Significance: Four years after it decided to exit the handset baseband market, Texas Instruments looks set to move another step away from the mobile space with which it is so heavily associated, scaling back its OMAP app processor activities on smartphones. However, it will not, as often rumored, sell the platform, but will instead refocus it on embedded devices and industrial markets such as automotive. There is internal logic to this ‘ TI’s areas of growth and leadership are increasingly in industrial and analog chips rather than mobile devices, where its competitive landscape has change dramatically since the launch of the iPhone and the decline of its major customer Nokia. Its decision to quit basebands has left it weakened against makers of integrated SoCs like Qualcomm, and it also faces new challengers like Nvidia and Intel. But in shifting its efforts to the embedded space, TI is doing more than admitting defeat in smartphones. It is following a far broader industry movement, in terms of growth, value and power plays, towards the internet of things. As the mobile device segment consolidates around a few leaders, there are far bigger opportunities in the vast embedded and M2M areas ‘ not just for chipmakers like TI and Freescale, but for other players in the mobile food chain, notably the software platforms. Just as those forced into secondary positions in processors can seek new chances in embedded, so can the OS suppliers squeezed out by Apple and Google ‘ RIM, HP and most importantly, Java.
Texas Instruments may be preparing to exit the mobile processor market, where its OMAP platform is a venerable player. The US company fervently denied reports, earlier this year, that it might seek a buyer for OMAP, but now says the opportunity is ‘less attractive as we go forward’. If it follows through on the comments, it will highlight two important trends which are changing the balance of power in the mobile silicon space.
One is the upturning of the traditional order in the handset chip space, as a result of the move from voice-focused cellphones to mini-computers with full processing power. The need to pack in advanced processing, graphics and connectivity functionality while barely increasing power consumption has delivered the market to the masters of integration, notably Qualcomm, even as the list of high end customers has shrunk for everyone. Apple and Samsung, which account for about half the smartphone segment, are self-sufficient in processor terms, while conversely, the decline of Nokia has hit its major vendors, including TI itself and ST-Ericsson.
The second trend is the shift of growth and market power away from conventional mobile devices and towards the embedded sectors, and ultimately the ‘internet of things’. TI does not plan to abandon app processors but to refocus its efforts on the automotive and other embedded areas, a change of emphasis which echoes that of Freescale. ARM and Intel are increasingly battling in the ultra-low power platforms which these sectors demand, as well as in smartphones and servers, and the shift is not all about chips. Players in other parts of the mobile chain, especially the operating system, are also looking for growth opportunities in embedded devices as their prospects on smartphones are squeezed by Google and Apple.
In an interview with Reuters this week, Greg Delagi, Texas Instruments’ SVP for embedded processing, said the firm would invest less in the smartphone and tablet segments in future, and focus its processor efforts on markets such as automotive. As with its huge bet on the analog chip space, this shows TI putting all its resources into areas where it has natural leadership, rather than forcing itself to be a challenger. It is already the leading semiconductor maker in the industrial chip sector, according to IHS iSuppli. Importing its mobile experience to some of those bases makes a lot of sense, as the embedded systems in cars and homes become more sophisticated, following handsets along the road to incorporating computing power and browsers.
The industrial market is a fragmented one ‘ TI may have the lead, but its share was only 7.3% last year (up from 6.4% in 2010). Offering more options, and higher value systems, to key customers could help increase that further, and there is plenty of growth to be had ‘ TI’s revenue from the industrial business rose by 25% to $2.2bn in 2011. Delagi said growth in the embedded area would be slower than in mobile, but would “generate a more stable, profitable long term business”, targeting gross margins of 55% to 60% (in an uncertain timeframe).
This means that TI is likely to remain in the mobile market only in the two remaining areas where it can also claim significant market share ‘ wireless infrastructure chips, where its KeyStone architecture is extending its reach from the base station upwards to the network processor and downwards to the small cell; and smartphone connectivity, where it battles Broadcom, Marvell and others for combo chips integrating Wi-Fi, Bluetooth, GPS and other radios.
It is all a far cry from the pre-iPhone days when TI dominated the cellphone chip market, largely thanks to its symbiotic relationship with market leader Nokia. When Nokia started to decline, and also opened up its supply chain to a wider range of chipmakers, TI began to pull back slowly from the space as well as to build up its businesses elsewhere. In 2008, months after Nokia shifted a large part of its business away from custom partnerships with TI and towards several merchant providers, it announced it would quit the cellphone baseband business (apart from some of that remaining custom activity). Not long afterwards Freescale made a similar decision, having been sold off by Motorola, losing its near-exclusive at a vendor which, like Nokia, was losing ground anyway.
