

3D Mark, Ice Storm Unlimited score

GPU Performance

(GFXbench,T-Rex
Offscreen)

GPU Performance

(3D Mark, Ice Storm
Unlimited score)

To See, To Believe

HTML5 Fish Bowl delivers visualized performance metrics to show true system capability. With the same settings, and a bowl containing 250 virtual fish, faster-swimming fish means faster performance and the ability to handle complex tasks — and ZenFone 2 wins out.

50~60

ZenFone 2

22~26

Competitor A

8~11

Competitor B

50~60

ZenFone 2

22~26

Competitor A

8~11

Competitor B

World’s 1st Smartphone with 4GB RAM

Ever experienced sluggishness when using your phone, and felt the need to kill background apps for a speed boost? With the 4GB RAM version of ZenFone 2, you will never have to worry about that again!





There’s high and an increasing demand to run powerful apps and multitasking on smartphones, yet many models are equipped with just 2GB or 3GB of memory. ZenFone 2 is the only smartphone with 4GB RAM capability — making it a the industry’s monster performer!And with a 64-bit processor, ZenFone 2 fully unleashes the power of the Intel® dual-channel RAM architecture, by delivering two simultaneous 32-bit data channels between processor and memory — and that means the best-ever performance from ZenFone 2.

4GB RAM for Incredibly Smooth Multitasking

Instant switching between apps
(26 apps running in the background)

Dialer

Email

Calendar

Camera

Chrome

Facebook

4GB model wins all apps switching speed compared to 2GB model. Example: ASUS Browser, Gallery, Gmail.




Faster-than-Ever LTE for Up to 250Mbits Downloads

ZenFone 2 has the latest 4G LTE Advanced technology for the highest-ever mobile data speeds — up to a staggering 250Mbit/s*! ZenFone 2’s 4G LTE Advanced also supports carrier aggregation*, exploiting the bandwidth of multiple carriers for ultra-reliable, ultra-high speeds wherever you go. With over 31 countries already supporting carrier aggregation and many more set to do so, ZenFone 2 is a world-leading smartphone — with world-leading mobile speeds!


* Varies by country, region and carrier support

433Mbit/s 802.11ac Super-Fast Wi-Fi

ZenFone 2 is equipped with 802.11ac Super Fast Wi-Fi, which is a supercharged version of 802.11n (the current Wi-Fi standard that your smartphone and laptop probably use). It offers speed from 433Mbit/s and operates in up to eight spatial streams (MIMO), employing an advanced technology called ‘beamforming’ to increase throughput and reduce power consumption.

Class-Leading 60ms Touch-Response Time

Encounter a stunning touch experience on ZenFone 2, with a record-breaking response time of just 60ms — way ahead of competitor devices that typically take 80-100ms to respond. Scrolling is smooth and free from lag so the screen feels alive to your touch, and games respond with speed that helps you secure victory!

Dual SIM, Dual Active Smartphone that Never Misses A Call

ZenFone 2 has a Dual SIM, Dual Active technology, so you can fit two SIMs and have both active at all time. So, while talking to your friend from one SIM, you’ll still receive messages and calls from the other. If you need to switch focus, just put the first call on hold to deal with the second!

Five Antennas by Laser Direct Structuring

Five Antennas by

Laser Direct Structuring

ZenFone 2 has five innovative laser-engraved 3D antennas to dynamically improve signal quality and create space. The latest Laser Direct Structuring (LDS) technology is used to mark ZenFone 2’s inner plate, which is then dipped into three different liquid metals to complete the formation of the antennas. The result is greater signal coverage for effortless video streaming and web browsing.




Five Satellite Systems Supported

ZenFone 2 supports five global satellite-navigation systems, including GPS, GLONASS, BDS, QZSS, and SBAS. With the ability to track up to 60 satellites, ZenFone 2 provides you with the most precise and accurate position no matter where you are, wherever you go — even in built-up urban areas!


GPS Satellite Cover


5 Satellite System Cover

Full HD IPS Crystal-Clear Display

Full HD IPS

Crystal-Clear Display

ZenFone 2 is equipped with a high-resolution 1920×1080 Full HD IPS display that offers wide 178-degree viewing angles. The staggering 403ppi pixel density renders every image in eye-delighting detail.

72% NTSC Color Range

ZenFone 2 ultra-sharp display reaches 72% of the NTSC gamut so you see more colors, just as precisely as your eyes enjoy the real world. Photos and video are brighter, clearer and as vivid as can be — and so much more lively than competing smartphone displays!

Full Screen Lamination with

ASUS TruVivid Technology

For the best clarity, brightness and touch responsiveness, ZenFone 2 incorporates exclusive ASUS TruVivid technology. This seamlessly transforms the conventional four-layer design into an innovative two-layer design, with full lamination. Fusing the cover-glass and touch-panel layers eliminates the air gap, bringing everything closer to ZenFone 2’s surface — and closer to your touch.










Corning® Gorilla® Glass 3 Tough, Yet Beautiful

Gorilla Glass 3 with Native Damage Resistance™ (NDR) improves damage resistance and toughness, so ZenFone 2’s display is up to three times (3X) stronger than other glass compositions — and up to three times (3X) more damage-resistant than Gorilla Glass 2. NDR is a unique feature of Gorilla Glass 3 that provides better resistance to deep scratches that cause glass to break.


0.7mm Corning
Gorilla Glass 3

Reinforced 0.7mm Soda
Lime Glass
Reinforced 0.7mm Soda Lime Glass


Responds to Every Touch —

Even with Gloves

ZenFone 2 supports ASUS Glove Touch technology, so it responds to your every touch, swipe and gesture — even if you’re wearing gloves.

Expansive Audio, Tuned by Experts

Expansive Audio,

Tuned by Experts

Tuned by the renowned ASUS Golden Ear team to provide you with the very best smartphone audio, ZenFone 2’s SonicMaster technology delivers crystal-clear audio — always. Whether you’re on headphones or speakers, you’ll experience richer, more detailed audio in everything you do.




Size Matters

ZenFone 2’s sound chamber has been improved in many ways. It is now 25% larger than before, with a three-magnet construction and detail, metal voice coil that delivers more powerful sound in every scenario.

The Smart ASUS AudioWizard for

Every Scenario

The Smart ASUS AudioWizard for Every Scenario

ASUS AudioWizard is the smart software that provides the best listening experience in different scenarios, including Movie Mode, Gaming Mode, Keynote Mode, Music Mode and Smart Mode.

• Movie
  Mode
• Music
  Mode
• Gaming
  Mode
• Speech
  Mode
• Smart
  Mode






Next

See What Others Can’t See


Learn more about Sharp Vision





The death of Intel’s Atom casts a dark shadow over the rumored Surface Phone

Intel’s plans to discontinue its Atom chips for phones and some tablets may not have killed the dream of a Microsoft Surface phone—just the piece of it that made it so enticing.

In the wake of a restructuring that relegated the PC to just another connected device, Intel confirmed Friday that it has cancelled its upcoming SoFIA and Broxton chips. That leaves Intel with just one Atom chip, Apollo Lake, which it had slated for convertible tablets.

Microsoft has never formally commented on its future phone plans, save for a leaked email that suggests that Microsoft is committed to the Windows 10 Mobile platform and phones running ARM processors. But fans of the platform have long hoped for a phone that could run native Win32 legacy apps as well as the new UWP platform that Microsoft has made a central platform of Windows 10. The assumption was that this would require a phone running on an Intel Atom processor. Intel’s decision eliminates that option.

Unless Microsoft has some other trick up its sleeve, the most compelling justification for a Win32-based Surface phone appears to have died.

Shawn Morgan / Intel

At Mobile World Congress, Intel CEO Brian Krzanich broke out a drone to show the kinds of devices that Intel would eventually base its company upon.

Intel gives up on the smartphone

Intel’s decision was first reported by analyst Patrick Moorhead, and confirmed by IDG News Service and PCWorld. Intel told PCWorld that it plans to kill the “Broxton” Atom platform as well as all the flavors of its SoFIA chips, which combined Atom cores with 3G and LTE modems for smartphones. The company said it will continue to support tablets with a 3G derivative of the SoFIA chip, the older Bay Trail and Cherry Trail, as well as some upcoming Core chips.

Microsoft uses a Cherry Trail chip inside of the HoloLens, but it’s unclear whether or not that will be affected. 

“I didn’t get the sense that they’re going to exit the tablet space—I felt like there was [a message] of ‘more coming soon,’” Moorhead said. “On the phone stuff, I just don’t think they’ll continue to do that, because, you know, it’s a business.”

Intel said recently that it plans to refocus on the data center, the Internet of Things, memory, 5G modems, and connectivity. To so, Intel will lay off 12,000 workers over the next year or so. But some Atoms apparently fell in the wake of a comprehensive program by Venkata Renduchintala, the new president of a newly created Client and Internet of Things (IoT) Businesses and Systems Architecture Group, to examine the viability of products across Intel’s client businesses. 

Intel

Intel once had plans to retake the smartphone market with an integrated 4G/microprocessor SoC called “SoFIA.” Those plans are now dead.

Intel chief executive Brian Krzanich spoke optimistically about the phone business in 2014, and a year ago appeared prepared to do whatever it was necessary to make phones succeed. Now, Intel has acknowledged its defeat on the phones front and vowed to fight elsewhere.  

“The reality is that if you’re going to spend money, you have to take it from somewhere else,” Moorhead said.

Windows Phone: running out of options

So does that mean that the dream of an Intel-based Windows phone is dead? “Yes, if they’re not going to go after the smartphone, I don’t see how that would be possible,” Moorhead said.  

If Microsoft does remain committed to an X86-compatible Surface phone, it does have some options, however unattractive. Intel recently announced its next-generation Apollo Lake Atom architecture, which Intel positioned as the foundation for, not a phone, but an “entry level PC.” The Apollo Lake chips will be branded as Atom, Celeron, and Pentium, Intel said.

Intel hasn’t said exactly how much power the Apollo Lake chip will consume, though it’s made on the same 14-nm process as the existing “Cherry Trail” Atom chips. The Atom X7-Z8700 used in the Surface 3 is rated at just 2 watts of power. Does that mean that it (or Apollo Lake) could be squeezed into a phone or phablet running Windows 10 Mobile?

Could the Surface 3 be turned into a phone?

Intel seems to have left the possibility open. “In terms of Cherry Trail, form factor boundaries are increasingly blurring in the mobile computing market, and we no longer look at tablets as a stand-alone segment,” an Intel spokeswoman said via email. “Our product roadmap reflects that.”

Analysts, though, say it looks doubtful. “I have a hard time visualizing the package that would be compelling: with a phone running a processor that has never been used as a phone before that’s running as a fake desktop,” Wes Miller, an analyst with Directions on Microsoft, said in an interview. “And then what are you going to do for battery life? It’s very perplexing.”

That also means that that phone would have to have a discrete modem built in. Kevin Krewell, principal analyst at Tirias Research, noted that the SoFIA chip Intel cancelled was its first to have an integrated modem, which would make the phone even larger.

Unless AMD can somehow cut the power of its embedded chips—an aging AMD G-series LX still consumes about 5 watts—the only way to enable Win32 compatibility may be through virtualization. That’s the strategy HP’s Elite X3 is reportedly using, though that virtualization is back up through the cloud and back down to the phone. It appears, based on this Windows Central FAQ, that a business will have to tailor this compatibility with HP itself on an app-by-app basis.

Melissa Riofrio

HP’s Elite x3, shown here resting inside its Desk Dock, will ship this summer.

Even then, however, it’s hard to imagine a scenario where a phone that needs to communicate with the cloud to perform virtualization in a faux-desktop environment wouldn’t end up sitting next to a computer or thin client that does the same thing, Miller pointed out.

So… same old, same old?

The leaked memo from Microsoft’s OS chief, Terry Myerson, certainly makes clear that Microsoft is committed to Windows 10 Mobile. (A Microsoft spokeswoman declined to comment on Intel’s announcement.)  “Let me be very clear: We are committed to deliver Windows 10 on mobile devices with small screen running ARM processors,” Myerson wrote. “We are currently in development of our next-generation products and I wanted to reconfirm our commitment to Windows 10 Mobile. ”

Which, of course, is exactly what Microsoft is shipping today: Windows 10 Mobile on ARM processors. But sales are still plunging, and Microsoft is giving its phones away as a result.

Rob Schultz

Consumers just never fell in love with the Lumia 950, and Microsoft’s Windows 10 Mobile OS is still hovering at less than 10 percent share of all Windows phones, according to AdDuplex.

Slapping a Surface brand on an existing Windows phone won’t save Microsoft’s phone business. Microsoft needs a compelling message that it can uniquely deliver: “the tablet that replaces your PC” was the tagline that sold the Surface lineup. It’s possible that Microsoft has another moonshot in place for its mobile business—who knows, maybe it’ll make the Microsoft Band its “phone” of the future. Stability issues aside, the Surface Book demonstrated that Microsoft can craft innovative hardware using the same components its competitors use. 

For right now, though, it appears another door has closed on Windows Phone. And that’s just not good news for a platform struggling as hard as it is.

Additional reporting by Gordon Mah Ung. Updated at 5:02 PM to note that Microsoft uses a Intel “Cherry Trail” Atom inside the HoloLens, according to an IDG News Service report.

ASUS Zenfone 2 Review — Intel Atom Powered Smartphone!

Earlier this week ASUS launched the Zenfone 2 and we were able to get our hands on this brand new smartphone that costs less than $300 outright. Yes, we are talking $299.00 unlocked smartphone here that doesn’t require a contract that will work on the AT&T and T-Mobile carrier networks here in the United States (Sprint and Verizon are not supported). While the price tag on this device might be less than half of the price of a Samsung Galaxy S6 Edge and Apple iPhone 6 Plus, the specifications of Zenfone 2 look pretty damn good!

