Your browser is not supported. Please update it.

15 October 2007

Samsung and TI line up behind ARM to shape the fightback against Intel

The next generation of communications will see the PC and smartphone architectures converging, and the battle is heating up to dominate the trend, with contenders from both the PC/IP and the cellular side of the fence. Intel has used its existing dominance in the computer world as a springboard to push forward its UltraMobile PC (UMPC) as the default approach to high end mobile devices, but so far this has gained traction primarily in its traditional base. Intel will never achieve the hegemony in mobile devices that it has in PCs, but it needs to ensure it has a dominant influence over how the new breed of products evolves. This will be an equally vital goal for the cellphone baseband majors, and last week saw several of their number banding together behind one of their few unifying factors, support for the ARM microprocessor architecture. Intel dropped its own ARM-based architecture last year and decided to base all its mobile efforts on its own x86, creating a serious, if high risk, alternative to ARM, which is in well over 80% of cellphones. It will seek to dominate the new mobile device both at chip level and by driving software standards, increasingly around Linux. With ARM as a mascot, the cellphone chipmakers are starting to respond, setting out their own moves to harmonize an optimized processor/Linux architecture for smartphones, and using their claims of superior mobility and power management to fend off Intel’s killer arguments ‘ its unified platform and its vast developer base.

While Intel’s first stab at the next generation mobile market, the UMPC architecture, has mainly gained acceptance among traditional notebook makers, most of the handset manufacturers, which will be the key decision makers along with the consumer electronics giants, have yet to show their hand. A year ago Nokia might have been assumed to be sticking with its old familiar, ARM licensee Texas Instruments, for its shift into the UMPC-class space between cellphones and notebook ‘ the key device category for the mobile internet ‘ but with its chip purchasing strategy now increasingly multisource, the Finnish giant will be one of the most important influencers of the shape of the new mobile device sector. Last week we saw signs that Intel might get some of the benefit of Nokia’s favours after all ‘ the Finn’s Internet Tablet is the first product in an ambitious roadmap geared to controlling the device agenda itself, potentially at Intel’s expense (see Wireless Watch October 8 2007), but despite possible conflicts of interest, Nokia will use Intel chips in the WiMAX versions of the Tablet.

Such moves make it even more important for TI to form its own new commercial and political alliances, and the autonomy that it has reluctantly won from Nokia’s agenda is reflected in its early support for the ARM initiative ‘ one that will challenge the Intel bid to put UMPC at the forefront of the new mobile client world. The group also contains Samsung and Marvell, plus open source specialists MontaVista and Mozilla, and Movial, a creator of mobile Linux development platforms.

Specifically, the partners will create an open source Linux-based system for UMPCs and other high end mobile devices, that will include the operating system, application development framework and browser. This will directly take on the efforts of Nokia, Intel and others to create similarly integrated stacks, using combinations of their own developments and major open source products, and optimizing them for selected chip platforms. The importance of the joint effort is that it pools the resources of several giants around an officially neutral force ‘ ARM ‘ and so should accelerate the creation of a new platform and reduce the damaging fragmentation in mobile Linux. This, in turn, will help answer one of Intel’s most telling arguments in favour of its x86/Linux combination, that this is a fully unified platform. Greater harmony among the cellphone chipmakers would help them, and their customers, attract more developers, and ARM’s involvement enhances the UK company’s position as it rises to the Intel challenge, making it a harmonizing factor across its licensee base.

The ARM group’s plan is audacious and takes on several existing industry powerhouses in the fight to create a compelling mobile internet user experience around Linux, at an affordable cost so that UMPCs and internet smartphones can go mass market. This attempt is key to the future of mobile broadband and so of the business models of most device makers and operators. Key attempts so far include:

‘ Intel bidding to take the lead with Windows and Linux versions of UMPC, and with Moblin, its take on defining de facto standards for a Linux-based mobile internet platform (key Linux partners in this are Ubuntu and Red Flag).

‘ Nokia moving heavily to Linux for certain devices, notably Internet Tablets, backing open source products like Maemo and Safari. The chances of a Nokia-Intel axis with the open source software optimized for Intel processors are growing. The two vendors are the key supporters of the recently formed Gnome Mobile and Embedded (GMAE) project to bring the Gnome Linux user environment to small devices.

‘ Motorola taking a guiding role in the most effective group so far to aim to unify fragmented mobile Linux implementations and interfaces, the LiMo Foundation. Broadcom recently became the first silicon major to join.

‘ Sun announcing JavaFX Mobile, a powerful mobile software system that runs on top of Linux, and may be adopted by Samsung.

‘ Google potentially releasing a Linux-oriented software platform tied to its own hardware design for internet phones.

‘ Both former halves of Palm ‘ the PDA maker itself and Access, owner of its former software arm – seeking to reassert influence by combining elements of the old Palm user interface with Linux.

‘ Other initiatives include the Linux Phone Standards (LiPS) industry body, which mainly focuses on middleware and recently teamed up with the OMA.

The ARM/TI/Samsung effort will either take on, or find common cause, with many of the above, as most of them move beyond hobbyist enthusiasm or pure research to start to create commercial class platforms and sign up their major handset vendors and chip partners.