With that move, TI took a gamble that, as smartphones morphed into small computers, vendors would buy app processors separately from the modems, and that proved true for a short period. However, the recent trend has been to integrate the processor, GPU and baseband back into a single system-on-chip, as epitomized by Qualcomm, in order to reduce cost and power consumption. The first wave of a new device category, whether tablets or LTE smartphones, tend to have separate elements, but after that, as the form factor commoditizes, the race is on for full integration. Several processor specialists have acquired baseband firms ‘ Nvidia bought Icera and Intel acquired Infineon Wireless ‘ in order to challenge the broad portfolios of Qualcomm, Broadcom, Renesas Mobile and Marvell. Intel itself is likely to offer its first all-in-one LTE smartphone product next year, leveraging the Infineon basebands fully at last.
This left TI somewhat isolated, despite the strong established presence of OMAP, now entering its fifth generation and set to harness quad-core technology and the new ARM Cortex A-15 architecture. TI says it will continue to support key customers ‘ a recent flagship win was in the Amazon Kindle Fire ‘ but it seems the space has become too competitive to keep OMAP as a central business unit. Qualcomm is consolidating its lead; Apple and Samsung are increasingly relying on their own platforms, removing them from the addressable market (ironically, they are the two players which defy the all-in-one trend for now); former PC specialists like Intel and Nvidia are taking some share, especially in the higher value tablet segment; and there are new challenges as Chinese manufacturers move up the food chain.
These shifts have also put intense pressure on other former giants of the handset supply chain, like ST-Ericsson, and Freescale has already refocused its own processors on the embedded space and on specific segments like ereaders.
Future development of the OMAP architecture will now be optimized for industrial devices, although there is already speculation that TI would like to sell the handset platform, perhaps to one of the Chinese or Taiwanese firms, such as MediaTek which wants to expand its role in high end handsets. Assuming it does not, though, it is certainly following the flow of mobile industry growth towards the internet of things. This has been epitomized by ARM’s increasing extension of its architectures not just into microcontrollers and ultra-low power processor cores, but also security platforms for IoT. And where ARM goes, it must be confident most of its mobile processor customers want to go too.
If embedded and IoT segments are to provide new growth hopes for those disappointed in handsets, the same goes for the casualties of the smartphone OS wars.
RIM has already talked about targeting its upcoming BlackBerry 10 platform at embedded devices, presumably as a back-up plan should it fail in its last ditch attempt to reverse its decline in smartphones. This is not as desperate as it sounds, given that BB10 is partly built on the QNX platform RIM acquired a couple of years ago, and that was originally conceived for the automotive sector. At the recent GigaOM Mobilize conference, RIM’s Sebastian Marineau-Mes, SVP of QNX engineering, said the dream of the internet of things, all communicating all the time, would only come true with open standards. And both iOS and Android are ill-suited, he argued ‘ iOS because it is too closed, and Android because it is too anarchic for such an immature space. What is needed is ‘curated openness’, he added, with a few core functions standardized to ensure interoperability, and then a wealth of different apps on top. A combination of BB10 and HTML5 would be ideal, in RIM’s world view, but there will be plenty of other smartphone OSs manqu’s making similar arguments.
One may be Hewlett-Packard and webOS. Just when everyone thought HP had thrown in the towel in mobile products when it dumped the former Palm offerings last year, CEO Meg Whitman wants to have one last try, with a strategy which has strong echoes of the one that cost her predecessor, Leo Apotheker, his head. Not only did she say the firm would have to offer a handset ‘eventually’ and was actually working on a device, but she has revived the idea of embedding software in emerging form factors optimized for constant access to the cloud.
It was in this latter category that Apotheker hoped to harness the Palm OS and differentiate HP’s offering, providing a stripped-down device platform optimized for always- on connectivity, which could be embedded into all kinds of endpoints including HP’s own printers and PCs, plus third party products from phones to machine-to-machine modules. His strategy sounded appropriate to the emerging cloud-oriented era but was poorly communicated and executed, and Apotheker virtually closed down the activities before resigning.
