The past week has seen the titans of the processor world. Intel and Qualcomm, both hosting their annual developer events, their main showcase for the products and directions they will take in the coming year. Inevitably, the flurry of announcements from Intel Developer Forum (IDF) and Qualcomm Uplinq highlighted the fact that the two firms really are head-to-head in the bid to dominate the post-PC world, from PC replacements to embedded devices. Intel is credible, for the first time, in smartphones and tablets, but at a time when ARM-based processors are moving to 64-bit and onto Intel’s core territories (is the souped-up A7 processor in the iPhone 5S really a try-out for an ARM-based Macbook?) And both companies are determined to steal a march in the new world of the internet of things (IoT). This was the most interesting aspect of both firms’ shindigs, and are both looking well beyond just chips, and towards creating fully fledged frameworks. The processor wars will intensify, and these offerings will remain the core of both companies’ activities for years. But for a battle on entirely new turf, where the trenches have not yet been dug, they are looking to the enticingly broad, but very immature, IoT space.
One notable feature of Uplinq was the broadening scope of Qualcomm’s ambitions. Of course it has been pushing Snapdragon into tablets, notebooks, set-top boxes and other non-handset products for some years, but now it is looking beyond the processor and modem. It wants to control more aspects of the device platform, as witnessed by its stepped-up RF activities, in the shape of its RG360 offering, and the intensified focus on its Hexagon DSP (digital signal processor). This is a firm which wants to control every element of an increasingly integrated platform (it also has its graphics processor technology plus Wi-Fi/Bluetooth/powerline connectivity from Atheros).
And then it wants not only to adapt that platform for a wider range of target devices, but to create an ecosystem around it. So we see Qualcomm unveiling a smartwatch to showcase its own screen technology, an area where it has serious ambitions considering its alliance with Sharp; creating media streaming platforms akin to Apple AirPlay; and binding it all together with the software framework AllJoyn. This is reminiscent of the days of Brew (the developer system for low end handsets which was originally the focus of the Uplinq event). As in those earlier cellphone days, Qualcomm knows that real power comes from controlling a full platform, and especially its developers, not just providing the silicon.
Intel knows that too, and these are the only two players which really have both the will and the resources to build that top-to-bottom environment, and target it at multiple segments from mobility to home media/TV to machine-to-machine. Samsung Electronics could do it too, and is well ahead in some elements of the platform, such as display and memory. It has not yet acted on such a grand scale as the big two, but they will certainly be watching the Korean giant carefully in this regard.
The critical test for new Intel CEO Brian Krzanich, hosting his first IDF, will be to prove Intel can stay ahead of Samsung and Qualcomm in this all-in-one platform game. This is not about mobility any more ‘ the chance to dominate smartphones is gone, and that space will be one of dying margins and rewards anyway. Now the battle is to provide a platform for the internet of everything, in which handsets are just one ‘thing’ alongside media players, tablets, cars, health monitors, security systems and so on.
Perhaps Krzanich’s biggest moment since taking the helm was his launch, at IDF, of the Quark 1000 family of chips, Intel’s most significant offering for the internet of things. ‘Our strategy is very simple,’ he said. ‘Our plan is to lead in every segment of computing–servers, PCs, tablets, phones and beyond. Segments that are still being developed such as the internet of things.’
Intel has a well established embedded division and some important technologies, not least from its Wind River acquisition. But this goes well beyond attempts to squeeze Atom into smaller, ultra-low power form factors. Quark is a whole new design, just 20% of the size of Atom and 10% of the power consumption, and with an ‘open architecture’.
This last feature is the most interesting, but also the least precisely defined by the chip giant. It was not clear whether it had broken with x86 for this market ‘ having struggled for so long even to get its main architecture down to handset size and power levels ‘ or had created an ultra-low power variant. Either way, it seems it will be prepared to open up elements of the platform to others, recognizing that for the hugely diverse internet of things, one vendor will not make every type of chip required. It will do better to sit like a controlling spider at the center of a web of fabless semiconductor firms and associated software developers, as ARM does in the handset world.
Not that Intel is licensing the core itself, at least not yet. Intead, it will offer ‘fully synthesizable’ chips with hooks in the silicon to add others’ IP blocks, to customize it for specific tasks. It also unveiled a software stack that includes security, manageability and connectivity features optimized for IoT.
A bolder move might follow. If Intel follows through aggressively on this ‘openness’, it would certainly be a fundamental change in its business model, but one that is necessary in a world where IP and common platforms are the route to power and profit, not churning out every low margin chip and sensor that an IoT application might require. And like Microsoft, it needs to accept that its main platform may not be flexible enough to adapt to all the multitude of use cases in the internet of everything, and be ready to leave x86 behind in some areas.
It would also see Intel moving with the tide in the internet of things. While ARM licenses its microntroller cores in many segments, other conventional CPU makers are also adopting the customized approach. AMD has a semi-custom strategy to broaden its reach, and recently launched its embedded processors, which can use both x86 and ARM cores. And IBM is trying to reignite interest in the Power architecture by offering a new interconnect that allows partners to add their own technology. Soon, it will also license the whole architecture to a wider base of customers.
