Handsets and notebook PCs still account for the bulk of computing device sales, and will do so for years, but both are the victims of commoditization and margin collapse, and the latter is in slow decline. That means growth for the firms which have supplied most of the chips in these categories must come from new form factors, the ‘post-PC’ products which will combine the connectivity of smartphones with the computing and multimedia power of PCs and other consumer devices like consoles.
Intel and Qualcomm, as dominant powers in the older segments, have the added challenge of ensuring they are not just replacing one device type with another, perhaps at lower price points, rather than generating additional sales. So while it is vital for both to thrive in the post-PC market, in its fluid and unpredictable form, to offset the squeeze on their traditional device strongholds, they also face the challenge of companies which were not significant concerns in PCs or handsets, but may have a renaissance in tablets, Chromebooks, hybrids and so on.
AMD’s extremely late entry into mobile processors, while largely the result of incompetence, could now be post-justified as waiting for the right moment, the market which the company could address with a unique solution, not a me-too. For the first time, with its new mix-and-match x86/ARM architecture potentially able to target any number of emerging form factors, it may be a genuine worry to Intel, rather than a useful cushion against accusations of x86 monopoly. Similarly, the particular strengths of Nvidia (graphics) and Broadcom (connectivity) may have greater impact in post-PC devices than they have done in handsets.
So the timing is awkward for the big two, and that’s without the added problem of having to compete head-on with one another for the first time, both desperate to be the kingpin in larger-screened mobile computing devices. They will support almost any form factor, in the hope it will be the next big thing, and under their control. Sometimes this fails (Qualcomm’s smartbook, Intel’s UMPC), sometimes it succeeds, though often because they are really just extending an existing platform (Intel’s Ultrabook, Qualcomm’s tablet designs). The real deciders in this game will be the platform providers, and the chipmakers know that their biggest opportunity lies in creating the post-PC version of Wintel.
Ironically enough, Qualcomm is working most closely with Microsoft, whilel Intel becomes Google’s ally (and they both beg for crumbs from Apple in the meantime). These new alliances reflect the new landscape in post-PC. Qualcomm’s leadership in handsets, and the ties to Google that implies, will not change, or be severely threatened by Intel. Similarly, in notebooks, even in their new guises (Ultrabooks), there is no point trying to erode Intel’s dominance nor its marriage to Microsoft.
But in the new devices, the dymamics are very different. Chrome OS is Google’s key weapon in its bid to ensure that ‘post-PC’ is not just a rethinking of the Windows device, but a fully cloud-based platform running its browser/OS, and therefore showcasing all its web services and adverts. In anointing Intel as the primary chip partner for Chromebooks and Chromeboxes, it is blatantly seizing the Wintel architecture and ecosystem of vendors and suppliers, from under Microsoft’s nose.
Intel will certainly hope that the Wintel platform will live on, and the Ultrabook and the x86-based Surface tablets are important in its strategy. But to a large extent they are replacements for existing business. By taking a bet that Chrome will be a significant platform, Intel is burning a lot of boats with Microsoft, so its gamble is a serious one. It must believe not only that Chrome will provide replacement revenues as PCs decline, but that it will create additional markets, where Intel’s architecture can be the default. This is a powerful statement that Intel believes Google to be a more likely generator of new markets ‘ additional users, applications and services ‘ than Microsoft.
Qualcomm did announce support for Chrome OS in 2010 and, like Intel, will leave no potential platform untapped. However, it has a reverse strategy. The heart of its post-PC push is the tablet, a popular design in which Intel has largely been relegated to the challenger position it has in handsets ‘ partly by Microsoft’s decision to create Windows RT for ARM, partly because most tablets are just stretched smartphones and so naturally use smartphone chipsets. Qualcomm hopes to turn the tablet into more than a big smartphone and see it performing all the functions of a post-PC device, hence the investments in souped-up processors, advanced displays and graphics and so on. Tablets need to knock out PCs, not sit alongside them forever, and while for Google that means making Microsoft irrelevant, Qualcomm’s approach is to help Windows survive in new form factors, but taking the old Intel role for itself. So as Microsoft smarts under Intel’s increasingly vocal support for its would-be destroyer, Chrome OS, Qualcomm has done much of the work to create optimized systems for Windows Phone and Windows RT.
