Mobile Semiconductors: In Search of the One Chip Solution

Print Friendly, PDF & Email

SEE LAST PAGE OF THIS REPORT Paul Sagawa / Artur Pylak


sagawa@ /

twitter.jpg @SecSovTMT

March 13, 2013

Mobile Semiconductors: In Search of the One Chip Solution

  • Portable devices will eventually push PC architecture into its dotage. This paradigm shift is creating huge opportunities for semiconductor companies, with processors, modems, radios and sensors the biggest categories, and with footprint and power efficiency paramount design considerations, as device OEMs cope with increasingly demanding apps, evolving wireless standards, and a growing roster of spectrum bands. Within this, the processor drives much of the measurable performance of the device, and device makers have largely standardized on ARM’s widely licensed core architecture, leapfrogging each other generation to generation. Chip makers have 3 main competitive levers: integrating multiple elements onto a single “system on a chip” SoC; delivering superior design for elements of the solution; and exploiting proprietary manufacturing process advantages. Qualcomm was first to integrate an LTE modem to its SoC, which has added performance advantages from its proprietary ARM-compatible processor vs. rivals, including Apple and Samsung, using ARM’s reference designs. Intel’s new 14nm process is inherently faster and more power friendly than the 28nm used by Asian foundries, but its CISC architecture wastes much of that advantage and is incompatible with the favored ARM standard. Qualcomm is also playing for the lead in radio chips, recently announcing a 40 band RF solution that allows OEMs to support most carriers with a single device.
  • Portable device architecture is becoming dominant. Smartphones, tablets and keyboard equipped devices applying the same network reliant philosophy, are on a trajectory to squeeze out PCs and other older era platforms, first with consumers and, eventually, in the workplace. This paradigm shift, which includes the integration of portable device architecture into products like vehicles, appliances and TVs, is creating huge opportunities for semiconductor makers.
  • Processors, modems and radios. The main categories of portable semiconductors are central processors, graphics processors, baseband modems, radio transceivers (RF) and sensors. Most OEMs have built their software for the widely licensed ARM processor standard, establishing a serious hurdle for vendors, like Intel, with alternative processor architectures. Modems must cope with rapidly evolving wireless standards, and the variants that pop up, rewarding serious wireless chops. RF is growing more complicated, with a global expansion in the bands dedicated to wireless forcing radios to be able to tune to a much wider range of frequencies. Sensors – e.g. cameras, touch screen controllers, etc. – are a fast growing category with new functions – e.g. gesture recognition, environment and vital sign monitors, etc. – likely to emerge.
  • Low power, small, powerful and cheap Portable architecture raises the importance of power efficiency and footprint for chip makers looking to exploit the paradigm shift, while the performance and price of the solutions remain obvious selling points. The fast product cycles of the portable device market also reward vendors that are able to bring new solutions to market quickly, maximizing the time that an innovator can exploit advantages as rivals move to leapfrog them with their own new solutions. There are three main levers that semiconductor players use to gain these advantages: integration, design innovation, and process innovation.
  • Integrating to a System on a Chip (SoC) Replacing multiple chips with one saves a lot of space and power in a portable device. Current leading-edge SoCs feature multiple cores. For example, Samsung recently announced its Exynos Octa, combining four high speed cores with four low power cores, separating tasks with different needs to maximize performance without taxing battery life. Most high end SoCs also integrate graphics processors. Qualcomm was first to integrate an LTE modem to its SoC, helping it gain a dominant position in smartphones. To date, only Nvidia has announced similar capability. The ultimate SoC would also integrate analog RF with the digital processors and modem, a difficult challenge requiring a tricky mixed-signal process and a very complicated design. In the interim, chip makers can combine SoC and RF chips into a single package, an improvement over the entirely separate configuration.
  • Design leadership. Mobile processors, modems and RF are complicated, giving chip makers opportunities to differentiate based on innovations in the way the circuits are laid out. For example, most SoC vendors, including Samsung, Nvidia, and Apple, license ARM’s Cortex reference designs, integrating the standard cores with proprietary logic. Qualcomm takes it one step further, using its own compatible Krait processor design for its cores, a choice that gives it slight footprint, power draw and performance advantages. Qualcomm also recently announced an RF solution that supports 40 different frequency bands, allowing device makers to support most of the world’s wireless carriers with a single radio configuration, a considerable benefit.
  • Process innovation. Reducing the size of transistors on a chip improves its speed and reduces its need for power. Improvements to fabrication processes enable these smaller geometries on a typically predictable development trajectory. Today, most high performance chips are manufactured 28nm processes that will hit a physical limit with next move to 22nm. Intel has solved this problem with a 3D process that will allow it to build parts at 14nm 12 to 18 months ahead of its rivals. This may improve Intel’s SoC performance by more than 50%, an advantage partly mitigated by the inefficiency of its complex instruction set (CISC) design. We note that most device OEMs have considerable investment in ARM architecture, and Intel’s process advantage does not appear long lived enough to break ARM’s hold on the mobile market.
  • Qualcomm is the likely big winner in SoC. Qualcomm has design leadership in processors, modems and, as of its recent announcement, RF. This leadership, combined with its head start in SoC integration and lucrative IPR licensing business, gives it real advantage, even as device makers like Apple and Samsung push to vertically integrate. Nvidia has been a step behind, but has strength in graphics processing that has played well in the tablet market. Qualcomm’s 40 band radio has the potential to leapfrog the 4 traditional RF players – RFMD, Triquint, Skyworks, and Avago. The sensor market is still undefined as new applications emerge, but OmniVision has established itself as a leader in camera sensors. We are not optimistic that Intel, Texas Instrument, Broadcom, and others are positioned to challenge for leadership.