When it comes to general features you are looking at a large 5.5-inch 1920 x 1080 Full HD IPS LTPS display, 2.33 GHz Intel Atom Z3580 ‘Moorefield’ quad-core 64-bit processor, 64GB of storage and 4GB of LPDDR3 RAM running Android 5. 0. The specifications on this phone are pretty impressive and it should be competitive with many of the big name premium phones on the market today. Having 4GB of memory on a smartphone is pretty crazy as there are millions of desktop PCs and laptops that are running with that amount of memory or less right now.

The ASUS Zenfone 2 will also have a $199 version that has lesser hardware inside. On that model ASUS went with the 1.83GHz Intel Atom 3560 64-bit quad-core processor, 16GB of storage and 2GB of LPDDR3 RAM. The lower cost model also doesn’t feature the 9V / 2A QuickCharger with BoostMaster technology that promises to charge this phones 3000mAh battery from dead to a 60% charge in under 40 minutes. For a $100 difference in price it looks like the higher-end sku is well worth the additional price as you are getting a faster processor, quadruple the storage space, double the RAM and a wall charger that will charge your phone quicker. Both are still arguably stupid cheap for what you are getting though!

ASUS let us try out the ZenFone 2 ZE551ML in gray, but you have a choice between Osmium black, Glamour red, Sheer gold and Glacier gray when you go to buy one. The version we are trying out is of course the high-end $299 premium model that has all of the bells and whistles.

The ASUS Zenfone 2 comes with the USB wall charger, required USB cable for charging and backup purposes and a small user guide that will help you get going. One thing that we noticed right away when using the Zenfone 2 is that it was mentioned on the packaging that this is a “64-bit Super Quad Core Processor” powered device running Android 5.0. The thing is that the OS is still only 32-bit and that despite the fact that our phone has 4GB of memory it shows up as having 3.8GB of addressable memory space.

With a 5.5-inch display you’ll quickly find that the Zenfone 2 is a fairly large phone at 152.5 x 77.2 x 10.9 mm thick. That is similar in size to the LG G4 and the iPhone 6 Plus! The Zenfone 2 also has a heavy feel to it since it weighs in at 170g, but for some reason the construction feels a bit cheap. On the bottom you have three capacitive buttons for back, home and recent along the bottom of the Gorilla Glass 3 display. Up top you have a front-facing OmniVision 5-megapixel camera that takes 2560×1920 pictures as well as 1080P 30FPS video.

The back of the Zenphone 2 has the single 0.7W speaker located at the bottom. ASUS went with the Realtek ALC5647 Audio DAC for their audio solution and the phone has two microphones for active noise cancellation technology. Up top you have a Toshiba 13-megapixel rear-facing camera with flash. This camera has a f/2.0 aperture that has a maximum picture resolution of 4096×3072 and video resolution of 1920×1080 at 30 frames per second with video stabilization. You also have the volume rocker located on the back of the phone directly below the camera lens. After using the Zenfone 2 for a number of days the cheap feel might be because the curved brushed aluminum back cover isn’t firm since it isn’t really metal and has a hollow sound when you tap on the plastic.

The camera itself works pretty well indoors in poorly or oddly lit areas like the image we took in a hotel lobby of a flower setting that they had. The next two images shows a wine glass in a bar with dim lighting in auto mode with the flash disabled and then in low light mode. The ASUS PixelMaster 13MP camera has 16 different built-in camera modes so you’ll be able to get the shot you need right. The included modes are are auto, manual, HDR, panorama, effect, low light (2M), selfie panorama (140 degree selfies to fit everyone), beautification, time rewind, miniature, depth of field, smart remove, all smiles, gif animation, night and time lapse.

The power button for the Zenphone 2 is located at the the phone as is the 3.5mm audio jack. This location might feel awkward for some, but all phones have a learning curve when it comes to key placement. The button feel isn’t that crisp as the button is built into the phone case and is therefore removable. Most phones have a power button that has a crisp tactile feel to it and that is not the case with the Zenfone 2.

If you were hoping to see a USB 3.0 port on the bottom of the Zenfone 2 you’ll be disappointed as ASUS went with the traditional and more cost effective micro USB 2. 0 port for handling data transfers and battery charging duties. The good news is that ASUS included an 802.11ac wireless solution inside that is capable of 233 Mbps speeds (29MB/s) and the controller also supports Bluetooth 4.0 LE.

With the battery cover removed you’ll see a battery that will take some disassembly to get out, dual active micro-SIM slots and a microSD card slot. This means that you can use two separate SIM cards with different phone numbers and voice plans at the same time. You can even put a call on hold to answer a call from the other SIM, so there is no need to carry around two phones unless you want to! The primary SIM card slot supports LTE/4G/3G data while the secondary SIM card slot is limited to 2G/texting/voice only. ASUS went with the Intel XMM7260 LTE-Advanced modem in the Zenphone 2, which is rated for 150Mbps download speeds and 50Mbps upload speeds on LTE Cat 4 (LTE Adcanced) networks.

The MicroSD card slot supports SDXC cards up to 128GB in capacity. You can also see the single speaker that is hidden by the back cover in the lower left hand corner. The rear facing audio solution isn’t stellar, but it works good enough in most situations. ASUS says that the internal 64GB EMMC storage is rated at 165MB/s read and 95MB/s write when it comes to sequential data transfers. The non-swappable 3,000mAh battery easily lasted a day for us with normal use, but battery life greatly varies on how you use the phone.

We also got a chance to try out the optional ASUS ZenFone 2 View FlipCover Deluxe that allows up to 7 apps to be shown through the cover as well as adding a bit of extra protection from scrapes and scratches. Pricing is unknown for the Deluxe case, but we expect it to be made available for purchase in the weeks ahead.

The Zenfone 2 was found to be very responsive and it wasn’t sluggish during any time in our testing. Since the is an ASUS phone you’ll find proprietary apps like ZenCircle and MyASUS and a fair number of third-party applications like Amazon Kindle, Clean Master, Omlet Chat, Party Link, Dr. Safety, Jawbone Up, TripAdvisor and so on. We don’t like to see any bloatware, but unfortunately the Zenfone 2 comes with a fair bit of it. When we used the phone for the very first time we got pop up notices from Clean Master saying that we had apps at idle. This doesn’t make for a good user experience as I don’t want to see any messages coming up on a brand new phone that isn’t really relevant. The good news is that you can remove most all of it! ASUS likely gets a kick back from having these apps installed, so hopefully they are passing along some of that money to consumers by lowering the sale price.

The Android 5.0 ‘Lollipop’ OS with the ASUS ZenUI should be familiar to familiar to most. The responsiveness when swiping through menus and apps was excellent. This phone doesn’t feel sluggish at all. The icons are a bit large though and the drop down menu has nearly 20 options on it.

Gaming and benchmarking on the Intel Quad Core 64-bit Atom Z3580 2.3GHz processor and PowerVR Series G6430 GPU with OpenGL 3. 0 support was an overall good experience. The 5.5″ screen with the 1920×1080 resolution looked good. The pixel density of this display is 403 PPI and it has a first touch response of 60ms. The Corning Gorilla Glass 3 has an anti-fingerprint coating on it, but we still saw the usual smudges and fingerprint marks on the display as you can even see some in the picture above right in the middle of the display. The display has a brightness of 400nits, so it is a little dimmer than some of the other phones we have been using lately. The iPhone 6 is rated at over 600nits and the Samsung Galaxy S6 have around 550nits.

Benchmark performance on the ASUS Zenfone 2 was pretty mix with scores all over the place. We ran Geekbench 3, Vellamo, GFXBench 3.1, AnTuTu v5.7, Lightmark and 3DMark Ice Storm Unlimited. Benchmarks like 3DMark Ice Storm Unlimited are run using OpenGL ES 2.0 at a 720p resolution and the Zenfone 2 scored 20013 points and averaged 104FPS in game test 1 and 74FPS in game test 2. Pretty solid benchmark scores! On other benchmarks like Lightmark showed we averaged just 7.42FPS on the normal 1080p benchmark. The SunSpider 1.0.2 JavaScript benchmark has the ZenFone 2 coming in at 632ms, which is a solid middle of the road score. You can check out all the scores in the gallery above and these are all free benchmarks that you can do at home on your current smartphone to see how much faster the Zenfone 2 is than your current device.

ASUS Zenfone 2 Final Thoughts and Conclusions:

If you are looking for an affordable unlocked phone with near-flagship hardware specifications running on Android 5.0 it is hard to ignore the ASUS ZenFone 2. The Zenfone 2 has the large 5.5-inch display that is craved by many these days and the user experience was found to be overall positive. For the price there aren’t glaring issues that we would consider deal breakers. After using the Zenfone 2 for about a week we can nit pick things like the screen brightness isn’t on par with flagship phones, how ASUS installed too much bloatware or that the faux metal case back should have been made from real metal, but that is about it.

When it comes to pricing the premium $299.00 shipped sku would be the one we reviewed today that has an Intel Atom Z3580 (2.3GHz) processor, 4GB of RAM, 64GB of internal storage and the 9V / 2A QuickCharger. The other open would be the $199 shipped version that features an Intel Atom Z3560 (1.86GHz) processor, 2GB of RAM and 16GB of internal storage. We were left satisfied be the performance

Keep in mind that this Intel-powered unlocked smartphone is compatible with AT&T, T-Mobile, MetroPCS, Cricket Wirless, Brightspot, Net10, h3O Wireless, Straight Talk, Giv and other GSM cellular carriers. It is not compatible with Sprint, US Cellular and Verizon!

Legit Bottom Line: For the $299 price point you’d be hard pressed to find a phone that has the overall features and performance of the ASUS Zenphone 2!

Questions or comments?
View this thread in our forums!

Intel splits on Atom after the mobile relevance of x86 whacked by Apple’s Ax



AppleInsider is supported by its audience and may earn commission as an Amazon Associate and affiliate partner on qualifying purchases. These affiliate partnerships do not influence our editorial content.


Intel is ending production of its mobile Atom processors as the reality sets in of a mobile world where virtually all the profits are inhaled by devices sold by Apple and powered by the iPhone-maker’s own custom Ax series Application Processors.

Atom has Intel thinking outside the mobile box after failing to gain any traction

Intel initially intended for Atom to scale down its legacy Wintel desktop x86 processor architecture for use in efficient mobile devices such as phones and tablets, but that strategy has been effectively abandoned as the chipmaker now moves to refocus its sights on modems, data center, Internet of Things and memory chips.

The decision to cancel Atom as a mobile strategy boils down to profitability.

Despite having about as many employees globally as Apple, Intel’s revenues were a quarter of Apple’s last year, and the Mac-maker reported profits 4. 7 times higher on sales of finished devices.

The majority of Apple’s products are now powered by Ax chips the company designs internally. Other smartphone and tablet makers earn little to nothing, and virtually all have adopted the ARM architecture Apple initially co-developed back in the early 1990s for use in its Newton MessagePad tablet, offering little opportunity for Intel’s rival Atom architecture to make any inroads.

How Intel lost the mobile business to Apple’s Ax chips

The move kills Intel’s once enthusiastic plans to muscle its way back into smartphone devices after first fumbling the ball in 2006, when its former chief executive Paul Otellini overlooked the prospect of supplying chips for Apple’s original iPhone as not worth doing.

Four years later, Intel sought to win back Apple’s attention with Silverthone, a new x86 chip it expected the company to use in its initial iPad. Intel subsequently rebranded the mobile x86 chip as Atom.

However, Apple passed on Intel’s Atom and instead acquired the expertise to develop its own custom ARM Application Processor, known as the A4, which it used in the first iPad, then in iPhone 4 and Apple TV.

Intel’s Atom jilted by Microsoft, failed by Android

The following year, at the beginning of 2011, Microsoft also snubbed Intel by showing off what would later be named Windows RT, a project to get Windows working on ARM chips— including the Qualcomm Snapdragon, TI OMAP and Nvidia Tegra chip families— in order to keep low end Windows PCs, netbooks, tablets and hybrid 2-in-1 devices competitive with Apple’s increasingly popular iPad.

Intel’s Atom chips were clearly unable to match the A4 iPad’s battery life, but Windows RT, in the form of Microsoft’s own Tegra-powered Surface RT hybrid, also failed spectacularly.

Jilted by Microsoft, Intel subsequently jumped to support Google’s Android 3.0 Honeycomb also in competition with iPad, offering makers fat subsidies for building Android tablets using Intel Atom chips.

By the end of 2011, Google and Intel had also announced plans, as optimistically reported by Reuters, to «work together to optimize future versions of Google’s Android mobile software for Intel’s Atom processors, hoping to speed the development and time-to-market of future Intel-powered smartphones.»

However, Android did nothing to help Intel gain any traction in smartphones. And despite spending billions to subsidize Atom-powered Android tablets, those never made any money either.

In 2014, Intel shifted its plans to instead license its x86 Atom, 3G and LTE baseband IP to Chinese fabless chip designers including Rockchip, Spreadtrum and RDA Microelectronics, intending to follow a licensing model similar to ARM or Qualcomm. Those plans were branded as «SoFIA».

As AppleInsider outlined a year ago, the prospect for Intel to turn Atom around and make x86 relevant in on mobile devices again was essentially zero.

End of the road for fantasies of a Microsoft Surface Phone running Windows apps

As noted yesterday by PC World, the cancelation of phone-scaled Atom chips by Intel also «casts a dark shadow over the rumored Surface Phone,» a vaporware product that imagined packing a desktop PC into a smartphone, which could then be plugged into a monitor and keyboard using Microsoft’s «Continuum» strategy to run legacy x86 software.

Intel will kill off its «Broxton» Atom and SoFIA architectures, leaving behind only a low end chip using the Atom brand aimed at basic desktop PCs, effectively pulling out of the business of powering iPad-class tablets or smartphones.