The bet Intel is taking on Chrome OS, and more importantly on Google’s ability to drive the post-PC market, was seen clearly at this week’s launch of new Chromebooks. The event was Google’s big Chrome push of the year, and as well as new, faster models, its main themes were closeness to Intel, and functions which would help Chromebooks be a complete alternative to a PC. For all Google’s dream of users keeping all their data, content and applications in the cloud, accessing them from a mobile, inexpensive browser appliance, the reality is that most consumers are not ready for this. The norms of the PC ‘ significant local storage and processing power ‘ die slowly, as many attempts to create the thin client (Sun’s network computer, the netbook, now the Chromebook) have illustrated.
In its bid to make Chrome OS as dominant in enterprise and large-screened environments as Android is in handsets, Google cannot afford to be religious, and while cloud services still lie at the heart of the Chromebook proposition, there are many concessions (which also involve price tags closer to those of notebooks than originally envisaged, at least for high end models like Google’s own touch-enabled Chromebook Pixel). This week, the search giant announced more offline functions for the platform, making it more comparable to the conventional PC. Users now have the ability to edit videos and watch full movies offline. Caesar Sengupta, VP of product management for Chromebooks, acknowledged: ‘As the ecosystems evolve, more and more developers are writing apps using Chrome APIs so they work offline.
Google is targeting Chrome OS very aggressively at PC replacement now. Initially the devices were largely bought as companions, or by the education section, where there is evidence of replacement of Apple, the darling of schools and colleges (the Chromebook’s share of notebook sales to the US education sector has risen from 1% to 19% in 2012 to 2014). Now the web firm is going after the PC’s enterprise heartland, especially the opportunity provided by the death of XP. US retailer Woolworth’s has adopted Chromebooks instead of PCs for all its employees, for instance.
As well as the cloud-based approach, which can save costs and increase controllability for large enterprises, Sengupta touted the benefits of the constant updating of the operating system, in response to user demands. ‘It is an OS that updates every six weeks. It keeps getting better,’ he said. Google is also bringing more and more features from Android to Chrome, including Google Now, its rival to Apple Siri, which uses speech recognition to process requests.
Google is also being more proactive than it was in Android at pushing Chrome OS into new form factors, partly because its browser-based design makes that relatively easy. The most important example has been the Chromecast HDMI dongle, a challenger to Apple AirPlay and rapidly becoming a full apps platform supporting games, adverts and other new models. The Chromebox is a mini-desktop format while another new direction in 2014 will be Chrome OS-based digital signage products which can be updated automatically from the cloud, and clearly fit well with Google’s ad-driven business.
So while Chrome OS started life powering toys and appearing to promote a conflicting approach to that of native Android, it is now comfortable in its own identity and, as much as it will converge more with Android over the years, having a clearly different role in the quest to squeeze Microsoft out of the post-PC market.
Google also used the event to boast of its close alliance with Intel and to show off new models, from a range of vendors, which will run on Intel’s new Bay Trail design. This architecture, which greatly increases power efficiency, will enable Chromebooks to ‘grow up’, said Intel. This is an important hedge against the tablet remaining an ARM-based form factor ‘ apart from the x86-based Surfaces, which look unlikely to worry the iPad, Intel has been slow to get into Android tablets, but in Chromebooks, it starts in pole position, and ARM-based chip providers are the challengers.
Early ARM-based Chromebooks received poor reviews for functionality though Samsung went a long way to address those criticisms when it launched its Chromebook 2 model earlier this year. But Intel has responded to the enhancement of the ARM Chromebook proposition convincingly with Bay Trail, which offers similar power and cost efficiency. There are four current Chromebooks running Intel processors, and the firm says that figure will rise to 20 this year. The new models will either Bay Trail or Core i3 chips, and will ship from well-established Intel partners such as Lenovo, Acer, Dell, Toshiba and Asus. This line-up of vendors may highlight the desperation of the PC vendors to find a ‘next big thing’ which they, not the phonemakers, control, but it also indicates that Chromebooks are getting sufficient scale to become a challenge to Windows PCs and post-PC devices.