Portable Device Semiconductors – A Big opportunity for Small Solutions

Sometime after the 2010 launch of the iPad, it became clear that portable device architecture would displace the venerable x86 Wintel PC as the paradigmatic hardware platform for a new computing era. Defined by the ARM processor, integrated wireless communications, flash storage, and a growing reliance on cloud-based data centers, portable architecture is proliferating. The rise of smartphones, tablets, ultrabooks, netbooks, thin desktops, next-gen set-top-boxes, connected TVs, game consoles, automobile infotainment, wearable computers, and other yet-to-be conceived product categories, all committed to this new architecture is the opportunity of a generation for semiconductor vendors. Leaders have come to the front for the biggest parts of this gold rush – processors, modems, radio frequency (RF) transceivers, sensors and flash memory – already a $48B annual market for chip makers, growing at a healthy 16.2% annual clip.

At the center is ARM, which does not make chips, but rather licenses its power efficient RISC processor core architecture to many others. This wide support has been a key factor to establishing ARM as standard, allowing device makers and application developers to tie their software tightly to the instruction set without locking in to a single silicon supplier. Mobile chip makers, almost all licensing ARM’s technology, compete by implementing it within proprietary system designs – integrating more functionality to the central System-on-a-Chip (SoC), finding clever ways to eke out additional performance within the standard, and by pushing to new manufacturing processes that shrink the size of chips and improve their performance.

For example, while most licensees use ARM’s reference design, Qualcomm has designed its own ARM-compliant core, delivering better performance in less space with less power. It is also the first to have integrated an LTE modem onto its SoC, a huge space and power advantage, and recently announced a radio solution that supports 40 different bands on a single chip – previously, OEMs had to insert different RF chips for different carriers. The next step will be integrating the single chip RF solution into a common package with the processor and modem, saving further space and power, with the ultimate goal of integrating the analog radio with the digital SoC still a significant design and manufacturing challenge. These advantages have taken QCOM to the top of the heap, where even rivals like Samsung and Apple use its solutions for at least some of their products.

The other major players are more specialized. ARM makes a royalty on every core. Apple designs its own processors, but buys the rest of its solution. Samsung makes processors, sensors and flash, but has struggled with modems and RF. Huawei has started using its own processors, but that’s all. Nvidia, who’s graphics prowess has made it a strong player in tablets, recently announced its own SoC with an integrated LTE modem, but it has no expertise in RF or sensors. Intel is pushing a non-ARM SoC, relying on its process advantage to compete on specs – we are not optimistic. Qualcomm is dominant in standalone 4G modems, with 3G modem competitors, like Broadcom and ST-Ericsson, playing for scraps thus far. RF has been a free-for-all, with Avago and Skyworks having recently had the upper hand vs. RFMD and TriQuint, but Qualcomm’s new 40 band RF is a threat to all of them. The sensor market is wide open – specialist OmniVision is squaring off with Samsung and Sony in image sensors, but most sensor categories are too new to be well defined. Finally, flash is a well-established commodity with fast growing volumes and painful periodic price wars.