That leaves Microsoft in the position of selling Windows 10 Mobile only on ARM, and Windows 10 desktops and hybrids (like the Surface Pro 4) only running desktop x86 processors. Without the ability to emulate x86 code on ARM, that leaves Windows focused entirely upon «UWP,» the Universal Windows Platform introduced for Windows 10 intended to stretch generic software across a variety of device types and form factors, from touchscreen phones to a mouse driven desktop.

That’s a strategy that hasn’t worked well for Android, where apps generally target phones and don’t offer specialized optimization for tablets or other form factors.

Microsoft’s UWP hasn’t exactly seen enthusiastic adoption either, with legacy software still bound to mouse driven desktop PCs running x86, and mobile devices running Windows 10 so commercially irrelevant that developers have little reason to bother targeting them.

It is, however, good news for Apple, considering that the company now has a very large platform for iOS, with clear optimization for iPhone and iPad hardware. That has resulted in enthusiastic Enterprise adoption. In its last earnings call, Apple noted that «the latest data published by IDC indicates that iPad accounts for 72% of the U. S. commercial tablet market, comprising business, government, and education.»

Intel cancels new Atom processors for smartphones

3DNews Technologies and IT market. News mobile phones, smartphones, cellular … Intel cancels the release of new processors … The most interesting in the reviews 05/01/2016 [00:00], Anton Testov Four years after the release of the first SoCs for smartphones, Intel is phasing out two key solutions for this market: Broxton and SoFIA 2. In fact, Intel no longer wants to create solutions for these devices, since the gross profit from selling them constantly is declining, and phone manufacturers are demanding all kinds of support in creating devices based on Atom, which translates into losses for the company. The cancellation of two key projects at a late stage means that the company no longer believes in success in the smartphone processor market, but will try to focus on creating solutions for the fifth generation (5G) cellular network market. Atom chip Atom for tablets and smartphones cost $9 billion over three years 2012 with the support of Lenovo and Motorola Mobility. Since then, the company has released another generation of SoCs for mobile phones (Merifield/Tangier and Moorefield/Annidale), announced a family of low-cost SoFIA chips, and also agreed with Chinese developers Rockchip and Spreadtrum to create very cheap chips for affordable devices. Despite the high computing performance of advanced models, the low cost of solutions developed and manufactured by third-party companies, support for the Google Android operating system and software, Intel Atom has not become somewhat popular on the smartphone market. Two reasons for Intel’s failures are the constant delays in the release of new SoCs, as well as the lack of demanded features. Aside from several Lenovo and Motorola phones, the only notable consumer of the Intel Atom for smartphones is ASUSTeK Computer, whose Zenfone 2 has become quite popular in a number of countries. ASUS Zenfone 2. Possibly the last high-performance smartphone based on Intel Atom Attempts to popularize Atom among smartphone and tablet manufacturers cost Intel a lot of money. So, the company gave huge discounts on the processors themselves, invested in advertising, and also helped partners develop motherboards for end devices. All this resulted in the fact that the operating losses of the Mobile and Communications Group division, engaged in mobile processors, amounted to $1.776 billion in 2012 (on revenue of $1.791 billion), increased to $3.1 billion in 2013 (on revenue of $1.375 billion), and then jumped to $4.2 billion in 2014 (on revenue of $202 million), according to a report to the US Federal Securities Commission (security and exchange commission, SEC). In 2015, the company stopped disclosing the losses of the mobile division, creating a giant Client Computing Group that sells chips for all types of client devices. At the same time, if Intel managed to become one of the leading SoC suppliers in the tablet market, then the company could not take any significant share in the smartphone market. Apparently, in order to cut costs and free up engineers for more urgent projects, Intel has decided to abandon the production of a number of chips from the Atom family. SoFIA LTE and Broxton will not enter the market Thus, Intel will not release low-cost systems-on-a-chip SoFIA 3GX, SoFIA LTE and SoFIA LTE2 for affordable smartphones and tablets. As expected, Intel Atom x3-C344X with built-in 4G/LTE cat 6 modem (SoFIA LTE) will be based on Intel Silvermont microarchitecture and ARM Mali-T720 graphics core. Other SoC features include a single-channel LPDDR2/DDR3L memory controller, support for 802.11ac Wi-Fi, and Bluetooth 4.1 LE. Although the specifications of this chip look good, its release was supposed to take place in the middle of last year. In other words, the company missed the window of opportunity, and now this processor would not be very successful even in the segment of low-cost smartphones. It is impossible to talk about the prospects of SoFIA 3GX and SoFIA LTE2 chips, since almost nothing is known about these processors. However, one would hardly expect breakthrough functionality from them, given the history of all SoFIAs. It is noteworthy that today Intel does not want to talk about the fate of future Atom processors developed by Spreadtrum and Rockchip. Apparently, negotiations with these developers are still ongoing. Intel’s plans that will not come true If the SoFIA family was largely inherited by Intel from Infineon, then the Broxton project was created from scratch. It is believed that Broxton was a fully modular system-on-a-chip architecture that made it possible to quickly reconfigure microcircuits in accordance with market requirements by replacing «ingredients», as well as adding various third-party intellectual property if necessary. This approach is very similar to the ARM approach, which offers customers a range of components (CPUs, GPUs, memory controllers, specialized processors, accelerators, etc.) for a variety of needs, as well as technologies for their interaction (CoreLink bus and interfaces for connecting various devices), significantly reducing the time required to create chips. With Broxton, Intel could release new SoCs for mobile devices much faster than it is now. Intel plans that will not come true The first chip based on the Broxton platform (SoC codenamed Morganfield) was to have four cores based on the Goldmont microarchitecture, an LPDDR4 / Wide IO 2 memory controller, and also be manufactured by Intel using 14 nm process technology with FinFET transistors. Given the advanced microarchitecture and manufacturing at Intel’s factories, it’s clear that Broxton was targeting high-end smartphones first and foremost. Unfortunately, Broxton was delayed first due to problems with 14nm production rates, then due to unavailability. If Broxton came out at the end of 2016, it would face competition both from high-end SoCs developed by high-end smartphone makers (Apple, Samsung, Huawei/HiSilicon, Xiaomi, etc.) and from high-end processors built by Qualcomm , MediaTek and even Rockchip. With serious competition, Intel would either have to offer its processors at attractive prices, reducing profitability and revenue, or resort to the already used tactics of low prices and various partner incentives (marketing, development assistance, etc. ), which could mean losses. It should be understood that most well-known companies release flagship smartphones in the first half of the year, and to get into expensive devices in 2017, Intel would need to introduce Broxton in the spring or summer of this year. Apparently, Intel no longer intends to be unprofitable in a market where it has failed for several years. 5G is the future The cancellation of new SoFIA and Broxton chips means that Intel will leave the smartphone SoC market for at least the next couple of years. This does not mean that the company will stop producing solutions for mobile phones altogether. Intel will continue to make a variety of modems and even intensify investment in 5G modems, as CEO Brian Krzanich outlined earlier this week. 5G application possibilities. EC slide 5G networks are expected to be more penetrating than 4G networks, and modems will be installed in a wide variety of devices. Thus, Intel’s 5G modems will not be released exclusively for mobile phones (tablets, or laptops), and therefore Intel sees great prospects in this direction. While Qualcomm has virtually monopolized the market for high-performance 4G/LTE modems for smartphones, the company’s advantages may not be so obvious in the case of 5G, which means opportunities for Intel. As for the possible release of SoC for mobile phones in the future, Intel’s potential in this direction will be determined not only by the company itself and the development of its processor microarchitectures, but also by the devices themselves. In the event that Intel has the opportunity to sell its chips for smartphones at a profit, the company could return to the market. However, given the current competition, the corporation will not do this in the near future. Sources: If you notice an error, select it with the mouse and press CTRL+ENTER. Related materials Permanent URL: https://3dnews.ru/932311 Headings: News Hardware, processors, mobile phones, smartphones, cellular communications, communicators, PDAs, SoCs and controller chips, Tags: intel, atom, broxton, sofia, 5g, goldmont, airmont, silvermont ← past To the future →