Intel, of course, cannot back one horse too exclusively, nor can it loosen the ties with Windows too hastily. Intel VP Navin Shenoy told the launch event that supporting Chrome is part of a broader strategy to get the firm’s chips into ‘anything that computes’, regardless of operating system or form factor. ‘We will embrace multiple operating systems,’ Shenoy said. ‘We want our processors to work best with any operating system in the market.’
But Chrome OS has certainly been promoted to the premier league of Intel’s supported platforms. Intel, it was revealed, is the second biggest contributor to the open source Chrome OS after Google itself. Shenoy pointed to heritage in this market ‘ the first prototype Chrome OS device, the CR-48, ran on Intel silicon ‘ and also to advantages over ARM, including 64-bit support.
Just to reinforce the ironies of Wintel morphing into ‘Googletel’, on the same day that the friends held their event, reports surfaced that Microsoft was planning to introduce a ‘Surface mini’ tablet, for the first time using Qualcomm’s Snapdragon system-on-chip.
This is not a direct blow to Intel of course ‘ there are separate implementations of Surface for x86 and ARM architectures, with different Windows 8 variants ‘ but any progress by Qualcomm within Microsoft, and tablets in general, is worrying for the chip giant. Intel will have been pleased by the poor response to initial ARM-based Windows RT devices, clinging to the hope that Windows tablets would remain firmly x86-based, especially with the significant advances in the architecture’s power efficiency. A smaller, cheaper, slicker Surface could redress that balance, and in that scenario, Intel would rather see Nvidia ‘ the incumbent provider of processors for the Surface RT ‘ get the win, rather than its b’te noire, Qualcomm.
Sources told Bloomberg that Microsoft would unveil the new Surface on May 20 at a New York event where it is also expected to detail some of its plans for the recently acquired Nokia devices. Although Microsoft CEO Satya Nadella seems less interested in hardware than his predecessor, Steve Ballmer, he has made ‘mobile first, cloud first’ the mantra for his leadership and, for now at least, he is stuck with Ballmer’s belief that an integrated hardware/software platform was essential for mobile success.
Turning to Qualcomm would be no surprise ‘ the two companies have worked closely together for years and the chip designer was one of the few influential supporters of Windows Mobile in its dark days. But it would be an important win for Qualcomm, which has struggled to repeat its smartphone success in tablets. Nvidia put tablets at the heart of its mobile strategy, pushing high end credentials in graphics and multicore processing, and focusing on a sector where its lack of a fully integrated modem was less significant. Now that Nvidia has integrated the modem technology it acquired with Icera, and Qualcomm has beefed up Snapdragon, the two are in more equal combat for the rather squeezed tablet space between the iPad (which uses Apple’s own processors, and Qualcomm modems in the cellular models) and the low cost Android models, which increasingly use chips from Chinese vendors.
The other new factor in the post-PC landscape, at least from next year, is likely to be AMD, which has consistently failed to make an impact in the mobile world beyond powering a handful of Windows/x86 tablets for second tier vendors, and some lightweight mobile notebooks. It has not, until now, come up with a post-PC platform of its own, akin to Intel’s Ultrabook, nor penetrated the handset world.
But now, it has unveiled its ‘ambidextrous’ strategy, combining ARM and x86 platforms and instruction sets (see separate item), and is going agnostic in operating systems too, bringing native Android support next year.
AMD has clung to its Windows past for far longer than Intel, but is facing the same problems of PC decline, and has made even less headway in mobile devices than its rival. It has been getting closer to ARM, and has now announced that it has taken an architectural licence and will base its growth strategy on Skybridge ‘ a platform which will mix and match ARM and xi6 cores and instruction sets to create new and targeted offerings.