Portable Architecture – It’s Not Just for Smartphones Anymore

First there were cell phones, pagers, PDAs and iPods, which begat the Smartphone. While the term was bandied about in the pre-iPhone era, Apple’s bold stroke in 2007 redefined the concept and began the paradigm shift, inspiring Google’s Android, Microsoft’s Windows Phone, and a host of wannabes to follow. Smartphones now account for more than 40% of total mobile phones sold, with a total global volume of more than 700m units in 2012, up more than 45% YoY (Exhibit 1).

Exh 1: Global Smartphone and Tablet Production, 2012-2016

Apple struck again in 2010 with the iPad, a 10 inch riff on the iPhone concept, a rampant success that shocked industry greybeards with memories of Microsoft’s humiliating tablet flow a decade earlier. Once again, Google and Microsoft followed Apple’s lead, – their OEM licensees have delivered a cavalcade of tablets in multiple form factors and price points. 2012 global tablet sales were more than 140m units, roughly double 2011 volumes.

Looking further ahead, ultra-books and netbooks – basically tablets with a good keyboard – could be on the verge of being a “thing”. Enterprises, already serving some users via “virtual machines” running on a network resident data center, could use the thin portable architecture for inexpensive desktop terminals. Television boxes and gaming consoles are veering toward the portable approach. Automobile makers may be finally ready to embrace the new century with modern, upgradeable infotainment systems. Google Glass, Apple’s iWatch, and other wearable form factors built on portable architecture may soon augment our reality. Buzz about the Internet of things, reminds us that many things that are hard to imagine as connected may one day be connected. Meanwhile, the markets for feature phones and consumer PCs are withering in the face of the portable device onslaught, shaking out the semiconductor makers that supply the building blocks for both the new architecture and the old ones that it is making obsolete.

From a hardware perspective, four things stand out as defining the new portable architecture. First, ARM Holding’s RISC processor technology is standard. Second, wireless communications, often across multiple types of networks, are essential. Third, storage is either solid-state or in the cloud – no disc drives allowed! Finally, sensors, e.g. touch, image, gesture, movement, etc., are more important forms of input than a mechanical keypad. These definitional characteristics also point to the key semiconductor product opportunities created by the paradigm shift – processor centered System-on-a-Chip (SoC) solutions, digital baseband modems to manage wireless communications, radio frequency (RF) transceivers, flash memory storage, and sensors. Collectively, these categories make up 56% of the bill of materials for a typical smartphone, and generate $48B in annual chip sales across all mobile devices (Exhibit 24).

Exh 2: Device Teardown Comparison

Exh 3: Component Composition of a Premium Smartphone

Exh 4: Portable Device Semiconductor Content, 2012-16

ARM Inside

Apple’s first foray into portable devices, the ill-fated 1992 Newton PDA, was ahead of its time for more than the obvious reasons. At the core of the Newton was a reduced instruction set (RISC) processor designed by a subsidiary of the struggling British computer maker Acorn, in which Apple, and its partner Olivetti, had taken a 47% equity stake. Unlike PCs, which used Intel’s iconic x86 complex instruction set (CISC) processor, the Newton would have to run on batteries, placing a substantial premium on power efficiency. The Acorn RISC Machines approach streamlined the set of computations supported by the CPU, allowing a simpler, smaller design that saved considerable space and power draw. The short list of computational functions ran with extraordinary speed and power efficiency, while more complex instructions would be broken down into combinations of simpler instructions to be completed over multiple CPU cycles. In contrast, the x86 and its CISC architecture could run a laundry list of functions in a single cycle, making it faster and more efficient for applications that called for a lot of complex functions, but much slower and power hungry for applications that did not. Since Apple projected that the primary applications on the Newton – e.g. calendar, notepad, contact lists, etc. – would not require heavy computation, the choice was obvious.