New LG smartphones will run on Intel Atom processors

Main 12:58 2015-10-26 / Angelina Gor / Phones and smartphones LG Electronics has decided to diversify the supply chain of processors for its smartphones. The vendor is negotiating the possible use of Intel chips in its products. It is planned that within one or two years, Intel Atom processors will be installed on some smartphones of the Korean vendor. According to The Korea Economic Daily, Intel and LG Electronics have begun joint development of a system-on-chip (SoC) that will later be used in LG smartphones. The companies began talks to co-create chips earlier this year, with work on the SoC starting as recently as September or October. The system-on-a-chip is expected to be manufactured by Intel using a 14 nm process. Details about the processor and the timing of its appearance on the market are not specified. LG Electronics is trying hard to become one of the leading suppliers of premium smartphones, becoming on a par with Apple and Samsung. The popularity of the iPhone and Galaxy S is due to the fact that these devices not only come with excellent software, but also have a number of unique qualities and technical characteristics. Many of the features of the iPhone and Galaxy S are only possible through the specialized hardware of these phones. While LG can develop high-quality screens and a number of other things, it can’t compete with Apple and Samsung’s engineering teams when it comes to microprocessors. Apparently, the partnership with Intel is aimed at combining high-performance x86 cores with LG’s unique intellectual property that will set the company’s phones apart from the competition. Intel currently offers two families of ultra-low power Atom chips for different types of mobile devices: the Intel Atom x3-C3000 (SoFIA) and the Intel Atom x5/x7-Z8000 (Cherry Trail). SoFIA is a low-cost development that combines Silvermont generation Intel Atom cores, a single-channel memory controller, an ARM Mali graphics core, and a range of I/O technologies. Some SoFIA chips are developed by Rockchip or Spreadtrum, all manufactured in-house by TSMC. Cherry Trail is aimed at more expensive devices, and therefore based on Airmont generation Intel Atom cores, has a dual-channel memory controller (except for the cheapest model), the latest generation Intel HD Graphics graphics core and a number of advanced I / O technologies. Next year, the company plans to release new versions of SoFIA, which will be produced at its own production facilities, as well as a Willow Trail system-on-a-chip based on Goldmont cores. While these Intel designs may be successful in some segments of the smartphone market, they hardly meet LG’s requirements for flagship models, even when tweaked to increase performance or reduce power consumption. Apparently, Intel and LG Electronics are jointly developing a special version of the system-on-a-chip related to the Broxton generation. The latter not only integrates x86 Goldmont cores, supports Wide I/O 2 multi-layer memory (peak throughput of a single chip is 68 GB/s), but is also based on a fundamentally new modular architecture that allows SoC designers to relatively easily add the necessary standard and specialized components. (as it happens with systems on a chip based on ARM technologies). Broxton will enable Intel to offer superior performance to LG as well as the ability to integrate the smartphone maker’s unique intellectual property. It is likely that initially Intel processors will be used exclusively in the flagship models of LG Electronics phones. Since the latter needs to unify its software and hardware, over time, certain Atom models may be used for other LG handsets. LG Electronics controlled about 5.8% of the global smartphone market in the second quarter of this year, according to TrendForce data. Although Intel Atom processors are used by a number of handset manufacturers, the share of Intel processors in the phone market is very small. If the company manages to negotiate with LG to use its processors inside its smartphones, Intel will capture a significant part of the market for processors for such devices. It is possible that the total market share of processors for smartphones, which will be controlled by Intel, will not exceed 10-15% in the coming years. However, Broxton’s architectural capabilities, high-performance x86 cores, industry-leading workflows, and a desire to create unique chips could enable Intel not only to increase its share, but also to gain a number of strategic advantages. In particular, Intel’s position in the markets of personal computers, tablets, smartphones and wearable devices will allow the company to strengthen the position of the x86 architecture (against ARM and MIPS), its own graphics architecture (against ARM Mali, AMD Radeon, Imagination PowerVR, NVIDIA GeForce, Qualcomm Adreno , Vivante, etc.), as well as guarantee the constant utilization of production facilities. Similar stories / 6 New iPhones will run on 7 nm processors Phones and smartphones / Nikita Svetlykh New Sony smartphones will work on the MediaTek Helio P20 platform Phones and smartphones / Stas Kuzmin 90 Apple computers005 will run on their own ARM processors Computers / Angelina Gor Yota smartphones will run on modern Qualcomm 9 chips0007 Hi-Tech / Stas Kuzmin Intel showed Android devices based on the Atom processor Company news / Vipt Huawei smartphones will work on the Russian Aurora operating system Phones and smartphones / Angelina Gor 7 Intel Medfield Platform Atom Z2460 for smartphones Contents Performance and energy consumption Conclusions Introduction It is common knowledge that the computer industry is changing very quickly. One technological leader replaces another, and such processes sometimes take place so rapidly that people do not even have time to notice and remember everything that happened for a long time. For example, who now, except for the rare active users of smartphones with Windows Mobile, at least sometimes remembers this operating system? And on whose chips was the bulk of the PDAs that appeared at the beginning of this century based? Of course, inveterate enthusiasts remember these solutions and know the answer, but the market has gone further, and now Windows Phone is not the dominant mobile operating system, and now few people know about ARM cores from Marvell (although they are still used, for example, on BlackBerry phones). The bottom line is that any of the markets in a couple of years may look completely different than now, with other leaders and outsiders, who were among the first just recently. As another example, you can pay attention to the smartphone market, where Nokia was the leader not so long ago, but now it is inferior to several companies at once. In the system-on-a-chip market for mobile devices, things can change just as quickly. For example, now the leading supplier of such solutions is Qualcomm, and many other well-known companies are very active there, but Intel is practically not represented on it. But everyone knows that this company can achieve a lot in any market if they take it seriously, because they have both the resources and (more importantly) the technological capabilities that other suppliers do not have. Until recently, x86 processors were hardly energy efficient enough for ultra-mobile devices such as smartphones. Yes, Intel has had unsuccessful attempts to enter this market. But most importantly, mobile systems-on-a-chip (SoC) will remain one of the most important and attractive markets in the next few years, and Intel will do everything to become a significant player in it. And perhaps its leader, although now it is still not too hard to believe. The SoC market for smartphones and tablets now includes NVIDIA’s Tegra, Texas Instruments’ OMAP, Qualcomm’s Snapdragon, Samsung’s Exynos, and Apple’s SoCs. The latter are not available to third-party manufacturers, but are very widespread in the company’s own products. Most modern mobile SoCs are similar in composition. Most often, manufacturers buy a license for an ARM core like Cortex-A9, as well as their own Mali graphics core or an equivalent from PowerVR, combine the ARM core with a graphics one, adding image processing, video encoding and decoding blocks, I / O interfaces, etc. d. — that’s all. But not all SoCs are built from off-the-shelf building blocks. After all, ARM is just a set of instructions, and it is absolutely not necessary to make the core exactly the same as everyone else, you can improve it or even make your own, but using the ARM instruction system. For example, Qualcomm, in the new Krait architecture, made computing cores based on its own architecture, without using the developments of ARM. And NVIDIA will soon do the same with its Project Denver, an ARM-compatible processor of its own architecture. And there is also a video core in the SoC — there are many of them on the market, and they all have different characteristics. But more often than not, companies do not design the video core themselves, only NVIDIA does this in their Tegra, and Qualcomm has its own Adreno. It’s hard to predict now which company will do the best job in the mobile system-on-a-chip market over the next few years. However, we can assume that the development team that has the most experience in designing various chips and advantages for their production will have an advantage. For example, Intel has both, and very strong development teams and their own chip factories. Yes, and with the resources they have everything in order, and enough experience. So why don’t they try to become leaders in the mobile market? Intel’s advantages Naturally, Intel has its advantages in the competition with ARM-based systems. In the mobile market, Intel is on a collision course with companies designing ARM processors. They have different goals, Intel needs to reduce the power consumption of their high-performance CPUs, and ARM-system manufacturers need to increase the performance of their computing cores, which are characterized by very low power consumption. It is likely that it will be somewhat easier to increase the power efficiency of existing high-performance x86-architecture cores than to increase the computational speed in small ARM-cores. Indeed, in the process of complicating processor architectures, companies designing ARM-based CPU cores have yet to go the same way that Intel has already left behind. In the next generation of ARM processors from Qualcomm and ARM, their computational cores should become much more complex, for better execution parallelism and more efficient use of CPU resources. ARM’s position is simpler here, since this company already has experience in designing the Cortex-A9 core with completely out-of-order execution of commands (OoO — out-of-order). But even though the Intel Atom computing core has not changed for several years, all other things being equal, this processor with in-order architecture still turns out to be more productive than any ARM-based competitors. As performance requirements for mobile systems increase, Intel’s years of experience in improving the performance and efficiency of the x86 architecture can be implemented in future versions of Atom. But ARM, Qualcomm and others will have to invest significant resources in the coming years to increase CPU performance. The x86 processors followed a similar path a few years ago, and if you do not touch on the issue of power consumption for the time being, there is no doubt that Intel can offer processors with higher performance compared to CPUs based on the ARM architecture. The same applies to bandwidth and memory access latencies. Almost all modern ARM-based SoCs have mediocre throughput. And only in the following solutions from ARM, TI and Qualcomm, high-performance memory buses are supposed to be used. But the same Tegra line from NVIDIA does not have a fast enough memory bus, which is why it is sometimes inferior to its competitors in terms of performance, having excellent computing capabilities. Intel has always paid enough attention to improving performance in general, including the memory subsystem, which is clearly seen when comparing the performance of the Atom memory subsystem and mobile SoCs in some synthetic tests — the memory bandwidth and access latency of the Intel solution are much better. They now need to deal with reducing power consumption, and not increasing productivity, with which everything is already in order. So, with performance, it’s more or less clear that Intel has experience and resources here. As for the support of wireless networks, they still have no experience in working in 3G / LTE networks, but with WiFi, everything should theoretically be in order, since they have been working on wireless networks using this technology for a long time. Well, for mobile networks, they acquired Infineon Wireless, having received an experienced team with the development of equipment for 3G and LTE networks. In addition, they now have Silicon Hive, which deals with digital signal processing solutions, as well as McAfee, an anti-virus software manufacturer, and in general, Intel’s advantage in the “software” part (compilers and libraries) is difficult to dispute. And although Qualcomm has been the leader in wireless technologies so far, in the case of LTE networks, its advantage is no longer so noticeable. In addition to Intel and Qualcomm, NVIDIA has now become a fairly strong player, having acquired Icera, a manufacturer of mobile communication chipsets. So now Intel enters the market much better prepared than it was before. Energy Efficiency and Process And here we come to an issue that we have temporarily put aside from consideration. This is a matter of energy efficiency and low power consumption, a must for ultra-mobile devices like smartphones. And so far, ARM-based systems-on-a-chip have left the Intel Atom far behind. And this is the main reason why ARM now dominates the mobile processor market. Until recently, Intel could not provide the market with a sufficiently energy-efficient system-on-a-chip. After all, performance alone is not enough, a mobile device must be able to work enough time to remain useful to its owner. Therefore, a mobile SoC should consume little power in idle mode, and when working with a computing load, the appetite should not be too immodest. Moorestown, Intel’s mobile-centric system on a chip, was first reported in 2007. In the following years, Intel showed prototype devices at Moorestown at IDF, and in 2009-2010, even ready-made solutions from some manufacturers were announced. Unfortunately, then things did not go further than this, and probably largely due to the too high energy consumption of that system. But now, with the announcement of a platform code-named Medfield, Intel has shown that it has developments that allow it to come closer in energy efficiency to existing ARM-based solutions. Manufacturers of ARM-based systems have indeed been able to achieve very low power consumption from their systems, and before the advent of Medfield, these companies clearly outperformed Intel’s early solutions for the smartphone market in this parameter. But still, a few years after the appearance of the first Atom processors, Intel was able to create a system-on-a-chip that is quite capable of competing with existing similar solutions based on ARM. But we emphasized the word «existing» for a reason. At the moment, Medfield is comparable to ARM solutions only of the current generation, which has been sold for some time in the form of ready-made smartphones and tablets in the retail market. And to assess the competitiveness of Medfield, you also need to consider the next generation of such systems. Which, most likely, will still be somewhat better in terms of energy efficiency. But don’t write Intel off again. This company owns, perhaps, the best factories in the industry, which allowed it to successfully compete with AMD even in not the most successful periods of time for itself. Medfield’s SoCs on a Chip are now manufactured on a 32nm high-k process that has very low leakage, which is why Atom has been able to match or even outperform existing 40/45nm ARM solutions. technical processes. Intel split its process technology into low power and high performance a few years ago, and Medfield is using the «low power» 32nm LP process. Intel quotes 10 times less leakage current than the best 45nm products. Compared to the previous Moorestown solution, the new Medfield consumes 43% less power or is capable of running at 37% higher frequency for the same power consumption. It is noticeably more difficult to compare it with the technical processes of other companies, but it seems that 32 nm LP at the Intel factory is much closer to 28 nm LP than to 40 or 45 nm at the same TSMC or Samsung. Well, Intel’s next step is the transition to a 22 nm process technology using the so-called 3D transistors (3D Fin Field Effect Transistor). This transition should give a significant advantage when running chips at low voltage, and 22nm chips will be about half as energy efficient as 32nm. The impact of this transition will soon be seen in Ivy Bridge desktop processors, and it will be even more important for mobile systems, because the issue of power consumption is the most important for them. And if Intel continues to improve its manufacturing at this rate and further, then it will be ahead of its competitors in manufacturing technologies by several months, which will give it a significant advantage. With higher transistor densities, the improved process technology will give Intel more options to reduce power consumption and improve energy efficiency. After all, manufacturers of systems-on-a-chip on ARM do not have their own production facilities (except for Samsung, perhaps) and compete to ensure the release of SoCs on the best technical process. Apart from Intel and Samsung, all other companies depend on external developments from companies such as TSMC and GlobalFoundries. By the way, Samsung is also not averse to outsourcing its production to other chip developers. All of these companies lag behind Intel in terms of upgrading their production. So, if Intel has been supplying chips with high-k dielectrics for several years, then GlobalFoundries did not commercially supply chips using this technology until last year. And the yield of suitable chips produced using this technology is not as high as it was planned. Intel, on the other hand, has a wealth of experience in manufacturing chips using high-k dielectric: from Penryn to Sandy Bridge, and after the transition of all manufacturers to 28 nm, Intel will already switch to 22 nm and once again benefit from clear performance and energy efficiency. Graphics cores in mobile systems In recent years, several video cores can be noted among mobile graphics solutions: PowerVR, Mali and Adreno from Qualcomm. Interestingly, the roots of PowerVR come from very old solutions of the last century, developed then for