The firm took an ARM licence in 2012 but has now upgraded it to an architectural deal, which allows it to design its own cores.
Among the results of its new approach will be 64-bit ARM-based chips running native Android, though unlike Intel, AMD does not plan to run the Google OS on its x86 processors. While Intel has worked closely with Google to optimize Android for its architecture, AMD maintains that the operating system naturally works best with ARM-based designs.
The ARM/Android products will be targeted at tablets and thin clients, as well as embedded devices, but not smartphones, said Lisa Su, general manager of AMD’s global business units. AMD has clearly missed the boat in handsets, though the official reason given was that it wants to focus on devices that rely on strong computing and graphics performance. That would play to another of AMD’s recent differentiators, the tight integration of CPUs and graphics processors (GPUs). Skybridge will enable ARM Cortex-A57 64-bit cores to be integrated with AMD’s own Graphics Core Next (GCN) GPUs.
Now the strategy of allowing customers to be very flexible in how they configure their systems could enable AMD to support experimentation with new form factors and variants, driving new ideas and customers in the post-PC space. Launching its new Mullins and Beema mobile chips (see inset), AMD showed off and experimental nano-PC and a ‘discovery tablet’ based on the new processors, to showcase the flexibility of its platform in addressing whatever form factors its customers might dream up.
The chipmaker will be hoping for greater success in the tablet space if it supports Android. Its previous offerings in this segment, Z-01, Z-60 and Temash, found their way into only a few Windows slate models, and none was a big seller. Mullins only supports Windows in native mode, though Android can be added as a virtual layer through the Bluestacks emulator.
Amid these attacks from all sides, should Intel really take seriously the opinion of JP Morgan analysts, that it should throw in the towel altogether in mobile? The influential analysts issued a statement this week calling for Intel to shut down its mobile and communications group in order to improve profitability. ‘We continue to believe mobile is unprofitable for Intel,” the statement read. ‘We continue to believe Intel will lose money and not gain material EPS (earnings per share) from tablets or smartphones due to the disadvantages of x86 versus ARM and overall low profitability of the tablet and handset processor market. If Intel were to shut down its mobile business, we estimate it could unlock roughly $0.50 in 2015 EPS.’
It’s true that Intel has tried multiple times to be a big force in mobile, and has failed, and that the mobile and communications group suffered a $3.1bn operating loss in 2013, and losses of $929m in the first quarter of this year. However, it has a stronger platform this time around, and mobility is no longer a separate market from PCs, but the continuation of that segment. There is an argument that Intel has been too little too late in handsets and should not focus on that segment just at the moment when it is becoming so commoditized. But if it moves away from tablets and Chromebooks, it will eventually be exiting devices altogether, as these products replace PCs. That would be a very big and risky shift from volume to profitability, and not one new CEO Brian Krzanich shows any signs of taking. Julie Coppernoll, VP of the mobile and communications group, told EETimes that her division was clearly ‘an investment – a well understood strategy for Intel’s computing leadership perspective. We feel that we have a plan, we understand how we’re going to deliver, and are making investments that improve our particular competitiveness from a platform cost perspective.’
For the first time, that rings true. Intel has managed to be competitive with ARM in power consumption at last, and the new Atom generation, Broxton, will be capable of surpassing, not just matching, rival offerings from Qualcomm and co, with new CPU and GPU cores, 64-bit support and continuing power efficiency improvements. Coppernoll pointed out: ‘We’re confident that our Atom (Bay Trail) processors will continue to be the performance and performance per-watt leader. Keep in mind we are also manufacturing these chips now at 22nm, and we are in the process of starting up our 14nm process.’
The weapons are being assembled and sharpened on both sides, but it remains unclear exactly which device types Intel and Qualcomm will end up battling for. Sengupta would not comment on persistent rumors that a Chrome OS tablet is on the cards, but he did sum up the way all players feel about the nascent post-PC space, concluding: ‘The world is in a state of great flux right now.’