The Newton was put out of its misery by 1998, but Acorn RISC Machines, by then renamed Advanced RISC Machines or ARM, went public in that year, netting Apple $800M for its share. ARM pursued an unusual strategy, licensing its RISC processor designs for other chip makers to incorporate into their product lines rather than selling them directly to OEMs as chips. While the licensing only strategy conceivably left money on the table, it was a key factor in the wide adoption of the ARM architecture as a common standard for processors used in high-end cell phones. These proto-smartphones, led by Nokia from one direction and Palm Computing from the other, began combining the functionalities of cell phones, pagers and PDAs. By the time that Apple introduced the iPhone in 2007, beginning the modern Smartphone era, the ARM processor architecture was legion in the portable device world.

Today, iOS, Windows Phone 8, BlackBerry10, ChromeOS, and even new hopefuls like FirefoxOS, Tizen and Ubuntu, are all written specifically to the ARM processor architecture. Even Android, which Google pledged to support on both ARM and Intel’s CISC architectures, greatly favors ARM given that its OEM licensees have invested to customize the OS to their own needs, with that customization and many leading apps tied to an underlying ARM processor design. It would be costly and time consuming to make a switch, and with so many chip makers supporting the ARM standard, ARM architecture is, more or less, a defining characteristic of the portable device era (Exhibit 4).

Exh 5: Mobile Processor Power / Performance Improvement Over Time

Today, iOS, Windows Phone 8, BlackBerry10, ChromeOS, and even new hopefuls like FirefoxOS, Tizen and Ubuntu, are all written specifically to the ARM processor architecture. Even Android, which Google pledged to support on both ARM and Intel’s CISC architectures, greatly favors ARM given that its OEM licensees have invested to customize the OS to their own needs, with that customization and many leading apps tied to an underlying ARM processor design. It would be costly and time consuming to make a switch, and with so many chip makers supporting the ARM standard, ARM architecture is, more or less, a defining characteristic of the portable device era (Exhibit 4).

ARM generated £214B ($339.3B) in licensing fees for its core technology in 2012, up 19% YoY, driven in large part by a trend toward integrating multiple cores into a single processor solutions. In 2012, dual core solutions for both smartphones and tablets were commonplace, with quad core implementations hitting the market in the 4th quarter. Recently, Samsung announced its 8 core Exynos Octa, which uses ARM’s BIGlittle technology to separate tasks requiring high performance from those where power efficiency is paramount. In total, mobile processor chips, including SoCs including graphics processors, digital modems and other integrated system functions, are a $25.6B market, growing at 9% (Exhibit 6).

Exh 6: Global Mobile Phone SLI/SoC Revenue, 2010-2015

Talk to the Hand

While cellular connectivity is the primary directive of a smartphone, Apple’s first iPhone was designed to be tethered to a Mac, which was used to synch applications, updates, and downloads. However, that minor indiscretion was soon rectified, with modern portable devices designed to make maximum use of the airwaves. All smartphones and many tablets contain digital cellular modems, typically supporting all of the various 2G and 3G standards and sometimes supporting 4G LTE as well. WiFi, in its various flavors, is almost always supported by portable devices of all ilk. Bluetooth, for connecting peripherals over short distances, is also nearly ubiquitous. GPS, which must communicate with satellites in orbit to fix location coordinates, is increasingly important. Near Field Communications (NFC), a highly secure, high speed, short distance standard is showing in some new smartphones and tablets, generally in support of using the device as a means of payment in stores equipped to handle it. The preference for wireless operations is strong enough that cordless magnetic induction battery charging technology is quickly becoming the next “must have” in this season’s newest smartphones.

Exh 7: Major Semiconductor Categories present in a Typical Smart Portable Device, 2013

The job of coding and decoding all of the digital information traversing these wireless connections is performed by baseband modems, specialized processors designed to perform routinized operations to a digital signal in real time. Of the various wireless services incorporated into a portable device, digital cellular is by far the most complex, with every subsequent generation – 2G to 3G to 4G – multiplying the demands on the baseband modem many fold. Traditionally, the cellular baseband has been implemented separately from the baseband processor handling all of the other required digital communications, but recent solutions from Qualcomm have not only integrated WiFi, GPS, Bluetooth and NFC into its 3G and 4G baseband, but have integrated the entire multiservice baseband into the processor SoC. This is a huge potential advantage in cost, footprint and power usage relative to a multichip implementation (Exhibit 7).