desktop PCs with x86 architecture. Imagination Technologies (at the time called VideoLogic) developed several video chips for desktop PCs at the time, which used the original 3D rendering system and were manufactured by NEC. PowerVR became the first video chips on the market to use deferred texturing technology, when only visible polygons are textured, and areas covered by other polygons are discarded. At that time, mainly classical graphic architectures were used, and in comparison with them, PowerVR technology allows you to increase the efficiency of using the memory bus by increasing the effective fill rate (fillrate). But due to numerous driver issues, as well as poor performance under certain conditions (for example, when rendering scenes with a lot of geometry), PowerVR products failed in the PC world at that time and the company left the high-performance graphics market. But for products with low power consumption and high requirements for the efficient use of memory bandwidth, their architecture fits perfectly! In the future, PowerVR stopped producing chips itself, only developing and selling licenses for its cores — as ARM did. This allowed the company to focus not on production problems, but on the architectural features of PowerVR. Since then, the original architecture has been used in various processors (including serving as the basis for some video cores integrated into Intel processors), but it has become especially widespread precisely as mobile graphics. Qualcomm and its Adreno graphics solutions are the second most important players in this market. Qualcomm bought Adreno technology from AMD, which, in turn, acquired Adreno along with ATI. And ATI got the technology, which was included in the Imageon product line, after the purchase of BitBoys, a well-known company in narrow circles. The history of which deserves a separate article. So what happens, besides PowerVR there is nothing else? So far, Qualcomm’s position is quite strong — they have Adreno 225, which will be implemented in the Krait platform. This is a very powerful solution that is clearly faster than the leader of its time, the PowerVR SGX 543MP2. Both of these solutions are clearly faster than the NVIDIA Tegra 3 video core, the graphics architecture of which, unfortunately, is already pretty outdated. And if Intel is unlikely to be able to use Adreno and Tegra video cores in their SoCs, then PowerVR always remains an excellent option for them, especially since they have shares in this company — so they made their choice a long time ago. Medfield platform As we have already mentioned, Intel has already attempted to enter the ultra-mobile device market with its Moorestown platform. Unfortunately, devices based on it did not go on sale in the end, and only one smartphone from LG was actually announced, which never saw the light of day. Unlike Moorestown, the Medfield platform is not just offered as components for smartphones and tablets, but also exists as a reference design device that can be taken as the basis for your developments. The new platform has already received support from several manufacturers such as Motorola and Lenovo. They announced deals with Intel, including the release of smartphones and tablets, and Lenovo has even shown the K800, a smartphone model for the Chinese market. Back in September last year, Intel showed off a Medfield-based smartphone and tablet at IDF running Android 2.3 and 3.x, respectively. It was then that Intel and Google announced they were working together to optimize Android for mobile Intel Atom processors. Google’s support is very important for Intel, which is still poorly represented in the mobile device market, and the joint efforts of the companies are designed to accelerate its entry into the SoC market for smartphones and tablets. In December, Intel announced the readiness of Android 4. 0 for devices based on the Medfield system-on-a-chip. The simplicity of this adaptation is evidenced by the fact that it was possible to launch Android 4.0 on Medfield within a day after the publication by Google of the source code of the OS of the new version. Drivers for tablets and smartphones based on Medfield have already been offered to OEMs, but Intel continues to work with device manufacturers to help optimize the system for their Atom mobile based products. All Medfield components are on a single chip, unlike the previous platform, which contained two chips: a processor and a Langwell system hub. The evolution of Intel’s ultra-mobile platforms can be seen in general terms in the diagram: So, the basis of Medfield is one CPU core, the maximum clock frequency of which in turbo mode reaches 1.6 GHz. Intel claims that the single-core CPU configuration was deliberately chosen, since most applications still cannot use more than one processing core anyway. However, smartphones based on the dual-core Atom should hit the market next year. It is probably then, according to Intel, that multi-threaded applications should suddenly appear. With a smaller single 32nm chip and reduced power consumption, as well as greater CPU core performance, Medfield also provides LPDDR2 memory support with twice the bandwidth, as well as improved video and still image processing capabilities. For the latter task, Medfield uses developments from Silicon Hive, which specializes in parallel computing technologies for systems-on-a-chip. This company was spun off from Philips Electronics in 2007 and bought by Intel in March last year. A more detailed map of Medfield showing changes from Moorestown. The blocks that were already in the previous chip are highlighted in blue, the changes since then are green, and the new blocks that appeared in Medfield have an orange fill: As you can see, the CPU core, display controller and camera support have changed a bit. But there are even more new blocks in Medfield, because now it has become a full-fledged system-on-a-chip, which includes a fast video core, video encoding and decoding blocks, image and sound processing blocks, an LPDDR2 memory controller, an integrated modem part and other controllers that Moorestown were in a separate Langwell chip. Instead of two chips with a total area of ​​386 mm², all blocks were placed in one chip with an area of ​​144 mm² (including 1 gigabyte of LPDDR2 memory), which is 2.7 times less. At the same time, the package of chips decreased in size from 2200 mm² to 600 mm² — more than three times. Another important difference is that the composition of the new Medfield SoC includes 1 GB of LPDDR2 RAM, unlike its predecessor, which does not have on-chip memory. Although Intel does not give the size of the main die itself (144 mm² is the entire chip, and even with memory), but for the Atom Z2460 its area is approximately between Tegra 2 and Tegra 3 from NVIDIA, that is, between 90 and 49 mm². For a single CPU and a PowerVR SGX 540 video core at 32nm, that’s not a lot, given the five CPU cores in Tegra 3 and a slightly faster GPU, as well as the 40nm process technology. Of course, the same systems-on-a-chip from Apple have a large area, but they are also based on dual or quad-core video systems SGX 543 MP2 or SGX 543 MP4 and have two Cortex-A9 cores. Mobile devices based on the Intel Medfield platform with Android OS are expected this year, and the first of them were shown at the January CES 2012 exhibition. Thus, the prototype of the Lenovo K800 smartphone, which should go on sale one of the first, was shown at the exhibition. Another serial device on the Medfield platform should be the Orange London smartphone, and in the future there will be other products with Android 2.3, 3.x or 4.0 as the OS, and the choice of version remains with the device manufacturer. According to Intel itself, mass sales of such devices will begin in early summer. Enhanced Compute Core Medfield is the name of the platform, the specific SoC model is codenamed Penwell, and the CPU core inside it is called Saltwell. In itself, it is almost the same as the Bonnell core in the first Intel Atom processor, although it still made several changes aimed at increasing performance and reducing power consumption. The Intel Atom architecture was announced about five years ago, and it hasn’t changed much since then, so you can read the material from years ago and understand what kind of core is placed inside Medfield. This is still the same in-order architecture, which has support for Hyper-Threading virtual multi-core technology and is capable of simultaneously executing up to two instructions per clock. This is a fairly simple kernel that does not have dedicated integer multiplication blocks, they use the same hardware blocks as floating point calculations. To support Hyper-Threading, some of the blocks (like the register file) are duplicated, but the number of execution units in general is not too large. The Atom integer pipeline has 16 stages, double that of the Cortex-A9. Such a long pipeline in the Intel mobile processor is used to achieve acceptable power consumption. Intel’s approach to building a mobile Atom core turned out to be justified in the end, since in performance tests even one Saltwell core can quite cope with dual-core counterparts based on Cortex-A9. It appears that while the architectural difference between Atom and Cortex-A9 cores isn’t that big, some clever architectural decisions in Intel’s mobile core give it an edge. Perhaps Intel is right that while the Android operating system itself is quite parallelized, most of the common applications for it at the moment cannot take full advantage of the two or more processing cores already available in SoCs using the ARM architecture. Possibly when running at high frequency and loading two Cortex-A9 corescorresponding systems-on-a-chip and will be able to stand on a par with Medfield in terms of speed, but such use so far has little to do with reality and is artificial. But systems based on Cortex-A9 have been produced for quite a long time, and SoCs with Cortex-A15 cores should soon come to replace them. And when that happens, without significant architectural changes to Atom, Intel may already have some problems. After all, Cortex-A15 is a much more complex and modern core, also out-of-order, like A9, but noticeably more performant (ARM estimates up to 40% faster under equal conditions). Most likely, systems-on-a-chip based on Cortex-A15 cores will be able to surpass Medfield in its current form. Yes, and Qualcomm’s Krait should also be about the same as Cortex-A15, so the current advantage of single-core Medfield over dual-core ARM may disappear pretty soon. But for now, this advantage is there, and let’s see what Saltwell has done to improve its competitiveness. In terms of performance, there are minimal changes in it. Only minor tweaks have been made, such as increasing the speed of copying data from / to memory, improvements in the transition predictor, etc. We can also note the support for on-chip 1 GB LPDDR2 memory operating at a frequency of 400 MHz, which gave a double increase in throughput compared to the previous solution. A lot has been done at Saltwell to improve energy efficiency. As in other Atom models, the core has a 512 KB L2 cache, but there is also a separate static memory with its own power line. This memory stores the state of the CPU and data from the cache when the compute core itself goes into low power mode, which improves its energy efficiency bottom line. Even more interesting is the Intel Burst Performance (BPT) technology, which dynamically increases performance when needed. Similar solutions with turbo frequencies are now used everywhere, many manufacturers have similar technologies. BPT can temporarily increase the operating frequency of the CPU core to a set limit, as long as it does not go beyond a certain power consumption. The Medfield Universal Computing Core operates in several modes that differ in operating frequency and power modes. Its frequency and voltage are effectively scaled over a wide range, from C6 idle mode and low frequency mode (LFM) to high freq mode (HFM) and peak load mode at the highest possible frequency. for the CPU. Saltwell shows the lowest consumption level in C6 idle mode, in which the core and L2 cache are powered off (power gated) and their state is stored in the on-chip static memory. Therefore, the overall level of energy consumption in C6 mode is close to zero. The technology itself, called Intel Smart Idle (Intel SIT) is not new, the company has previously done similar functionality in Nehalem, for example. In active mode, the processor core is capable of operating at frequencies from 100 to 1600 MHz in 100 MHz increments. Moreover, the frequency of 1.6 GHz is a turbo frequency and cannot be maintained for a long time, just like Turbo Boost in Sandy Bridge desktop processors. The maximum frequency at which continuous operation is possible in the case of the first model of the Medfield platform is 1.3 GHz, but quite often it rises above this limit. Apparently, the output of suitable Atom Z2460 chips using the 32 nm LP manufacturing process has further improved, and now the maximum turbo frequency limit for them has been raised to the bar of 2 GHz, while maintaining the maximum «regular» frequency of 1.3 GHz. Probably, we can expect a further increase in frequencies for other models, but only in turbo mode, and in normal mode, the frequency will still be limited by the level of energy consumption. Video core and video data processing Instead of the PowerVR SGX 535 video core, known from Moorestown and Lincroft, a more advanced version of the GPU, the SGX 540 model, has become part of the system-on-a-chip of the Medfield platform. And although the fill rate (fill rate) of these video cores the same, but the updated video core is able to process more triangles and pixels. Its geometric performance has increased from 31 million triangles per second to 40 million, and its pixel shaders can run up to two times faster. The PowerVR SGX 540 video core is known to us from several systems-on-a-chip, such as TI OMAP 4460, except that in the chip from Texas Instruments it operates at a slightly lower frequency than in the first chip of the Medfield platform. The graphics processor used in the Atom Z2460 runs at 400 MHz, OpenGL ES 2.0 and OpenVG 1.1 are supported from the graphics API, so there is nothing new here either. It seems that the use of not the newest GPU model from PowerVR is due to the fact that Intel brings its mobile SoCs to the market a little later than previously planned. And fortunately, the upcoming dual-core version of Medfield already has a much more powerful PowerVR SGX 544MP2 video core. Three displays are already supported by Medfield chips for displaying images: two internal and one external (HDMI), instead of one internal and one external in Moorestown. HDMI 1.3a port with HDCP 1.2 is supported, the maximum resolution of the built-in screens is up to 1280×1024 (at 18-bit per color and 60Hz), and the external one is up to 1920×1080 at 30Hz. By the way, this resolution of the internal display may already be insufficient for tablets, especially considering the appearance on the market of models with a resolution higher than FullHD, which is simply not supported by Intel’s system-on-a-chip. But, in addition to the usual wired image transmission, Medfield supports wireless image transmission using Intel Wireless Display technology or WiDi for short. It allows you to transfer image and sound from a mobile device to a TV or monitor via WiFi. Of course, to transfer via WiFi, you have to compress the image on the fly, if it is not already a compressed Blu-ray video stream, and the quality will eventually be somewhat worse than with a “wired” connection. Plus, although resolutions up to 19 are supported20x1080, but at only 30 FPS. That is, WiDi is not well suited for all types of activities, but for many types of entertainment, such as movies and leisurely games, it is. Anything can be output from the phone to the screen, along with two-channel audio in 48 kHz LPCM format. Naturally, for this you will have to purchase a special WiDi-certified receiver for receiving a signal and outputting it via HDMI to an image output device, because most of them do not yet have built-in WiDi support. This technology is supported in the latest versions of the Android operating system, you can use the second monitor as a clone of the main one or stretch the desktop to two output devices. But in order to display some kind of video clip, you must first decode it. And from our numerous practical materials, you know that many SoCs in real use have problems with supporting hardware decoding of different formats. Suffice it to cite devices based on NVIDIA Tegra 2, which are not always able to decode video data in some formats in hardware. In Medfield, hardware support for encoding and decoding video data has been improved — thanks to the inclusion of video data processing blocks from Imagination Technologies in the SoC: VDX385 hardware decoding block and VDE285 video encoding, it now has full support for Full HD. If Moorestown was able to speed up the decoding of a single 1080p video stream and only 720p encoding, then Medfield can encode in 1080p, and decodes two 1080p streams at the same time. The Intel Medfield system-on-chip supports hardware decoding of the following video formats: H.264 Baseline (L3), Main and High (L4.2, up to 1080p at 60 FPS), MPEG4 Simple profile (L3), Advanced simple profile ( L5), VC1 Simple (Medium), Main (High), Advanced (L4 up to 1080p at 60 FPS), WMV9 Simple (Medium) and Main (High), and H.263 Baseline (L70 level). Declared support for decoding video data with a bit rate up to 50 Mbps, processing of two channels in FullHD format is supported, which will be useful for stereo mode and video conferencing. Medfield system-on-a-chip hardware encoding of video data is possible in the following formats: H.264 Baseline profile with bit rate up to 20 Mbps and resolution up to 720p and 1080p at 30 FPS, MPEG4 Simple profile with bit rate up to 10 Mbps and resolution up to 720p at 30 FPS, as well as the H.263 Baseline profile with bit rates up to 8 Mbps and resolutions up to 720×240 at 60 FPS or 720×288 at 50 FPS. Although this is a pretty good level for a modern system-on-a-chip, there is nothing supernatural in such support. Camera support and image processing For mobile devices, it is very important to have good support for taking photos and videos with built-in cameras, and even with their advanced processing. And Medfield has improved this support — if the previous Moorestown supported cameras with a sensor size of up to 5 megapixels, then the new SoC can process data from a 16-24-megapixel main camera (provided that such sensors are installed by manufacturers) and a 2-megapixel additional one. For the main image sensor, the MIPI CSI x4 interface is used, for the additional one — MIPI CSI x1. This is more than enough, since the front cameras usually have no more than two megapixels