Mullins may be AMD’s tablet breakthrough:
AMD has announced two new mobile accelerated processing units (APUs), as part of its new Puma generation of cores. The first chip, Beema, is a continuation of conventional AMD designs for notebooks, but the second, Mullins, may be its first credible offering for tablets, because of its reduced power consumption, combined with Puma’s increased processing power and integrated security.
“Our focus with Mullins and Beema in the third generation was overall performance per watt, fitting in to smaller form factors, and overall improved experienced,” said Kevin Lensing, senior director of mobility solutions, in an almost verbatim echo of Intel’s boasts for Bay Trail.
Mullins can be dual-core or quad-core with clock speed topping out at 2.2GHz and it halves the power consumption of previous designs, allowing for twice the system productivity, platform design engineer Ben Bates told EETimes. “Last year our high end products scaled up to 8 total watts of power. Now we’re delivering the same experience at roughly half of the power level, enabling more aggressive form factors.”
Improvements in power consumption are partly down to a silicon temperature tracking algorithm that allows for better management of the phone’s skin and can enable performance increases, on certain workloads, of over 60%. Also, energy is saved by
non-APU components completing basic tasks quickly, and increased memory support from low-power DDR3-1866 technology, which will enable Mullins to be used in high performance fanless tablets.
Mullins and Beema use AMD’s Enduro switchable graphics mode to optimize performance and power. The chips double the graphics performance of their predecessors, using the vendor’s GCN (Graphics Core Next) architecture. The APUs are also AMD’s first to have security integrated at chip level with a platform security processor (PSP) based on the ARM Cortex-A5 and ARM TrustZone technology. This creates a separate virtual CPU for sensitive processes such as payments authentication, and for storage of that data.
New software capabilities include touch-free gesture control, facial recognition and the virtual Android OS using BlueStacks. The first announced customer for the new APUs is Lenovo.
AMD’s mixed-core platform keeps all the design options open
When you’re late to market, you either have to come in with something very cheap, or something very different. AMD is pursuing the latter strategy in its very belated mobile push, refusing to target tablets under the $200 price mark with new processors like Mullins, and hoping its flexible architecture will drive new form factors and price/power/performance benchmarks, so that it can make up for lost time in the post-PC world (see lead article).
AMD’s bid for relevance will rest on its HSA (Heterogeneous System Architecture), which already boasts a Foundation of industry supporters. This is developing tools and standards around the concept of mixing and matching different cores to optimize performance and power consumption for each individual target machine ‘ from supercomputers and servers to post-PC devices to embedded products. The company announced HSA and started to gather partners almost two years ago, but is now showing off its own products and roadmap, harnessing the approach.
Processor design is all about mixing and matching different cores and accelerators to achieve the perfect balance of performance, power efficiency, graphics and functionality. This is a complex task which AMD hopes to simplify with HSA, and in so doing, greatly enhance its position in the SoC (system-on-chip) space, by providing an open approach to mixed-core design.
While SoC makers increasingly integrate multiple CPU and graphics cores, these are hard to program because they have separate memory pools. The problem can worsen when other accelerators for specialized tasks, such as DSPs (digital signal processors), are added too. HSA enables CPU and GPU cores ‘ and other elements in future ‘ to share resources such as memory.
Although AMD has often presented HSA in the context of mainstream computing and supercomputing, there is intense interest from the mobile sector, where devices are packing an increasingly wide variety of functions and chips, and where a highly integrated, power efficient SoC is key to success. ARM, Imagination, Mediatek, Qualcomm and Samsung are among the members of the HSA project. Imagination plans to support the architecture on all its cores, though it has not given a timescale, but this could aid its efforts to expand from being a GPU core specialist to ‘ via the acquisition of MIPS ‘ providing mixed-core platforms.
AMD’s own implementations will fall under its newly unveiled ‘ambidextrous computing’ roadmap, which aims to produce chipsets harnessing x86 and ARM cores, which are more than the sum of their parts. It has strengthened its hand by extending its 2012 ARM licence to be a full architectural licence ‘ something shared with only about a dozen others, giving AMD the right to design its own ARM-based processors rather than using the standard cores made available by ARM.