Of course, wireless standards evolve – the 3GPP industry organization, which governs both 3G and 4G standards, recently ratified its 10th major release, detailing a substantial update to the 4G spec, colloquially known as LTE Advanced. Modem makers have already incorporated this version into their top end baseband processors, well before commercial networks based on the technology have been deployed, and have begun working on implementing the draft standard of Release 11 in anticipation of its ratification later this year, with Release 12 just over a year after that (Exhibit 8). The frantic pace of technical change makes it very hard for new vendors to move down the learning curve fast enough to compete for leading edge business, helping Qualcomm sustain its dominance in 4G. Nonetheless, rival modem makers have made incursions into the older 2G and 3G baseband standards that still prevail in most emerging markets.

Exh 8: Wireless Data Standards Releases by Technology

What is the Frequency, Kenneth?

Radio communications requires more than a baseband modem. Radio Frequency (RF) transceiver chips are needed to interpret the signals captured by antennae, converting them into a digital stream of data that can be processed by the baseband modem. The RF transceiver also takes outbound digital streams from the baseband, converts them into an analog representation and steps them up to the appropriate frequency for transmission over the antenna system. The analog nature of radio communications requires a different approach to semiconductor design than is used for functions of an inherently digital nature, such as baseband or central processors. Moreover, silicon, the most commonly used substrate for semiconductors, has natural frequency limitations that render it inappropriate for chips that need to process signals out of its range, pushing RF makers to more expensive and arcane materials like Gallium Arsenide or Silicon Germanium. Because of this, integrating this functionality into a digital SoC would be a formidable challenge (Exhibit 9).

Good RF design is considered as much an art as a science, managing multiple signals over multiple frequency bands amidst interference is tricky, and the job has gotten geometrically more complicated with the expansion of the frequencies used for portable device communications. A 2G digital cell phone really had to account for four major frequency bands for near global coverage. 3G added another 7 or 8. However, as governments scrounge to find scraps of spectrum that can be reallocated for 4G wireless, the requirement for global coverage looks like 40 bands ranging from 450MHz at the bottom to more than 3GHz at the top. Thus far, device makers have employed separate RF configurations for various carriers in different parts of the world using different frequency bands, but this approach is inefficient and expensive, breaking up production volumes and complicating the supply chain (Exhibit 10).

Exh 9: Analog RF Flow Simplified

Exh 10: Global Allocation and Utilization of Cellular Frequency Bands

Enter Qualcomm and its new 40-band RF transceiver. Announced earlier this year, this single chip solution is the first answer to the 4G device maker’s quandary. If it works as advertised, it positions the company to extend its dominance of the 4G device market from SoC processors and baseband modems into RF, definitely bad news for the traditional RF players Avago, Skyworks, TriQuint and RFMD. Moreover, even though we believe that an integrated RF SoC is very, very far from reality, it is more than possible for Qualcomm to combine its “universal’ RF chip with its integrated 4G SoC in a single sealed package. This half-way integration would reduce footprint, improve power efficiency and cut manufacturing costs for device makers, giving Qualcomm unique advantage over its various rivals, all of whom specialize in one part of the solution or another.

We note that there are several other chips in the radio end of a portable device – discrete parts like filters and simple switches, and the power amp. We chose not to go into detail, as these parts are fairly generic and not necessarily unique to portable devices.

Thanks for the Memories

Portable devices don’t have disc drives. This wasn’t always true – the first iPod was built around an innovative 1.8 inch disk drive developed by Toshiba and signed as an exclusive by Apple. Nokia began its own quixotic quest to deliver its own disk drive equipped smartphone around the same time, and delivered it to market in 2003, even as plummeting flash memory prices had Apple planning to jump ship to a solid state design. With the introduction of the Nano in 2005, Apple turned the corner with its first flash memory iPod, and left disk drives for good in 2007.