and are used only for video conferencing and telephony. Some software and hardware features are declared for Medfield that improve support for photo and video shooting, such as high-quality noise reduction, support for high frame rate, which allows to achieve high speed continuous shooting, as well as high-quality stabilization in two axes for photo and video shooting. Naturally, autofocus, autoexposure and auto white balance are also supported. The main camera has a maximum processing performance of 15 fps from the 16-megapixel sensor, and video capture and recording up to FullHD at 30 FPS or 720p at 60 FPS. But if you don’t really surprise us with such video recording, then the possibility of continuous shooting at a speed of 15 FPS is quite rare for such a compact device. This speed allows you to catch any moment that is too fleeting to have time to press the camera button just once. Unfortunately, the speed of 15 frames per second is supported only for 10 consecutive frames, but from them it will be possible to choose the most successful shot in full resolution (see illustration). The programmable Imaging Signal Processor is responsible for most of the noted functionality in the system-on-a-chip. It is he who deals with noise reduction, image stabilization, continuous shooting frame processing, automatic white balance, face detection, digital zoom, chromatic aberration correction, etc. Intel uses an image processor with a performance of up to 240 megapixels per second in its mobile chip. According to the company, its capabilities are on par with entry-level SLR processors. And it is quite possible to believe in this, since not every “reflex camera” is capable of burst shooting at a speed of 15 frames per second, and not all cameras can shoot video with 60 FPS. But how does the chip have such advanced capabilities, if Intel has not been engaged in image processing before? We have already mentioned above in the text about the company they acquired Silicon Hive — it is their developments that are used in the image processor at Medfield. Atom models for ultra mobile devices A total of three Atom-based SoC models are currently planned for release: the Z2000 for low-end devices, the Z2460 for smartphones, and the higher-end Z2580 featuring a dual-core CPU and GPU. All of them are or will be produced using the 32 nm process technology with High-K/metal gate and differ from each other both in the number and frequency of computing devices, and in their capabilities. Model mid-2012 2013 The most interesting for us today is the Atom Z2460, which will give life to the first mass x86 smartphones. As we noted above, it is based on a single Saltwell core with Hyper-Threading support and its maximum turbo frequency is 1.6-2 GHz. A dual-channel LPDDR2 memory interface is used, and the PowerVR SGX 540 video core operates at 400 MHz. This solution incorporates the Intel XMM 6260 radio part, which supports HSPA +, but not LTE. The Atom Z2580 is the top model in the line, more than double the performance of the Atom Z2460, and includes an improved communication module that supports not only 3G and 2G, but also LTE. The base clock speed of two CPU cores with support for Hyper-Threading (two cores, four threads) is 1.3 GHz, and the turbo frequency can reach 1.8 GHz. The PowerVR SGX 544MP2 dual-core graphics core in this model is clocked at 533 MHz, which will greatly improve 3D performance over the Z2460. The SGX 544 is similar to the SGX 543 used in the Apple A5 (based on the iPad 2 and iPhone 4S), only Direct3D 9 support can be noted..3. With a GPU speed of 533 MHz, this SoC will be up to two times faster in 3D performance than the SGX 543MP2 in the Apple A5. Everything would be fine, but the A5X system-on-a-chip already has a quad-core PowerVR SGX 543MP video core, which is even faster. The Z2580 platform includes the Intel XMM 7160 LTE Enhanced Radio Unit, which is an upgraded version of the XMM 7060 with support for 3GPP Release 9 from the respective mobile telephony specification consortium. This solution supports many mobile communication standards: 2G, 3G, LTE, HSPA+, UMTS. This chip has already begun to be produced, but finished products based on it will appear only in the first half of next year. For use in low-end phone models, the junior model Intel Atom Z2000 is offered, which includes one CPU core operating at a frequency of 1.0 GHz and has neither virtual multiprocessing support nor operation at an increased turbo frequency. This low-cost solution should compete with low-cost systems-on-a-chip based on ARM architecture cores. The Z2000 has an SGX 540 video core running at 320 MHz (instead of 400 MHz on the average Medfield model), and video encoding capabilities are limited to 1280×720 (720p), although 1080p decoding is still supported. The model has an Intel XMM 6265 radio with support for HSPA + and two SIM cards, and the appearance of finished products based on this system is expected at the end of this or early next year. As you can see, it is planned to produce not only expensive devices based on Medfield. Thus, a ZTE smartphone based on the Intel Atom Z2000 system-on-a-chip, specially designed for use in budget devices, will cost about $160. Interestingly, all Medfield models have their own reference designs from Intel. For example, the reference device design on the Atom Z2000 has a different form factor compared to the reference devices on the Z2460 and Z2580, which is quite natural, since they are aimed at completely different market segments. And although most major manufacturers always make their design for mobile devices, reference products can be produced by mobile operators under their own brands, which will happen in the case of Atom Z2460. Smartphone reference design based on Atom Z2460 Smartphone reference designs for their SoCs were taken up by Intel for a reason, because it helps a lot in promoting the device to the market, even if individual manufacturers do not have the due enthusiasm for using them. The history of reference devices is closely connected with Mike Bell, formerly of Apple and Palm, under whose leadership the reference devices based on Medfield were created. Providing reference designs helps manufacturers by reducing development costs and shortening design times, just as it does with motherboards, for example. Years ago, Intel began designing and supplying motherboard manufacturers with reference board designs based on their chipsets. And they modified the reference design to suit their needs, making changes that added originality to the boards. In addition to reducing the time to market, such a strategy also leads to the fact that manufacturers of specific products do not deal with the same tasks for everyone, but do only what distinguishes their solutions from each other. The reference platform based on the Atom Z2460 does exactly the same thing for smartphones based on this system-on-a-chip from Intel. The company provides its partners with a solid and stable design of the smartphone, thereby reducing the hardware development time for them. And manufacturers have more time for other goals, including visual design, software development, and more. brands, as has long been done with many other high-tech products (and it is common in the mobile phone industry). Intel has attracted all the necessary resources to not only bring the Medfield platform to market on its own, but also to make very good reference smartphones based on Atom chips, which can be sold to third parties for sale under their own trademarks. As we noted above, since the modem part of this system is Intel XMM 6260, it supports HSPA +, but not LTE. Interestingly, the controller from Texas Instruments is responsible for WiFi communication in the prototype, since Intel’s wireless division does not yet have a ready-made solution with low enough power consumption for use in smartphones. But in the future, of course, they will appear. The Medfield reference platform, which was recently reviewed on our website, is available to Intel partners and they are already using it. Even the original models, like the Lenovo K800, are based on a modified version of the reference design from Intel, not to mention the simply branded operator phones from European Orange and Indian Lava. And that’s not all, pretty soon we can expect the appearance of other models based on the new system-on-a-chip from Intel. Software Compatibility and Binary Translation Compatibility is one of the most important issues in Medfield’s entry into the Android smartphone market. After all, most Android devices have ARM-architecture CPU cores, and not all applications will easily run on x86 without modifications. Although a lot has changed with the advent of Android 4.0, devices have appeared on processors with the MIPS architecture, for example. But their users immediately felt their inferiority, as some of the most common applications (for example, games) simply did not run on these mobile devices. No wonder that Intel has taken up this issue very closely. Back at IDF 2011, Intel and Google announced the start of cooperation to optimize the Android operating system for smartphones based on the x86 architecture. The OS itself isn’t a problem, as native x86 support was already introduced in Android 4.0, aka Ice Cream Sandwich. The joint work of the companies is to optimize the main applications for x86 and port existing software that uses the NDK. Intel has also developed power management software specific to their hardware. But the company is engaged not only in optimizing the Android code for their systems, but also fixing general errors in it, providing information about changes to the Android Open Source Project (AOSP), which deals with the overall development of this operating system. With these actions, Intel is helping to improve the OS even for ARM-based platforms. But the problem is not in the operating system, because all versions of Android will now be released for both ARM and x86. The main problem is the incompatibility of some applications. Moreover, already existing, released before x86 architectures entered the smartphone market, since software developed from scratch can already and should be developed simultaneously for different processor architectures. Most existing Android applications use the Dalvik virtual machine (the applications themselves are written in Java, but use a different bytecode format), and they do not need a specific CPU architecture to run at all. But some of the common applications and games use the Android Native Development Kit (NDK), a package of libraries for developing applications in C ++, which is necessary when developing performance-critical code. The main compatibility issue occurs when an application is specifically written for ARM and uses native code for that architecture. Typically, such applications are CPU-demanding games and some other types of software. Most Dalvik applications run on Intel Atom without any modifications, but «native» applications using the NDK need to be recompiled to run on x86-compatible systems, and some of these applications will require additional work to port the code. But there is nothing to worry about here either, Intel is actively working with all major application developers in order to port their products to the x86 architecture. In order to prove to potential users of x86 smartphones the viability of their ideas, Intel showed the work of one of the most important mobile applications on its smartphone — the game Angry Birds. It just uses the features of ARM processors and under normal conditions does not work on CPUs of other architectures. There is no doubt that Intel will do everything to make the vast majority of Android applications work on their products, even those that were originally compiled for a different processor architecture. Intel estimates that approximately 75% of all Android apps in the Google Play Store (formerly Market) do not use ARM-specific code and should theoretically run on x86 without issue. But the remaining quarter of applications is the main compatibility problem. So how did Intel make it so that everything works «out of the box», as they say? After all, the example of low-cost devices based on MIPS systems with insufficient compatibility is quite recent, and it is unlikely that anyone will want to repeat this sad experience. And it can take a very long time to wait for optimization of all applications for x86 from the software developers themselves. For such programs, Intel decided to use binary (binary) code translation. Collaboration with specialists from Google made it possible to ensure that even ARM-native applications on smartphones with Intel Atom run and work without the need to port them. Dynamic binary translation allows most NDK applications to run on Medfield and provides sufficient performance for normal use. Intel doesn’t really reveal the details and capabilities of binary translation, but it can be assumed that the compiler dynamically intercepts ARM code and translates it to x86 right on the fly, as they say. At the same time, Intel understands that they are unlikely to achieve 100% compatibility. It is assumed that with the help of binary translation, the share of compatible applications by the time devices enter the market will grow to 90%. Of course, dynamic translation itself is slower than the execution of «native» code for the CPU, but Intel is quite optimistic, and we can only believe and (if possible) check. Interestingly, all applications should be available in the Google Play store on x86 smartphones, just like on ARM devices. Usually, on mobile devices with a CPU of a different architecture that does not have the ability to binary translation, some of the applications are not available for download and installation. But in the case of products based on Intel Medfield, by the time they appear on the market, most applications (including those using ARM code) will work on them as on ARM devices. Performance and Power Consumption The importance of high CPU and GPU performance for today’s smartphones is unlikely to be denied, but even more important are energy saving technologies and, as a result, high energy efficiency and battery life. After all, few people will be satisfied with a phone that cannot “live” without being connected to the mains for at least one day. This is confirmed by the requests of users who most of all want more battery life: Now, after all, it just turns out that most smartphones with more or less active use (but without extremes, because active web surfing over 3G or a 3D game can drain the battery in a few hours) are able to work autonomously for only 1-2 days. Accordingly, one of the main requirements of users to phones is battery life. And this is the most important task of all manufacturers, they are all trying, if not to increase this time, then at least not to reduce it. What can we say about the performance and energy consumption of Medfield now, if we don’t have the final products yet? Practical comparisons are yet to come, but first, let’s look at the data from Intel itself. They give diagrams without numbers and specific models of competitors’ smartphones, in benchmarks and conditions of use chosen by them: phones, and Medfield products are just getting started. However, we note that if these figures are to be believed, the Atom Z2460-based reference device shows better performance and at the same time consumes less power than most of the compared incognito smartphones. Let’s not question the high results of the Intel solution in such tests as SunSpider and BrowserMark, because they are really strong, as the first tests of the reference platform showed. Where did such a difference come from? Typically, large gains in these tests come from optimizing Internet browsers rather than using more powerful systems-on-a-chip. Intel itself claims that their solution has several advantages. For example, although Cortex-A9 cores are quite high-performance by themselves, in many tasks their speed is limited by the performance of a slow memory interface. And Atom has an advantage in case of misses when reading from the cache memory, «getting» the data necessary for calculations faster. But if this is the only thing, then the imminent release of SoC based on Cortex-A15, it is quite possible that it will be able to change the situation. But so far, Medfield wins in such tests. Although its tests may well be helped by turbo mode, because in CPU benchmarks the core can be accelerated to 1.6 GHz for a short time, and it is difficult to determine here whether the architectural advantage in terms of memory bandwidth over ARM was affected either simply by the higher frequency of Atom and the single-threaded nature of the tests. The performance of the PowerVR SGX 540 GPU in other tests is not as good as depicted in the GLBench 2.1 graph. In other GPU tests, its performance surpasses only older systems or SoCs that have similar PowerVR video cores, but operate at a lower frequency. For example, in TI OMAP 4460 the video core operates at 300 MHz, and in Medfield at 400 MHz, so the winner in this pair will become clear whose decision. Compared to other systems, more modern GPUs like the Mali 400 MP or Adreno 220 are much faster than the Medfield solution, not to mention the quad-core PowerVR SGX 543 MP4. It is more difficult to compare with the NVIDIA Tegra video core, this traditional 3D architecture is very strong in some mobile tests, but seriously loses in others. In general, it must be said that there are very few high-quality 3D tests for Android, all of them are rather controversial and the generally accepted analogue of the «desktop» 3DMark is clearly not enough. Talking about the performance of mobile SoCs designed for smartphones makes no sense without taking into account data on power consumption. And here Intel has something to please potential buyers, Medfield really looks quite competitive. So far, we only have data from the company itself, but they are impressive. The consumption of Intel’s new system-on-a-chip is comparatively low under different conditions. By itself, one CPU core of the Medfield platform consumes about 50 mW at 100 MHz, 175 mW at 600 MHz, 500 mW at 1. 