But Whitman’s new ideas are not far from those of her ill-fated predecessor. As enterprise and service provider customers build out their data centers to support cloud computing, HP is preparing a compact web server with storage and network components all included along with key applications and data analysis tools. But the company wants to go end-to-end too. Although any architecture in the BYOD age will have to span virtually any cloud-connected device, HP still believes in its ability to control non-consumer endpoints such as printers and smart meters ‘ providing the intelligence and management for them, and even, like Cisco, delivering web applications to them.
Last month, HP revamped the webOS platform, which had been placed into an open source project, under the new Gram brand and said it would run that as a wholly owned subsidiary focused on building devices which would feature the OS. HP also tapped former Nokia executive Alberto Torres to run a new mobile division.
In the end, though, these OSs targeted at big, memory-rich smartphones may face the same challenge ‘ despite QNX’s automotive roots ‘ akin to that Microsoft had when squeezing Windows onto handsets. Better suited to the emerging class of modules and gadgets are the specific software frameworks created for the embedded world, spruced up with a degree of applications capability. This is where Oracle is aiming for a whole new sphere of influence for its Java technology.
Java, despite its huge presence on featurephones, proved too fragmented and low-level to succeed as a fully blown smart device system, partly thanks to various missteps by its creator Sun. Now Sun’s owner, Oracle, wants to revamp the platform for a greater role in the embedded market and has released a runtime version geared to microcontrollers, plus a suite of Java services aimed at gateways and other embedded products. The runtime is a variant of the Oracle Java Wireless Client, now also in release 3.2.
Oracle also released a software developers’ kit for the embedded code, targeting a wide range of industrial, consumer and energy systems, particularly in machine-to-machine applications. ‘We believe the next wave of innovation in IT will come in M2M and more devices connected to the internet,’ said the firm. Cinterion will demonstrate cellular modules using the code at Oracle’s JavaOne event next week and the database giant said it is in talks with microcontroller vendors about deals to bundle the software.
This comes just two weeks after Intel showed off a new framework for embedded systems, bundling its PC management code with security and realtime software from its McAfee and Wind River divisions. ‘Intel is reflecting what we also see as the desire for a complete embedded software stack so people don’t have to worry about the nuts and bolts when so many new devices are emerging,’ said the Oracle spokesperson, insisting that Java could coexist with the Intel software, which is at a lower level and closer to the chip. However, despite the talk of synergy, Oracle is mainly targeting ARM architectures – Java Micro Edition Embedded 3.2 is for systems with less than 1Mbyte of memory and as little as 130KBytes of RAM and 350KBytes of ROM, especially those based on ARM Cortex-M and Cortex-A cores.
More than 3bn devices now support Java, and the software is downloaded more than 1bn times a year, Oracle said. Such statistics are often forgotten in the mobile world where Java is regarded as old hat. But like TI, also seen as something of a dinosaur by the thrusting new mobile processor makers, the vast emerging base of the ‘internet of things’ may provide the best chance to move into the limelight again.
‘Radio-Free Intel’:
As is its wont at its annual developer forum, Intel chose this month’s IDF event to look well beyond the integration of processor, GPU and modem, and to revive its decade-old vision of ‘Radio Free Intel’. This was first posited at IDF 2002 by the then-CTO Pat Gelsinger, who helped steer Intel into being a serious wireless influencer, in vision if not always execution. He believed the company’s emergence as a powerful mobile force would rely on creating a digital radio platform which could incorporate any number of connectivity options in one unit. It sounds commonplace now, but back then the ‘wireless everywhere’ mantra was still young, especially when it would involve multiple connections. Gelsinger, Intel’s first ever CTO, set the R&D departments to work on digital radios which would be fully agile, picking up different frequencies and connectivity standards, and seamlessly and intelligently switching between them for optimal performance.
The industry is still away from that nirvana as yet (Gelsinger’s target of 2005 now seems naive), but the general vision has moved into the mainstream. The current CTO, Justin Rattner, told this year’s IDF about developments that would help Intel enable a world where ‘everything that computes will connect’ and just about everything, down to a light bulb or pet collar, will compute. He unveiled a series of technologies from the company’s research labs designed to connect all sorts of devices wirelessly, centered on Rosepoint, which he thinks fulfils Gelsinger’s requirements at last. A decade ago ‘it was more of a fantasy’, he said. “It has taken us 10 years but we think we did it.”