Along the way, Apple became the world’s largest buyer of flash memory, accounting for more than 28% of all volume in 2011. It has been noted that Apple’s highest margin product is the extra price charged on additional flash memory in its devices. This is a testament not only to the importance of flash memory to portable devices, but also how hard Apple has been able to push its flash suppliers for lower prices (Exhibit 11). Samsung may have more than 40% market share in the category, but leadership has not given it pricing power over what has become a commodity part. In 2012, the price of 4GB flash memory chips dropped by more than 50% YoY, with total industry revenues dropping 7% despite robust volume growth. The industry expects better stability in 2013, but it does not seem likely to be more than a respite from margin crushing cyclicality.

Exh 11: Apple Device Memory Component Cost Estimates ($ per GB)

Moreover, the trajectory of ever increasing flash storage configurations in mobile devices may have a logical plateau with the increasing importance of cloud-based functionality. The model of carrying a full archive of photos, song downloads, and family videos in your pocket is rapidly changing to one where these files are held in the cloud for access from any device with internet access. Demand for specialized solid state storage in the cloud may be strong, but the device driven volumes of commodity flash memory will likely decelerate.

Sensors and Sensibility

Qwerty had a long run. After pushing punch-cards to the sidelines in the ‘60’s, the keyboard happily bipped along through the PC era, even enjoying a renaissance of sorts several years ago, when T-Mobile’s Sidekick and then RIM’s Blackberry made Qwerty cool again and reminded power users of the limitations of predictive text input on 12 key numeric keypads.

Once again, along came Apple to spoil the party. The iPhone famously ditched the keypad entirely in favor of a single home button and an innovative “multi-touch” screen, dooming physical keyboards to the scrap heap of telescoping antennae and WiFi dongles, at least for devices smaller than a full-on Ultra-book. Instead, smartphones and tablets are incorporating an array of sensors to capture the world around you (Exhibit 12).

Exh 12: Summary of Mobile Innovations, 1996 – Present

Touchscreens require a sensor – typically a transparent capacitive matrix overlaying the display – and that non-semiconductor sensor requires a controller. These controllers are a $1.5B market growing at a mid-teens annual growth rate, led by Atmel, Cypress and Synaptics, but facing growing competitive threat from a variety of Asian market entrants. Cameras, now de rigeur for most portable devices, also require a sensor, this time a light sensing semiconductor. According to IC insights, image sensors are a $6.3B market projected to grow 13% on average over the next 5 years, with portable devices the biggest part of overall demand. OmniVision leads Samsung, Sony, ST and Toshiba in this market (Exhibit 13).

Demand for other types of sensors, such as motion, vibration, proximity, etc., is growing but nascent and the supplier base is highly fragmented. With the potential of wearable devices on the horizon, it seems likely that new opportunities for semiconductor makers will emerge.

Exh 13: Global CMOS Image Sensor Sales

How the Game is Played

Competition for the semiconductor opportunities being created by the paradigm shift to portable device architecture is playing out in three dimensions. The first is circuit design. In the processor market, most players have chosen to implement the Cortex processor core reference designs licensed by ARM, but Apple and Qualcomm have taken to design their own processor cores that adhere to the instruction set and operational specifications provided with each release of the ARM standard. This approach has given both companies a time to market advantage, typically completing their internal implementations before ARMs own engineers release their designs to licensees. Both also seem to have derived performance advantages as well, including smaller footprints vs. similar parts from rivals using Cortex. Qualcomm derives similar advantages in the baseband processor market, typically first to support new standard releases with superior performance and power efficiency. Qualcomm showed its design prowess in the RF market as well, with its innovative 40-band transceiver.