However, while AMD will ship its first ARM-based chips later this year, it will not have its own designs ready for market until 2016. Its first ARM chip is targeted at the nascent ARM-based cloud server space, signalling that the vendor aims to straddle both architectures even in a segment where it has x86 presence established.
The most important aspect of the hybrid approach in the device world will be the flexibility it gives OEMs and ODMs, when it is so uncertain which form factors will achieve critical mass. This is where AMD’s Project Skybridge comes in, promising a pin-compatible x86/ARM chip in 2015. That would allow a customer to swap out an ARM-based CPU with an x86 CPU on the same board, and this platform is firmly targeted at embedded, tablet and PC sectors, enabling their manufacturers to support two architectures with a single hardware design.
The ARM elements of Skybridge will be AMD’s first chips with native Android and will support HSA specifications for shared memory and queuing between cores. AMD has upgraded its current on-chip fabric, I/O and memory management units so they can support both cores.
Then in 2016, AMD will roll out K12, a custom 64-bit ARM core, possibly made using the upcoming 16/14nm process with 3D transitors, which it will plug into its SoCs. This will see AMD bravely taking on several of the other ARM architectural licensees, all looking to roll out new SoCs with custom cores at around the same time ‘ they include Qualcomm, Broadcom, Cavium, Nvidia and Applied Materials.
Analysts are impressed that AMD has wooed back veteran chip designer Jim Keller from Apple. He told EETimes: “We got the IP, SoC and fabric right, then worked on how to plug the cores into them to get the [SoC] plumbing right’. However, he conceded: “We are doing a new x86 and a new ARM core and a new SoC methodology. That’s pretty daunting.’
With this major push, AMD is gambling on the need for manufacturers to keep their options open, in a situation where x86 and ARM designs are increasingly competing for the same spaces, rather than occupying different ones. Unless Intel were to take an ARM architectural licence (something it divested when it sold one of its previous failed mobile efforts, XScale, to Marvell), AMD will be the only company able to design on both instruction sets.
On the server side, the interesting target will be the web giants. Google and Amazon, in particular, are reported to be hiring chip designers to build their own customized ARM-based architectures for their cloud data centers. Their aim is to ensure the designs meet their particular needs, and to reduce cost and reliance on partners like Intel. However, AMD ‘ armed with its architectural licence ‘ could save them a lot of investment and risk by becoming, in effect, a custom designer for at least one web major. When it launched HSA, it pointed to the way big players are swinging back towards the old computing days of heavily customized processors ‘ seen in devices (Apple’s A5/6/7 for instance) and now in webscale servers.
Its flexicore architecture is designed to support that shift and to expand its custom chip business, since its chances of leading a major off-the-shelf segment such as merchant smartphone SoCs are limited. AMD estimates that semi-custom chips will generate 20% of its 2013 sales, and rising ‘ for instance, it designs a processor specifically for the Sony PlayStation 4.
AMD is also looking even further to support new architectures which will evolve for webscale data centers and cloud platforms. Amazon has been reported to be considering not just its own chip, but a whole fabric for its data center. AMD has been creating a fabric, based on its acquisition of SeaMicro, bringing together dense arrays of low power Intel Atom chips, which can communicate via the fabric to share jobs and balance loads. Here, Intel could go the same way, increasing its own willingness to create semi-custom chips and fabrics around Atom or Core. But the dark horse may be IBM, which despite rumors of an impending exit from chips, recently announced a roadmap to push its Power architecture into the vanguard of the cloud server space.
Overall, the aim is clear ‘ to make up for failure to show up in smartphones by leaping ahead in emerging growth areas, even before the requirements for those areas are fully understood, hence the focus on flexibility. CEO Rory Read says that, by the end of 2015, over half of AMD’s revenues will come from new growth segments. Not handsets but semi-custom chips, embedded systems and ultra-low power clients and servers.