As with other activities, the key aim is to make Mobile Linux a more attractive proposition to end users, by improving the user interface and making more applications available; and to developers, by unifying the currently fragmented platform through a harmonized upper layer. With each handset semiconductor company implementing both the ARM designs and the Linux distribution differently in their particular chipsets, no single platform has sufficient bulk to attract mainstream developers, especially when the handset makers add their own variations. This is the major argument used by Intel for its attempts to put its x86 architecture into mobile products ‘ that it is unified and so provides a simple platform for developers, especially with the commonality with the PC. Of course, Microsoft used the same argument to back up its attempts to make Windows a dominant mobile OS, and found there was more to it than that, but certainly a higher degree of simplicity for the Linux applications community would bolster ARM’s cause in the face of a potential onslaught by Intel.

The two companies had until recently largely different territories and business models, but are now looking to lock horns directly ‘ a battle ARM CEO Warren East is careful to present as one between Intel and the 200-strong ‘ARM Collective’ of licensees, not between the California Goliath and UK David. Intel’s own major implementation of the ARM technology was in its XScale architecture, a successor to the StrongARM technology it inherited from Digital Equipment, and once designed to be the basis of its own attack on the cellphone sector. However, it sold that business to Marvell in 2006 and is now focusing almost exclusively on x86.

At the recent Intel Developer Forum, the company made its challenge to ARM far more explicit than it has before, prompting an equally direct response from East at his own company’s developers’ conference a week or so later, and sparking the public announcement of the Linux initiative. On Intel’s part, it showed off quad-core architectures, mobile web interfaces and platforms, and Linux efforts, all of which will combine, according to its executives, to create a superlative mobile internet platform that will build on the heritage of the PC and be a clear step ahead of systems from chipmakers using the ARM designs. ARM responded with its latest architecture, the single- and multi-core Cortex A9, and with the Linux plan, which showed three of the most significant ARM supporters, TI, Samsung and Marvell, lining up beside it, (though of course the addition of Qualcomm would have been even more intimidating for Intel. However, Qualcomm’s agenda is complex in relation to Linux and mobile internet, and it is unlikely to line up behind any mainstream standardization efforts as yet, even for the joy of rattling Intel’s plans ‘ see inset).

As well as pointing to the huge base of developers for x86 in the PC world (where AMD is the other major user of the architecture), Anand Chandrasekher, head of the Intel Ultra Mobility Group, fronted a string of direct attacks on ARM. These played down the power issue and went heavy on an area where Intel claims an advantage, suitability for a complex PC-style internet experience. One presentation counted the number of errors encountered while browsing the internet on Intel notebooks versus tablets or phone based on ARM cores, with Intel implying the latter did not cope well with full web pages because of compatibility problems. However, Intel would not identify the devices used in the comparisons, nor whether they supported other features that would be crucial to mobile internet performance, such as Flash or fully functional browsers.

Despite the developer advantage, it would be hard to fancy Intel’s chances of securing major gains against ARM licensees in the near future, unless it comes up with a more radical new approach to mobile device design. It will clearly gain a dominant market position among the traditional PC vendors, but their collective share of the next generation mobile market is set to dwindle not grow. If, for simplicity’s sake, we define current mobile internet devices as being wireless notebooks and PDAs from the PC camp, and internet enabled cellphones, the former has the larger share of the web using population, but this is starting to be eroded slightly by the webphones. This drip-drip effect will become a flood once the internet experience on truly mobile devices (non-notebooks) is sorted out, which is likely to happen ‘ unless Nokia, Apple and the others really drop the ball ‘ within four years. In the next decade, the boundaries between the notebook, UMPC and smartphone will have become merely ones of external design and user preferences, rather than being tied to specific architectures, so we should see Intel and the ARM supporters competing far more directly for the same customers, plus the addition of a new range of purchasers coming from the consumer electronics device arena.

In this free-for-all, it would be foolish to underestimate the influence Intel can have through its funding of clever partners, its active approach to standards bodies and industry groupings, its foundries and its vast ecosystem. But even so, in the past it has usually been able to leverage the grip it has on its existing channels and customer base, and the fact it has had no effective competition. Like the internet services players like Google, its mobile efforts will be hampered by its lack of experience and track record in mobility itself. And it will be trying to appeal to a new set of customers who may prefer the security of an array of vendors backing an increasingly harmonized mobile internet architecture, rather than over-dependence on a single chipmaker, with the risk of non-delivery.

If ARM continues on its current path, it looks set to retain greater market presence than Intel in mobile internet (even though it faces its own challenges in terms of the revenues and licensing structures it can continue to sustain in this more open world). Without an ‘ARM Inside’-type branding, the UK company remains largely invisible to the general world, a factor that may sometimes have damaged its rating by stock analysts, but which has helped save it from being perceived as an Intel-like quasi-monopolist, and so sparkling hostility and counterattacks from the mobile community. This changed once Intel decided to get serious about x86 as an alternative architecture for smartphones, but in many ways will just spur the ‘ARM Community’ into more coordinated action, as the Linux initiative indicates. ARM may well find itself the mascot for the efforts of TI, STMicro and others to keep Intel off their turf, a role it could not take when Intel was still using the ARM cores itself.