3 GHz, and up to 750 mW at maximum 1.6 GHz. In idle mode for the entire chip, the total consumption figure does not exceed 20 mW, while transmitting voice data over a 3G network, the chip consumes up to 750 mW, and in web surfing mode, about 1 W. That’s even a little less than 1.2-1.3 W for today’s top smartphones like Samsung Galaxy S II or Apple iPhone 4S under the same conditions. Yes, and most other systems-on-a-chip based on ARM cores consume about 1 W at the same load, so Medfield clearly managed to stay at a good level in terms of CPU consumption. Good, but still current — compared to SoCs from other manufacturers, made using 40/45 nm process technology. What will happen next — only time will tell. Most likely, future ARM chips for smartphones will consume the same 1-1.5 W as much as possible, but at the same time they will have noticeably more efficient CPU and GPU cores. If we move on to a more familiar assessment of the battery life from a single battery charge, then the numbers from Intel are as follows: up to 10 days in standby mode, up to 48 hours of music playback and about 4-5 hours of high-definition video playback or web surfing. This is a significant improvement for x86 systems, and the reduction in power requirements is due to the high integration of the new system-on-a-chip, since all the functionality of the chipset for previous Atom models was integrated into one compact chip manufactured using modern technologies. Medfield’s high energy efficiency is to blame for its advanced power management system, which we already talked about above. Blocks that are in idle mode are disconnected from power and consume almost nothing. After all, when listening to music, you do not need a display or a CPU operating at a high frequency, and in idle mode, you only need to keep in touch with the base stations of the mobile network, and everything else can be turned off. And it looks like Intel has been able to catch up with ARM-based systems in energy efficiency, for which it can only be commended. Conclusions The battle between Intel and companies using the ARM architecture has already begun. All of them are trying to win back the markets occupied by rivals. Intel is targeting the mobile market with its new Atom systems-on-a-chip, and ARM and its licensees are planning to break into the Intel-dominated market of desktops, mobiles, and servers. And gradually this battle becomes more and more heated. So, after a wave of success of tablets based on ARM-architecture chips, the so-called ultrabooks came into fashion, where more powerful solutions from Intel are installed — here the company feels in its own field. Well, they already have some success with smartphones — so far in the form of agreements with Motorola and Lenovo, which are going to release smartphones and tablets based on the new Atom models discussed in the article. On the other hand, HP and some others announced their intention to use ARM architecture chips in their servers in 2012. It is planned that such servers will be cheaper and the cost of their operation and maintenance (electricity, cooling) will be less. For application in servers, future 64-bit solutions planned by ARM will also be useful. And Qualcomm is still negotiating with mobile PC manufacturers to produce thin and light laptops running Windows 8, based on a Snapdragon S4 quad-core system-on-a-chip. That is, in all markets, the strategies of these companies clearly intersect and go against each other, and not everyone will benefit in the end. Most likely, not the most calm stage of development awaits us, when either one company or another will have an advantage. In any case, in the segment of high-performance desktop and mobile CPUs, Intel’s business now seems unshakable, and with the release of Medfield, they are making a good claim in the «smartphone» and «tablet» segments, not to mention the solutions of the future, when the technological advantage of the company begins to affect . We hope that buyers will only benefit from active struggle — after all, in conditions of fierce competition, progress is usually faster and prices are lower. Intel’s announced Medfield solutions are a more serious entry into the smartphone market than the previous Moorestown solution. The close attention to this market is understandable, because smartphones are becoming almost the most important computing devices for a large number of users. You don’t always carry a PC or laptop with you, and a smartphone usually lies in your pocket or purse. But in the phone market, smartphones still have not taken a dominant position, in 2009only 17% of the US phone market was occupied by smart phones, and in 2011 it was already about half. And this growth continues. With the release of Medfield, Intel has finally done what it can to help it enter the ultra-mobile market. After years of talking about them being about to release smartphone products, Intel has finally released a truly competitive product. Moreover, in the form of a reference design device, which will seriously help in promoting x86-compatible systems-on-a-chip. Moreover, Intel now has two strong mobile partners who plan to develop and release smartphones and tablets based on Intel Atom «smartphone» models. The main stream of announcements and market launches is yet to come, but it is already clear that the success of Medfield in the market will clearly surpass the “success” of Moorestown. The reasons for this potential success are simple — the fact that Medfield turned out to be a fairly powerful system-on-a-chip already solves a lot. And if this platform had entered the market six months or a year earlier, then it would simply become the best on the market and would conquer most of it. But even now, the single-core Atom Z2460 delivers high CPU core performance, which is higher than currently sold competitors in several recognized benchmarks. Although the GPU performance of the part is not as high as some competitors like the Adreno 220 and Apple A5X, it is quite competitive and sufficient for most current smartphones. In addition, the Z2580 model with a dual-core PowerVR SGX 544MP2 video core running at a higher frequency will soon solve this problem. But the Medfield platform and the Atom Z2460 model, in particular, stand out not only for performance. It turned out that the rumors about the problems with the high consumption of x86 computing cores are greatly exaggerated. In the case of the first Medfield model, it shows excellent energy efficiency. And if we still don’t know the energy consumption of end devices, then Intel’s system-on-a-chip turned out to be quite economical for use in smartphones. Of course, this is due not only to the work done to improve the energy efficiency of the SoC itself and its components, but also to a more subtle technological process compared to competing chips. But since Intel has such an advantage, why not use it? Moreover, it is precisely due to the advantages of Intel mentioned in the article that there are good reasons to believe that the Medfield platform and Atom Z2460 will become for the company only the starting point for an excellent start in the smartphone market. Intel already has a good chip, which is not only available to partners in its pure form, but is also offered as a well-developed reference design device, which can still lead to interesting consequences on the market. After all, if mobile operators sell reference phones of such a powerful company under their own brands, then they will be able to win a decent part of the market. But it’s definitely too early for Intel to relax, because it’s one thing to compare the solution with ARM systems that have been on sale for several months, but it will be a completely different thing when multi-core systems-on-a-chip based on modern Cortex-A15 computing cores come out, as well as with more powerful GPU cores from other mobile SoC manufacturers. You and I can only watch this fight for our wallets — and you can be 100% sure that it will be interesting! Intel Atom: a small favorite of a big corporation 05/11/2010 We are used to the fact that Intel Atom processors are used only in netbooks. However, the stereotype will have to be parted very soon: Intel plans to put «atoms» in smartphones, cars and even cash registers. In general, everywhere where ARM architecture processors are now in charge. But can Intel do it? On April 13, Intel Corporation announced record financial results for the first quarter of 2010. Usually at this time, the market «rests» after the rapid sales in the fourth quarter of the previous year, when buying activity peaks due to a series of holidays and the desire of the relevant departments of large companies to use the allocated budgets to the end. But in 2010, everything turned out differently, and in terms of net profit, the first quarter outperformed its predecessor by seven percent, reaching $2.4 billion. As for revenue, it amounted to $10.3 billion, and according to this indicator, we are dealing with the best first quarter of Intel in the 21st century. Let’s not delve too deeply into financial figures, since everyone can do it on their own by following the secret link. Let us just note that, despite the general growth, revenue from sales of Intel Atom processors and chipsets fell by as much as 19%, even compared to the not very successful first quarter of 2009, when the growing financial crisis forced mobile computer manufacturers to reduce production volumes. The impression of such a drop is reinforced by the news about the decline in sales of netbooks in general: they say, people, having played enough with low-power «atoms», are increasingly choosing compact models, albeit more expensive, but more productive. Skeptics have been talking about how Atom has done its job, planting in the minds of the need to get another computer, small and inexpensive, and now can leave the scene. Fortunately, even the ULV version of the Core 2 Solo from a year ago fits into a TDP of 5.5 watts, while demonstrating noticeably better performance. And if you transfer them to a 32-nanometer process technology … Modern Intel Atom processors based on the Pine Trail core Looking ahead a bit, it should be said that the prospects for Intel Atom in its current state are really very doubtful. And one could join the chorus of gravediggers, if not for one nuance. Both outside and inside Intel are receiving signals that the corporation is making a very serious bet on the platform, transferring marketing mechanisms primarily to the priority support of Intel Atom. And not only and not so much for its use in netbooks: according to Intel’s idea, «atoms» should appear in cars, and in TVs, and in set-top boxes (not to be confused with game ones), and in MIDs that are experiencing a rebirth with the release of the iPad , and even … in smartphones. Most Profitable Small Processor Before twirling your thumbs at your head saying “what other x86 smartphone?”, let’s take a breath and look back a bit. As we know, Intel Atom is, in fact, a slightly simplified and modernized Pentium III, made using 45-nanometer technology. That is, the costs of R&D are obviously not exorbitant, the cost of production is very low (due to the simplicity of the design and the well-established technology). However, the Atom N270, the most popular in netbooks, sold for $44 in lots of 1000 from the very beginning. And not only did it sell excellently, but also the Intel® 9 chipset was certainly bought with it.45GSE Express with ICH7M. The latter was announced at the beginning of 2006, and had become obsolete even by the time the first «atoms» appeared. However, thanks to his little friend, he is still on sale, and the N270 processor and chipset bundle costs $83 in quantities of 1000 pieces. Classmate PC is Intel’s ambitious project to promote sales of Atom processors. Interestingly, school netbooks are assembled not only in Asian countries, but also, for example, in Portugal. Considering sales volumes (about 50 million «the world’s smallest x86 processors» have been sold so far) and low production costs, within Intel itself, the Atom direction is considered not just profitable, but super profitable (of course, not in absolute numbers, but in terms of cost/income). And the drop in sales is not a disaster, but a signal. They say, no matter how good the idea is, you can’t extract superprofits from it indefinitely. After all, even for some Internet pages with flash inserts, the power of the N270/280 is not enough, and even such a luxury as watching HD video from a netbook is out of the question. In addition, everyone who wanted small mobile computers has already acquired, and the explosive demand in this segment is replaced by quite ordinary and calm. And here it’s not up to high prices for old chipsets. Plus, do not forget about the extremely interesting and growing popularity of the Nvidia ION platform, which has recently been updated to the second version. This «kit» not only makes Atom quite suitable for 3D games (though in most cases, with minimal graphics options), but also helps the kid in video encoding and illustration processing, shifting some of the load to the video adapter. Is it worth specifying that Intel treats the success of a competitor in the field, which was previously considered undividedly its own, without much enthusiasm? It is clear that only very thoughtful guys who see the future a couple of parsecs ahead are trusted to write such articles. But Intel also has such people, and they began to think about the fate of Atoms much earlier. According to some reports, back in 2006, when the corporation sold its division that produced XScale processors on the ARM architecture. After the announcement of the first Atoms in 2008 and the promises made in connection with this in the near future to settle them everywhere, including household devices, there was a feeling that the march to the fiefdom of ARM would soon begin. However, for almost two years everything was quiet and, perhaps, the only device where «atoms» seriously pushed ARM out was NAS. And even then only in the upper and partially middle price segments. But in 2010, the offensive began in earnest. After all, if Atom sales go from millions to billions (like competitors with ARM architecture), and profitability remains at a good level, the concept of “record revenue” at Intel may undergo significant changes. Phones, cars, then everywhere? The first herald of change was the smartphone LG GW990, demonstrated at CES 2010. The device, as a device, is only a little big, but it’s a normal variation on the theme of iPhone and LG Chocolate. It just worked on the Atom Moorestown platform running the Intel Moblin operating system. Compared to the previous version of the economical Atom, Menlow, power consumption under load has decreased by half, and in standby mode by as much as fifty times. Paul S. Otellini, President and CEO of Intel, at CES 2010 introduces an LG smartphone with an Atom processor inside Perhaps this allowed LG to promise that the communicator with a 4.8-inch screen will be able to work 4 hours of talk time or 300 hours of standby time from a battery capacity of 1850 mAh. True, it is easy to detect some … cunning in the numbers, but about it at the very end of the article. Atmophone LG near Back at CES 2010, they announced the launch of Intel AppUp, an app store optimized for the Intel Atom architecture. Optimization can be very different — from reducing power consumption by turning off unnecessary beauties of the program to optimizing the interface for a typical netbook screen resolution. Mere mortals can download a special client program (there are versions for Windows and Moblin), and with its help, with a couple of clicks, buy the programs they like, which will be installed directly on the netbook. Does this software trading model remind you of anything? Yes, I have exactly the same associations. Main page of the site on which Intel plans to build its App Store Further, more. At the end of February, Intel allocates a million dollars to stimulate the development of software for Atoms. In the first phase, which ran from March 1 to March 31, the money was literally given away in chunks of $500 for each application registered with the Atom Developer Program. Then, among the registrants, a hundred programs will be selected, the authors of which will be paid extra for each download, regardless of whether the application is paid or not. Finally, very good products will be chosen among the very best, and their authors will receive cash prizes from 10,000 to 50,000 dollars. It makes no sense to go into details, since everything is listed point by point at the link. Suffice it to say that Intel has never before been seen in the mass and open support of consumer software development. If a corporation suddenly decided to take the path of Apple, Google and, to a certain extent, Nokia, by leaps and bounds, this is alarming, right? Speaking of Nokia. As we remember, on February 15 at MWC 2010 it was announced that the Finns are joining forces with Intel, and a new MeeGo mobile platform will appear based on Maemo and Moblin. So, according to very reliable rumors, the first beta version of MeeGo with a graphical interface will be released in mid-May, both for ARM (with an eye, of course, on Nokia N900) and for Intel Atom. LG has already hinted that the GW990 shown at CES will work on MeeGo. In parallel, there were rumors about the upcoming «atomic» smartphone manufactured by Lenovo. But smartphones are just the tip of the iceberg. At IDF 2010 Beijing, which ended on April 14, the main topic was… cars. Doug Davis, Vice President of Intel, showed a prototype System-on-Chip (SoC) with an Atom core (codename Tunnel Greek), where the processor core itself, a memory controller, a graphics adapter, an audio codec, a decoder for HD-video processing and the PCI-Express interface, through which it is planned to connect all the peripherals. The code name for the platform, which includes the SoC itself, the hub (IOH) and the board on which everything is located, is Queens Bay. A slide showcasing the differences between the new Queens Bay platform designed to replace the failed 2008 Menlow It is promised that the graphics core will be 50% faster than the current one, and in terms of «performance per watt» the increase will be about 15%. Cheerful Chinese from HawTai Automotive hastened to announce that they will install a multimedia system in their B11 luxury sedan, where the Atom-based SoC will work under the control of MeeGo. True, there is no data on the SoC production technology and TDP of the new processor core. Most likely, it will remain 45-nanometer. Graphs from which you can draw some conclusions about the benefits of the Tunnel Creek core Back in Shanghai, they showed a slide that demonstrates the truly Napoleonic plans of Intel. Judge for yourself: the corporation is sure that Atom will come in handy absolutely everywhere, even in a cash register, even in a smart helmet of a soldier of the American army. Netbooks are, of course, great, but can you imagine how many cash registers are required around the world? What about slot machines? What about credit card and barcode readers? And, damn it, regular TVs? And almost everywhere there are now used ARM processors. But by the end of June, Moorestown should be released, and in the fourth quarter Tunnel Creek will go to the masses, and then, then … Atom: further — everywhere! Cold shower And now it’s time to ask — what then? And what is the likelihood that Intel will succeed? Let’s start with the most incredible application of Atom processors, smartphones. On June 1 last year, Qualcomm announced its new flagship, the 45nm Snapdragon QSD8650A, clocked at 1.3GHz. Devices