Rosepoint, previously described early this year, is a test chip, manufactured on Intel’s 32nm process, that combines a dual-core Atom with a Wi-Fi transceiver on the same silicon chip. This is a breakthrough because, as Gelsinger envisaged, the RF transceiver as well as the baseband is digital, which subjects the whole product to Moore’s Law (digital circuits get smaller, and usually faster and more power efficient, with each new generation of process technology). Analog circuits do not follow Moore’s Law and often get worse with scaling, hence why Wi-Fi and other connectivity options are on a separate chip within the handset.
Intel said it had to ‘rethink the radio’ to create a “Moore’s Law Radio’, with techniques to mitigate interference between its elements. Not that the result will be in products any time soon ‘ 2015 seems the best guess, around the time Intel gets to the 10nm process. The company often boasts of how its control over, and investment in, process technology allows it to drive rather than follow trends, and innovate more aggressively than fabless rivals like Qualcomm. If it gets to market ahead of others with the digital radio, it will have a powerful proof point. Qualcomm may be excellent at integrating different digital components such as basebands and CPUs, but Intel is talking about another dimension of cost, space and power savings. “We believe these radios will outperform the best of analog radios,” Rattner said.
Intel and Qualcomm throw new chips at their battle:
Some former mobile giants may be drifting towards the embedded market, but Qualcomm and Intel remain firmly focused on the smartphone processr. Their battle in the mobile world continues to intensify, and both firms have shown off new weapons this week, Intel demonstrating a dual-core Atom for tablets, and Qualcomm promising quad-core Snapdragons for high volume smartphones.
Intel is to unveil Clover Trail, a next generation Atom system-on-chip, at an event in San Francisco today, targeting tablets and high end handsets. It says there are more than a dozen tablets already lined up for the dual-core processor, though for now it will not be threatening Snapdragon in the Android territory. The first release of Clover Trail will be for the upcoming Windows 8, with an Android version to follow.
That could give Qualcomm the chance to stake out a strong position in the new wave of tablets by exploiting its deepening alliance with Microsoft to steal pole position in the ARM-based Windows RT platform (it already has a virtual monopoly in Windows Phone).
However, Clover Trail will put up a strong fight to keep tablet makers loyal to the x86 architecture, Windows 8 being the first fully fledged Microsoft OS to span both main processor platforms. The system-on-chip has early support from Acer, Asustek, Dell, Hewlett-Packard, Lenovo, Samsung and ZTE, which are all expected to show off slates, ultrabooks or hybrids of the two form factors, at the Intel event.
The chip giant claims its latest design will boast twice the CPU and graphics performance of the single-core Atom chips used in the current smartphone offering, the 1.6-GHz, 32nm Medfield. Clover Trail, also made in 32nm process, will come in two version, one at 1.8GHz, targeted at W8 tablets, and one at 2GHz, for Android smartphones. They are expected to use different PowerVR graphical processor cores from Imagination Technologies. The next big Atom upgrade will come in 2013 with Bay Trail, the first in the family to use the 22nm process with tri-gate transistors to reduce power consumption dramatically.
Over at Qualcomm, the company has added two processors to its Snapdragon S4 range, the MSM8225Q and MSM8625Q. Both will sample by the end of this year, targeting high volume smartphones and tablets.
Qualcomm has been more cautious than rivals like Nvidia about moving to quad-core CPUs in mobile devices ‘ though less conservative than Intel. The new products, upgrades to the current Snapdragon S4 Play MSM8225 and MSM8625 units, both come in four-core variants. These will support LPDDR2 memory, increasing the bus bandwidth for capabilities such as 720p display and video codec. The MSM8625Q features Qualcomm’s integrated multimode UMTS/CDMA modem, and the MSM8225Q has an integrated UMTS modem. Both processors also support Wi-Fi, Bluetooth 4.0 and FM connectivity using Qualcomm Atheros chips.
The San Diego giant also launched an SoC for China, the Snapdragon S4 Plus MSM8930, which supports all three 3G standards – UMTS, CDMA and TD-SCDMA. It will also support TD-LTE/TD-SCDMA, China Mobile’s 4G and 3G technologies, for midrange smartphones, enabling affordable 4G devices for the cellco’s ‘trial’ network early next year and trying to prevent local suppliers, and close Mobile allies like ST-Ericsson or Sequans, stealing the early glory.
Qualcomm will also be releasing reference design versions of all three processors, an increasingly important strategy to fend off low cost competitors in the midrange and emerging market smartphone sectors.