Along with market presence and the commercial and technological understanding of true mobility, the ARM faction claims low power as its other critical advantage over Intel, and much of the latter’s challenge in mobile devices will depend on it delivering on the ultra-low power technology agenda it set out last year. Once the microprocessor war shifted, with the advent of ever smaller devices, from mips per dollar to mips per watt as the driving force, embedded technology companies like ARM were thrust to center stage with the low power advantages gleaned from generations of their products. In this, as East put it recently, Intel has ‘significant catching-up to do’, though it has set out its stall. Next year it will introduce Silverthorne, an x86-based chip for UMPCs that is targeted to match many ARM chips by consuming under a watt of power. Following Silverthorne will be Moorestown, which was previewed at the IDF, and is part of an ultra-low power R&D effort geared to cutting idle power consumption by a further 10 times by 2010.

The other vital aspect of the Intel strategy to make x86 and UMPC ubiquitous is increasingly ‘ if belatedly ‘ Linux support, hence the importance of the ARM group’s latest move, which to some extent builds on ARM’s Connected Mobile Computing project, kicked off in early 2005, which looked to define its own platform for the emerging category that Intel calls UMPC, hoping to create this in the cellphone’s image rather than the PC’s. This battle is set to become bloody as real markets for mobile internet applications open up among enterprises and consumers, and as the device giants start to make their choices. Nokia and (reportedly) Apple are showing some support for Intel, while Motorola is likely to stay firmly with suppliers like TI. A united front displayed by some of the baseband majors, with ARM as a unifying factor, is a worrying factor for Intel, not to mention the unpredictable response of Qualcomm. But fragmentation remains the bane of the cellphone and the mobile Linux sectors, and TI and its allies will have to address this for real, not just through talking shops, if they are to rob x86 of its greatest advantage.

The Cortex A9 architecture:

ARM Holdings unveiled its latest processor architecture at its recent developer conference, topping up its range and targeting smartphones, mobile computers, consumer electronics, networking and in-car entertainment systems. Initial sign-ups for the design, which covers single and multicore processors, were NEC, Samsung, STMicro and Texas Instruments.

The Cortex-A9 processors can execute up to four instructions per cycle in devices clocked at over 1GHz. The single core variant promises low time to market for cost sensitive products like midrange phones. It delivers 2,000 DMips of performance when
designed in a generic 65nm manufacturing process from TSMC, and the core logic costs less than 1.5 square millimeters of silicon, ARM claimed. With the multicore implementations, performance scales up to 8,000 DMips, up to 16 times that found in most high end phones, so that complex web 2.0 and multimedia applications can be supported, the company said. This is a clear indicator of ARM’s planned response to Intel’s headstart in supporting high end, PC-class internet applications.

New features include an Accelerator Coherence Port for increased system performance and lower power; Advanced Bus Interface Unit for low latency in high bandwidth devices; Multicore TrustZone’ technology for hardware-based security; and a Generalized Interrupt Controller for software portability and optimized multicore communication.

Symbian, the leading supplier of smartphone operating systems, said it would support the Cortex-A9 Multicore and the ARM Symmetric Multiprocessing architecture to improve power/performance ratios in handsets running SymbianOS. The company said battery life will be improved by accessing some cores only when running demanding multimedia applications and powering them down when these are not in use.

Qualcomm and ARM:

Qualcomm, despite its current travails, will be one of the most significant obstacles in Intel’s path in the handset world, and therefore an important member of the ARM group, even if an awkward one.

Qualcomm is one of the only companies to have a full architectural license from ARM ‘ though Intel also had one via Digital ‘ and this enabled it to leap ahead in terms of modifying the design for mobility when it moved into this field in 2005 with the release of its Scorpion microprocessor, and subsequently with its Snapdragon architecture. It opened a chip design facility in 2003 and acquired the rights to modify the ARM architecture in order to optimize an application processor for a smartphone chipset, which emerged as Snapdragon.

The 65nm Scorpion ARMv7 core delivers 2,100 DMips peak, while consuming under half a watt of power. This showed Qualcomm quadrupling the power/performance of its products based on unmodified ARM cores. Key enhancements have focused on multimedia engines to accelerate processing of media applications while keeping power consumption low.

Scorpion is part of Snapdragon, which also includes a 128-bit SIMD (single instruction, multiple data) processor and a 600MHz DSP to accelerate multimedia applications. On the RF chip, Snapdragon will incorporate options for EV-DO, W-CDMA, HSDPA and HSDUPA, broadcast television and multimedia, Wi-Fi and Bluetooth.

This was unveiled a year ago and represented one of Qualcomm’s first moves to extend its reach beyond pure handsets into other consumer electronics and home media devices, and also a key challenge to Intel in the PC/smartphone crossover space.

‘Qualcomm is driving a shift that is taking place in portable electronics by adding ubiquitous connectivity with zero compromise for the user experience to an extended range of devices,’ said COO Sanjay Jha. Samsung was the first public customer, and is likely to use Snapdragon in its ultra-mobile PC, scheduled for early 2008.