For now, this new approach is starting conventionally with Intel products and reference boards, to be available towards the end of this year, so presumably appearing in devices from mid-2014. Initial Quark targets are the industrial, energy and transportation segments, said Krzanich, but he is also eyeing the emerging wearables segment, and showed off a prototype connected bracelet.
If Quark was the defining word at IDF, at Uplinq it was AllJoyn, Qualcomm’s peer-to-peer connectivity platform for proximal devices, around which it is building a fully fledged developer framework for home media and the internet of things. In its quest to get Snapdragon and other chips into every device, Qualcomm is increasingly becoming a software company. Two years ago it announced a program to release applications programming interfaces (APIs) to give web apps deeper links into its hardware. The APIs would be exposed directly into the browser, boosting performance and making the choice of OS less relevant, helping cross-platform apps players such as carriers. AllJoyn is one of the most important platforms for this effort and for building full environments around Snapdragon products.
This is best seen in two growth markets for Qualcomm, home media and IoT. On the former front, it has made a string of acquisitions to extend its technology reach. For instance, it acquired Silicon Optix in 2011 and added a TV display chip to its range, targeting smart and digital TVs. And of course, its most important acquisition for pushing beyond the smartphone has been Atheros, specialist in Wi-Fi, powerline and home connectivity. AllJoyn differentiates Qualcomm in video by supporting excellent screen interaction between the TV and mobile device, argues the firm, although some argue Wi-Fi Direct could provide the same functionality more openly. Qualcomm has also developed Skifta, a DLNA-based remote access service for shifting music around the home.
Now it has added AllPlay, its latest experiment with creating a wider home media platform around Snapdragon hubs. Like Apple AirPlay or Spotify Connect, this streams music from a mobile device to multiple speakers around the home and harnesses the AllJoyn framework for connectivity and also for third party applications. Music streaming provider Rhapsody is the first flagship partner. Qualcomm said it would partner with major audio manufacturers to launch AllPlay, and that developers will have access to the AllPlay SDK before the end of the year.
In another example of how Qualcomm is targeting its core mobile technologies at every device in the home, it has worked with set-top box maker Technicolor to create Svelte (Stimulating Video Experience over LTE), a media center which runs on Snapdragon 600 and supports terrestrial TV, Android and console-class gaming. Qualcomm says the combination will take the smart TV a step further by also making it a media hub for the whole home, with interactivity over 4G.
AllJoyn will play an increasingly important role in tying some of these disparate activities together in terms of connectivity, ecosystem and third party apps, so that the platform is unified at more than just silicon level. As well as AllPlay, other significant Uplinq launches relied on, and showcased the framework, the Toq smartwatch, which was designed, not to turn Qualcomm into a gadget maker, but to highlight what it can do in the areas of wearable computing, ultra-low power displays and ad hoc peer-to-peer connectivity ‘ all hallmarks of the next generation of devices.
AllJoyn also turned up in new chips from the Atheros division, providing a low power Wi-Fi platform for home appliances and consumer electronics, and eventually for IoT. The QCA4002 and QCA4004 integrate the IP stack and networking services on the chip to enable customers to add Wi-Fi to any product with minimal development effort or cost, said Qualcomm. An on-chip processor and memory avoid the need for a system controller, resulting in a package that is sufficiently simple and low cost to go into very basic connected gadgets.
Pushing Atheros’s Wi-Fi expertise into the internet of everything is not a unique strategy of course, but Qualcomm hopes to differentiate itself by creating an apps platform around AllJoyn too.With the QCA4004, customers can write their own applications on the platform, not just use it for connectivity. This will be important in the diversity of the IoT, where the connected devices will be used for millions of different tasks, not the relatively small number that run on a smartphone. Atheros is initially targeting major home appliances such as washing machines and has a deal with Haier, which will include the chip in a washer/dryer and an air conditioner. AllJoyn provides basic building blocks for building services, particularly those that relay on connectivity and peer-to-peer communication ‘ such as notifications.
This is a very different battle to the ones Intel and Qualcomm (or ARM) have been used to. Until now, each has been knocking at the walls of a fortress the other controlled, and was forced to adopt the rules laid down by the incumbent. In IoT, there are very few rules at all, and many are being set by a bewildering array of new partners such as carmakers or utilities, hospitals or washing machine vendor – not the old familiar OEMs and carriers (or certainly not by consumers). There is all to play for, and as Intel’s ‘open architecture’ and Qualcomm’s APIs indicate, they have both acknowledged that.
But there is a key difference. Intel’s mentality remains firmly in its heartland skills in silicon engineering and the manufacturing processes to go with it. Krzanich concluded: ‘The landscape of computing has never been bigger. We showed you everything from the biggest server with E5 to the smallest SoC Intel has ever built with Quark.’ Addressing the same point about the proliferation of devices, Qualcomm CEO Paul Jacobs was more focused on consumer experience (the Toq smartwatch was a personal pet project for the former engineer) and on software. His defining comment in his Uplinq addresses: ‘We’re trying to create the common language for the internet of everything’.