Integration is a second competitive lever. All things being equal, one chip is faster and more power efficient than two. Devices no longer sport a stand-alone CPU, rather the central processor is implemented on a single System-on-a-Chip, with at least system memory and graphics processors on board. Apple and Nvidia have focused on amping up the graphics capabilities of their SoCs, with Apple’s latest A6X chip packing 32 graphics cores capable of 76 GFLOPS and Nvidia’s announced Tegra4 claiming similar performance from its 72 cores (Exhibit 14). Graphics performance is generally a bigger consideration on the larger screen of a tablet than on a smartphone. In contrast, in 2012, Qualcomm delivered the first SoC with an integrated 4G LTE baseband modem, giving its solution a substantial advantage in smartphones, where cellular connectivity is critical. Qualcomm’s integrated S series SoCs dominated the merchant processor market last year, even Samsung, which prefers to use its own in-house Exynos processors wherever possible, picked Qualcomm for its LTE equipped smartphones. While Nvidia recently announced that it, too, had an integrated LTE SoC, we expect Qualcomm to sustain its dominance, particularly with LTE Advanced just around the corner. Finally, we don’t believe that SoC products with integrated RF are forthcoming due to the complexities of mixing analog and digital on the same die, Qualcomm’s 40-band transceiver opens the possibility of combining it with the Snapdragon SoC within a single contained package. A two-chip module would offer unique performance, footprint and power draw advantages that would be invaluable to smartphone makers.

Exh 14: Announced Flagship Mobile Processor Specs

The third main lever is manufacturing process. The fundamental driver of Moore’s Law is the ongoing reduction in the size of integrated circuits. With advances in manufacturing, the microscopic wires can be thinner and the gates placed closer together, reducing the time needed to complete a computational cycle and, at the same time, reducing the electricity needed to do it. So each move to ever lower line sizes, from 32nm to 28nm to 22nm and on down to 14nm, improves the performance, improves the power efficiency and lowers the cost. While this onward march typically iterates on well-established techniques for miniaturizing circuits, these techniques occasionally hit up against natural limits. Such a limit was reached with the move from 32nm to 28nm – going smaller would be impossible using the most widely used techniques, as interference across the gaps between lines would exceed acceptable limits. Intel solved the problem using a different technique that bridged over circuit lines in 3D, and has already moved to 22nm on its way to 14nm. While the big contract fabs in Asia – e.g. TSM, Samsung, Hynix, etc. – work to catch up, Intel has an advantage, which it is using to try to lever its CISC processor architecture into the portable device market. Thus far, the reception has been chilly – design advantages help the ARM-based competitors limit Intel’s performance lead, while the costs of embracing an entirely new architecture make the change unpalatable to many OEMs. Unless news of serious setbacks as the Asian fabs seek to close the gap emerges, we do not expect Intel to make headway.

Winners and Losers

With its unparalleled wireless expertise, its strength from RF to processor design, and its clear path to an integrated single-package smartphone solution, we see Qualcomm as clear leader in new architecture semiconductors. Its Krait ARM compatible core is state-of-the-art, its integrated baseband modem products are dominant, and its new 40-band RF solution is a potential game changer. We note that the two biggest smartphone suppliers, Samsung and Apple both produce their own processors, but that neither has the baseband capabilities in house to produce a competitive integrated solution to rival the Snapdragon line. Nvidia has developed an integrated baseband SoC, and time will tell how well it plays beyond Nvidia’s graphics-oriented tablet comfort zone (Exhibit 15).

Rivals in the baseband market, such as MediaTek, Intel, Broadcom, and ST-Ericsson, fare much better in the simpler and more stable 2G/3G modem segment, but don’t appear to be pressuring Qualcomm in the more lucrative LTE segment, where has a dominant market share. The RF market is more fragmented with top vendors Avago, SkyWorks, TriQuint, and RFMD, rising and falling with each design-in win and loss. Qualcomm’s new 40-band solution is a potential game changer here, as would be an integrated RF/SoC single package solution.

Exh 15: Winners and Losers

Insofar as there is any winner in the commoditized market for NAND flash for portable devices, it would be Samsung, which at least has the scale economies related to its 40% market share. Still, with device storage configurations likely to stagnate in the next generation or two, even leadership may not be particularly attractive.

Finally, sensors are a fairly nascent and fragmented market, with only touchscreen controllers and image sensors mature enough to make a call on winners and losers. Atmel, Cypress and Synaptics dominate touchscreen controllers, but could be vulnerable to a raft of Asian competitors targeting the opportunity. Image sensors are a more differentiated market, with OmniVision the leader and a pure play on the opportunity. The others, Samsung, Sony and Toshiba are small businesses within much larger organizations.

Print Friendly, PDF & Email