on QSD8650A should appear on sale at the end of 2010, but developers of smartphones and software are already working with samples. Comparing the performance of such different architectures as x86 and ARM is not a very rewarding task (plus, do not forget that few people have seen the new Snapdragon live), however, according to data received from one Israeli software developer for smartphones and communicators, QSD8650A is about 10- 15% faster than Atom Moorestown. Yes a little. But there is a caveat: the power consumption of Snapdragon in standby mode is 10 milliwatts, and under maximum load it grows, scary to say, up to 0.5 watts. At the same time, the chipset immediately contains support for GPS, Bluetooth 2.1, DVB-H and perfectly decodes video with a resolution of up to 720p inclusive. Atom Menlow based MID prototype shown at CES 2008. Never made it to production… According to Intel, today the most economical Atom consumes 30 milliwatts in standby mode, and up to 1.4 watts under load (we are talking about the Z515 with a frequency of 1.2 GHz). And this is ONLY the processor core, while a variety of peripherals will also require additional power. For a stationary system and even a netbook, the numbers are very good, but for a phone they are simply deadly. Let’s remember the «four hours of talk time» promised by the LG GW9 developers90 — by all indications, it’s just about using the device with the screen on, that is, surfing the Internet, reading a book or watching a video. Would you buy a phone that runs out of power after four hours? To be honest, I wouldn’t risk it. The most mobile Atom Moorestown compared to Menlow. Unlike Queens Bay, it has its own southbridge Yes, most likely in Atom Tunnel Creek, or rather, its thoroughly stifled version of Lincroft specially for Moorestown, something will be thoroughly tweaked. But, taking into account the fact that the core will essentially remain the same, it is too optimistic to expect even a twofold reduction in power consumption from it. Meanwhile, Qualcomm plans to release a 1.5GHz dual-core 8X72 chipset before the end of the year. It must be understood that even this ARM monster will eat less than the younger Atom. How to compete with this? Of course, the alliance with Nokia significantly enhances Intel’s potential in the smartphone market, because the Finnish experience will come in handy, and devices based on Atom, released under the famous brand, will certainly encourage smaller manufacturers to experiment. Yes, and Android already supports «atomic» smartphones. And yet far from «telephone» power consumption is unlikely to make a device oriented to the mass market, and not only to the audience that buys Nokia N900 today. Well, it can’t, the mass model can’t work on a single charge less than the duration of daylight hours in the European part of Russia. While Atom only appears in smartphone prototypes, its ARM-architecture competitors have long settled there and seem to suit many. Pictured is the Google Nexus One board. As for the fight between Intel Atom and ARM in the markets of less mobile devices, everything is not very simple here either. Imagine that a twenty-seven-year-old Harvard graduate has arrived in Samara. Tall, handsome, already with his own business in Silicon Valley. He goes to the local beach, dressed only in a gold Tag Heuer and Lui Vuitton swimming trunks for a thousand dollars, and begins to intelligently get acquainted with the girls. Yes, he will probably find a pair for playing badminton pretty quickly. But it’s also unlikely that a queue for a game session with a beautiful young man will line up, because many girls did not come to the beach alone, or they have long had a serious relationship with someone, or they simply don’t want to get to know anyone, because and so good. And in general, in Samara, a slightly different system of relations between people than in Harvard, as the stranger can soon see, coming too close to the «busy» girl. What I mean is that the markets that Intel plans to enter with Atom have long been established. Many companies work there, they have their own ideas about how things should be, plus today more than two hundred companies have a license to manufacture processors with ARM architecture, and they, to put it mildly, are not happy about the arrival of such a powerful player as Intel. Of course, everything flows and changes, and if a large corporation decides to enter a new market fully armed, this should be taken very seriously. Suffice it to recall Microsoft with its game console: how everyone giggled over the first version, how everyone praised the Sony Playstation. And now the former leader occupies only the third place in the world, in the second place ridiculed by Microsoft, and in the first place by Nintendo, which at the time of the appearance of the Xbox was generally written off. At the same time, Microsoft, despite all its resources, was unable to retain the market share of Windows Mobile, because the platform, having defeated Palm, could not find anything to answer the newcomers who had come running. Most likely, Intel will have to face very fierce resistance, both fighting and polishing their main weapon. And this fight will drag on for years — with an unpredictable result so far. And in the near future, the main advantage for end users will be the emergence of a truly cost-effective Atom platform for netbooks — with support for HD-video and more or less decent 3D. At least the long-awaited models on dual-core mobile Atoms will be announced before the end of the summer. Waiting for a quad core for nettops? P.S. When the article was already written, the official announcement of the Moorestown platform took place. According to 3Dnews, its characteristics turned out to be even better than expected. So, in standby mode, Moorestown consumes only 100 microwatts and the device on the new platform can work up to ten days on a 1500 mAh battery, up to two days in music playback mode and up to five hours while watching video. Plus Moorestown is 1.5-4 (depending on the task) times faster than its predecessor. Well, at the end of the year the first smartphones will appear, and in 2011 the great and terrible Nokia will get into the fray. I already held a prototype in my hands — it is similar in size to the iPhone, only the screen is still worse. But it feels like everyone will have fun 🙂 Yours sincerely, Sergey Vilyanov Forum discussion New Intel Atom platform for ultra-mobile devices and smartphones. You’ve run a 3DMark graphics demo at exemplary smooth frame rates. Now imagine you are watching 720p video at 8000 kbps at a steady 30 fps frame rate. By the way, the webcam in the video conferencing program gives the same constant 30 fps . Now imagine all these applications on one screen. Of course, these puzzles are not difficult for an advanced computer, although the overload will be worthy. And now imagine that these applications operate with the same level of performance on the phone in your pocket . Your first thought is «Impossible.» Today there is simply no phone in the world that would have the ability to play video with such performance, not remembering 2 other additional tasks in the absence of loss of speed. Intel has publicly announced the release of the Atom Z600 processor line. Perhaps the performance of netbooks did not delight you in any way. Perhaps the previous generation ultra-mobile platform Intel (including phones and mobile Internet devices, MID) left a mixed memory. «How? Has Intel released a processor for telephones?” I hasten to assure you that the Z600 processor is a completely new generation of processors. The company invited the editors of Tom’s Hardware to the Austin, Texas, US Ultra Mobile Research Center for a preview of the platform formerly called «Moorestown.» And the reason for the conversation was very solid. Intel left no doubt about the seriousness of its approach to this market segment, and representatives of the company were ready to explain in detail why Lincroft : known today as the A tom Processor Z6xx processor, is a 45nm «system on a chip» SoC that integrates a compute core, memory and display controllers, a graphics engine, and a video engine. Why 45nm in an era, , when the Core line is already moving to the 32nm process technology, and when the die can be made even smaller? Simply because the Moorestow n was already in development when the 32nm process came into use. In other words, , the time to market with the 45nm process was less than , and the company was still able to achieve its goals. Langwell: chipset Intel Platform Controller Hub MP20 , which is manufactured in 65nm. Today, when most of the functions of yesterday’s chipsets have been transferred to the CPU, the PCH remains, , mainly for I / O operations, although we are in for a few surprises. Graphics and video Conversations with experts and additional information from various sources led to some comparisons that she herself Intel would not like to be held during the Moorestown press briefing. The Palm Pre and Apple iPhone 3GS phones use the Texas Instruments OMAP3430 processor based on the ARM Cortex A8 core and PowerVR SGX GPU. Both devices are known to provide great graphical features and are renowned for their beautiful 3D interfaces, right? The Snapdragon processor found in phones such as the Google Nexus One, HTC Incredible and HTC EVO 4G delivers half the peak fill rate of TI’s processor. Believe it or not, the Menlow outperforms the OMAP3430 by 60 percent—you might just never have heard of it because 9The 0560 battery life (or lack thereof) of the overshadowed everything else. But the point here is that Moorestown, with its 400MHz GPU, is doubling Menlow’s fill rate, and Medfield, Moorestown’s successor, will double that rate yet again. To cope with this bandwidth, Intel had to upgrade the bus between the CPU and graphics core, increasing the speed to 6.4 GB / s for reading and 4.3 GB / s for writing — incredibly high values, according to Intel. Intel has supported many standards in the Thalia graphics core, including OpenVG 1.1, OpenGL 2.1, OpenGL ES 2.0 and DirectX 9.L The emphasis on OpenGL no doubt explains why Intel loves to show Moorestown running a demo of Doom 3 (around 100 fps), but it also speeds up vector graphics and related applications like the Adobe Flash Player. It also plays a key role in why the Moblin/MeeGo 3D Clutter user interface is so attractive. Menlow and Moorestown have very similar graphics support (including support for SSE3 instructions), and the same can be said for video. Lincroft can simultaneously decode SD and 1080p30 HD streams. If you’re willing to put up with MPEG-2 quality 480p streams, Lincroft can output up to six of these streams simultaneously. These powerful capabilities are made possible by Intel’s built-in acceleration features. On the video side, Lincroft offers hardware-accelerated MPEG-2 and H.264/MPEG-4 encoding and decoding. Add to that WMV and VC-1 hardware decoding, as well as MPEG-1, Xvid, Real Video and Adobe Flash Video software decoding. If you do a lot of photo work, you’ll appreciate Lincroft’s addition of hardware-accelerated JPEG encoding. Video enthusiasts are interested in the possible streams, so they’ll be pleased with Moorestown’s support for 20 Mbps streaming with any profile, from baseline 720p at 30 fps to HP’s 1080p at 30 fps. No other phone platform available today can decode 1080p video. Only a few manage to decode 720p MP at 10Mbps. Intel mentioned that the Moorestown-based Aava Mobile reference design has enough decoding bandwidth to deliver 40 Mbps, though you’re unlikely to ever play that kind of content. Most likely, the ability to simultaneously decode two streams of 20 Mbps may be useful to you. Intel may be enjoying the performance crown in video processing in this segment today, but the competition will also arrive fairly quickly. In early 2009, TI announced their OMAP 4 line based on ARM Cortex-A9. The A9 is said to offer a 7x performance boost, allowing you to record 1080p video, capture photos at 20 megapixels (if you can afford the right sensors), and have enough battery life to play MP3 files for one week. It would be strange for Intel to target Medfield before Moorestown, but A9explains the reasons for this step. Display and memory In the world of laptops, most displays use the LVDS (low-voltage differential signaling) connection standard for their operation. Lincroft supports LVDS up to 1366×768 resolutions — quite a few pixels for sub-notebooks where you can expect the Atom. But in the world of handhelds and mobile internet devices (MIDs), displays are moving to the Display Serial Interface (DSI) standard promoted by the Mobile Industry Processor Interface (MIPI) alliance. MIPI’s goal is to set industry standards for wired and wireless interfaces in our multipolar world, where desktop or server areas often have very different priorities. DSI uses a serial bus like LVDS, but tries to achieve the least expensive implementation for LCDs, especially for the ultra-mobile market. Intel currently indicates that Lincroft can provide MIPI-DSI output up to 1024×600. So even if one were to imagine a Moorestown-based device with a display large enough to output native 1080 video, Lincroft would only be able to deliver a maximum of 720p to the display via LVDS, and even less via MIPI-DSI. Why were we so excited about Moorestown’s ability to deliver 1080p video that other phones can’t? This is partly because 1080p decoding performance means being able to decode multiple 720p or lower resolution streams. Plus, if you have a lot of 1080p movies in your video collection, you won’t have to waste time transcoding your content for transfer to ultra-mobile devices. Just copy the movie to take it with you and the device will take care of the rest. Lincroft’s integrated memory controller supports two memory formats: LPDDR1 up to 400 MT/s and DDR2 up to 800 MT/s. Why do you need support for two formats, given that the memory is still soldered onto the motherboard? The answer lies in positioning for different market segments. Moorestown is targeting two groups of devices, one for communications and connectivity and one for entertainment and productivity. Communication-focused models will feature 400 MT/s LPDDR memory and are unlikely to run processors up to 1.9GHz in BPT mode. These processors are intended for a group of devices for entertainment / productivity, where 800 MT / s DDR2 memory will be used. Interestingly, only the last group will support 1080p decoding. At the time of writing, we still have no confirmation that the communication platform will decode 720p video, although everything points to it. A typical standby power consumption of a pair of Lincroft and Langwell is about 3 mW. This applies to both communication and entertainment/productivity groups. However, with active use, the higher clock speeds of the last group will begin to contribute. If we get a range from 300 to 500 mW for a communication platform, then in the case of an entertainment / productivity platform — from 450 to 650 mW. At least, this is true for the preliminary samples of the equipment. Langwell Platform Controller Hub MP20 As we mentioned above, the MP20 chipset can be called, for the most part, an empty and reduced version of the former Poulsbo chipset, since it now only has a set of I / O blocks. But in fact, not everything is so simple. Yes, Intel fit a three-port USB 2.0 EHCI controller into a 14x14mm package. There’s also a USB On-the-Go (OTG) host/device port for connecting Moorestown ultra-mobile devices to other USB devices. Adding HDMI support from Intel isn’t heavily emphasized in Moorestown’s documentation, but it will be a key factor for those users who intend to use the phone as a video player for their TV. What about drives? There are four ports for SD/MMC/SDIO devices. But more importantly, Langwell supports SLC and MLC NAND SSDs with page sizes up to 4KB and net capacities up to 64GB. While Langwell supports the CE-ATA specifications used in 1-inch Microdrive drives, there is no SATA support. Probably, this function was recognized as redundant for the intended use of the platform. However, Intel has provided instructions for developers who wish to implement an optional USB-to-SATA bridge. There are also instructions for connecting a full keyboard through I2C interface . Camera support and image processing are also included in the MP20 functions, and this applies to individual pictures and videos. Moorestown allows you to connect two cameras — one with a resolution of 5 megapixels, and the second with a resolution of VGA (640 × 480). The 5MP channel is a dual-lane MIPI interface capable of supporting RGB, YUV, and RAW color formats, although the practical implementation of this capability is up to the developers. Langwell’s sound engine is more complex than you might expect. Intel used its own Smart Sound Technology (SST), based on a 24-bit DSP, which is responsible for voice processing and audio codec acceleration. In particular, Langwell provides hardware-accelerated AAC-LC and PCM (WAV) encoding. In terms of decoding, AAC-LC, HE-AAC, MP3, PCM and WMA9 hardware decoding is supported. This optimization, along with other Langwell power-saving elements, is why Intel was able to claim to have achieved its goal of MP3 playback for a week. When the sound engine needs system resources, it creates an event in the power management system, but only the memory access interface is turned on. The sound engine updates its buffer, after which the platform quickly returns to a low power state. Might be of more interest to business users than general consumers, but Langwell integrates some security features and a crypto engine, including Trusted Platform Module (TPM) chip functionality. Intel speeds up AES, DES, 3DES, RSA, and other encryption/decryption operations, but the Secure Boot Environment is a curious benefit, as is the case with vPro systems. It seems to us that it will become more and more relevant as mobile devices become more and more victims of hacking and malware. Of course, content providers will be able to rely on DRM support features in the security engine. Briertown Mixed Signal IC Briertown may seem like the fifth wheel in the Moorestown platform, but this chip is critical to achieving Intel’s power goals. So far, there are no official Briertown designs at all. Briertown exists today as an architecture specification for supporting peripherals and providing power, including battery charging, which is now done in hardware rather than software. Look at the Briertown chip on the Moorestown motherboard and you’ll see that it’s significantly larger than the CPU/chipset combination. Despite this, Briertown requires about half as many components and takes up about a third of the area on the motherboard compared to a similar power supply scheme on Menlow platforms. Not surprisingly, the Briertown also costs about a third of the price of its predecessor.