Intel’s tablet challenge:
One of Intel’s key messages was its seriousness about the tablet and hybrid spaces, and it is certainly fleshing out its range of processors to follow whatever direction the ill-defined post-PC device may take. It is really taking the battle to ARM in mobile devices for the first time. Having achieved some support for its Atom-based smartphone platform at last, it is now upping its game in tablets, where it has had limited success. It is addressing the space with Z3000 SoCs, based on the Bay Trail platform and Intel’s first mobile quad-core SoC. This is the successor to the current Clover Trail and Clover Trail+ and addresses some of the shortcomings of those, especially in the area of battery life and slow performance, when compared to big tablet players like Nvidia Tegra 3.
The initial Bay Trail tablet processors are the quad-core Z3700 and dual-core Z3600 series. The former will target midrange Windows or Android tablets, the latter will be just for cheaper 7-inch Android models, and for high end slates and convertibles, Intel has Haswell, the latest in the Core platform. Bay Trail promises tablets with twice the CPU performance for generalized tasks and three times the graphics firepower. According to Intel’s internal benchmarks, a Bay Trail (Z3770) processor delivered twice the performance, in many areas, of a Snapdragon S4, a Tegra 3 or a Clover Trail.
Of course, apart from the skepticism that always attends internal benchtests, the comparisons are of a new, 64-bit platform with ARM-based options that will quickly become last year’s news too. Apple’s 64-bit A7 is just the first example of ARM being pushed into the high performance bracket, while continuing to harness the firm’s expertise in low power consumption. And while Intel may well get more uptake for Bay Trail than Clover Trail, it still faces the same problem as it does in handsets ‘ vendors are increasingly taking its offerings seriously for mobile devices, but they are adopting Atom for secondary models rather than flagships, keeping the main emphasis firmly on ARM. A critical challenge for Bay Trail will be to reverse that balance in at least a couple of high profile OEMs, before 64-bit ARM CPUs become commonplace.
Hermann Eul, general manager of Intel’s Mobile and Communications Group, also looked ahead to the Merrifield offering, which will debut in 2014 for next generation smartphones and tablets. This will use the 22nm Silvermont microarchitecture and promises 50% performance and battery life improvements over the current Clover Trail+ product. Next year will also see the long awaited integration of mobile processors with LTE modems from Intel, and Eul outlined the XMM 7260 modem, scheduled for 2014, which will support carrier aggregation, faster speeds and support for TD-LTE.
Of course, the Core family, and its new flagship Haswell, are the mainstay of what remains a critical strategy, to keep control of the PC sector even as it becomes unrecognizable in form factor terms. The 22nm Haswell will be followed by 14nm Broadwell in the Core family, still Intel’s bread and butter outside the server, but increasingly it will go into tablets, Ultrabooks, cloudbooks and other form factors yet to come. ‘The PC is in the process of reinventing itself,’ Krzanich said. ‘There’s more innovation in the PC than I’ve ever seen before. It’s battery life, it’s form factor, it’s capabilities.’
Products were demonstrated based on the Haswell microarchitecture, including a Chromebook platform. Indeed, the close relationship with Google ‘ which has usurped Microsoft in the old ‘Wintel’ alliance just as clearly as Microsoft has anointed Qualcomm as its new best friend among chipmakers ‘ was emphasized when Google SVP Sundar Pichai made a guest appearance. Doug Fisher, general manager of Intel’s Software and Services Group, told Pichai and the audience that a key Intel goal was to make the Android and Chrome experiences superior on Intel chips. He announced that Intel was now enabling 64-bit kernel support for Android, allowing for faster application response and new memory configurations while paving the way for UltraHD. All this was discussed on the same day that Apple was unveiling its own 64-bit A7 smartphone processor, sparking speculation that the chip could replace Intel in future Mac models. Of course, Intel will be working hard to avoid that, and some of the enhancements to its Core range will help. But in the event that Apple does decide to control its own silicon throughout its range, Intel will certainly hit back by throwing its weight behind developing an experience for Google’s OSs that, only a decade ago, would have been reserved for supercomputers.
It is interesting that the new Atom family includes configurations that allow Windows 8 and Android to run on the same device, indicating one of the many dilemmas of the post-Wintel strategy. Ideally, Intel would like Windows 8 Pro to be a great success across all kinds of form factors because this is the platform it owns. The relative failure, so far at least, of ARM-based Windows RT and WP8, have been a boost to Intel, but they have also thrown a shadow over the whole Windows OS. The chip giant does not want to find itself defending a dying operating system while rivals steal the spoils in the growth markets. Most emerging OSs, such as Firefox Mobile, are targeting ARM platforms, but the mighty exception is Google, which has collaborated on a deep level with Intel to optimize Chrome OS ‘ its big play for post-PC dominance ‘ and Android for x86.