Your browser is not supported. Please update it.

17 January 2013

ARM flexes its muscles quietly at CES

Qualcomm and Samsung were widely acclaimed as the winners at CES for upstaging their rivals, and for memorable keynotes, but it was the UK chip designer ARM that has been taking over the mobile world behind the scenes on the back of the tablet and smartphone boom. There was also a hint of things to come with the 25% increase this year in number of exhibitors of connected healthcare devices such as heart monitors, which will become more relevant for pay TV operators and broadband providers in years to come as they compete for the burgeoning smart digital home market.

The biggest product splashes were made by tablets, smartphones and ultra HD TVs, relegating PCs and laptops to the back of the stage, while 3D TV seems to have expired almost completely. 3DTV will be back one day with glasses free viewing around a room featuring multiview technology and also probably some of the display innovations on show at CES, but for now it has been recognized as a vain and premature attempt to create the next big TV wave after HD.

PCs on the other hand are a mature product still much in demand, but are in a state of flux as they morph into devices fitter for the mobile age with better connectivity, and tablet like portability including touch screen operation and detachable hinges. It was fitting that the greatest PC related interest at CES was generated by Tactus with its new touch screen embracing keyboard features designed to woo old school word smiths who still prefer laptops to tablets on the move because they can crunch out copy faster. The new Tactus technology is compatible with third party keyboard apps such as Swiftkey and Skype, which users can invoke by double tapping the screen while in messaging mode, causing physical buttons to extrude from the screen in a bubble-like format, with ability to adjust the layouts. Tactus said that consumer devices including laptop/tablet hybrids incorporating the technology could be available late 2013 or more likely into 2014.

PC makers gave the lame excuse that the timing of CES did not accord with their product release plans, although that’s not hindered them in the past. It is true that some launched products to coincide with the release of Windows 8, but if so they could just as easily have emerged with something to show for CES as for the Mobile World Congress (MWC) in Barcelona in a few weeks time. MWC matches the increased priority they are giving to mobility, which in a sense was rather hilariously highlighted by Microsoft CEO Steve Balmer’s cameo appearance in the infamous Qualcomm keynote. This may have served to remind delegates of Microsoft’s existence when it failed to exhibit for the first time since 1995, but the intention of highlighting the strength of commitment to mobility got lost in the farce of the presentation. At least that is how it appeared to cynical journalists and analysts as the keynote went viral amid comments on its cringe worthiness and inadvertent hilarity. Faultline is not the place to add to all the opprobrium fired upon Qualcomm, but it did look as if this was a flawed attempt to glamorize mobile chipsets when arch rival Nvidia was able to captivate its audience with an actual product in the shape of its new portable gaming device, code named Project Shield.

This showed off the new Tegra 4 chip also unveiled at CES by playing games and displaying video content on the Android platform, with the ability to send to an HD TV, as well as stream games to the device from PCs. This made a big impact while Qualcomm was floundering around with its incoherent keynote, when it might have been better pointing out the actual shortcomings of Project Shield. Despite its potential and technological accomplishment, it is still ultimately a dedicated gaming device when the force is increasingly behind more ubiquitous tablets and smartphones. While it will appeal to the most committed gamers, most others will opt to buy much less expensive add-on controllers that connect to an ordinary phone via Bluetooth. Smartphone users already have the Android gaming library to play with, and a number of the non-Android games cannot be accessed by the Project Shield device anyway. To an extent it seems to have sacrificed all the functionality of Android devices for modest gains in gaming performance, which suggests that its main purpose was to show off the capabilities of the ‘Tegra 4’ chip, although this objective was also fulfilled by a reference laptop built specifically for the purpose, and which does run Android.

The chip overcomes a big limitation of the preceding Tegra 3 found in various tablets including the first Microsoft Surface RT, and smartphones such as the Google Nexus made by Samsung, by adding support for 4G LTE connectivity. Without this it would soon have been ditched by tablet and smartphone makers, and even now it may struggle to hold on to some key accounts against a renewed onslaught from Qualcomm. The Tegra does boast six times greater headline performance than its predecessor, and is one of the first quad-core chips to adopt ARM’s new Cortex A15 architecture. It does support higher resolutions up to ultra HD, but even so there are some who doubt Nvidia’s claim that it is the fastest mobile processor in the world, trumping Apple’s A6X processor used in the fourth-generation iPad, will stand up to benchmark tests.

Qualcomm meanwhile launched two ARM based chips at CES, the Snapdragon S 600 clocking at 1.9GHz and nominally 40% more powerful than the preceding S4 Pro. The new top end Snapdragon 800, operating 75% faster and with somewhat greater energy efficiency, is a quad-core processor running at up to 2.3 GHz. Both chips incorporate Adreno Graphical Processing Units (GPU), a 320 GPU for the S 600, and 330 GPU with double the graphics performance for the 800.

There was buzz around the halls at CES over whether Qualcomm has done enough with these new Snapdragon chips to replace Nvidia for some of the next Microsoft tablets coming this year, or whether by the same token Nvidia has not made sufficient progress with Tegra to hang on in there. The consensus seems to be that Qualcomm will topple Nvidia as Microsoft’s chip maker of choice alongside Intel and perhaps AMD, at least for the forthcoming follow-up to the Surface RT, and possibly also a new device called the Surface Book. However these rumors may have all fanned out from an original leak reported by a Twitter tipster who has admittedly proved accurate in the past, suggesting that Microsoft will replace the Nvidia Tegra CPU in the existing Surface RT with one of the new Qualcomm Snapdragon chips in its sequel, the 8.6 inch Surface RT2 coming out later 2013.

The same rumor suggested Microsoft will replace the current Intel Core i5 CPU for the upcoming AMD Temash APU in the second Surface Pro, which will have an enlarged screen of 11.6 inches rather than 10.6 inches, and that there will be a new device with an even bigger 14.6 inch screen called the Surface Book, which will have an Intel chip, the 22nm Haswell chip, the successor to the current Ivy Bridge range of CPUs.

Microsoft was certainly not pleased with the reception to the Nvidia Tegra 3 chips in the first Surface Pro, which was slated for mediocre performance in the desktop mode and modest battery life, as well as lack of support for LTE. At least Nvidia has rectified the last of those failings with the Tegra 4, but it is not clear yet whether the battery problem has been fixed, which would almost certainly involve some cooperation with Microsoft over support for the so called phantom core for low power tasks.

This phantom core is a fifth ARM Cortex-A9 processor core that implements what Nvidia calls ‘variable symmetric multiprocessing’ (vSMP) designed for extremely low-power operation during periods when the end device is in active standby mode performing background tasks such as email and social media synchronization. This is based on the premise that users do not care how fast these background tasks are executed, and would much rather save on battery life. This is undoubtedly the case when these tasks are taking place in the background, for users are only sensitive to the response time of the application running in the foreground. But whether such innovations will stem defection to Qualcomm remains to be seen.

Meanwhile Microsoft was actually to be found somewhere else as well at CES even if it was again nothing to do with PCs. This time Microsoft high jacked Samsung’s keynote to unveil its IllumiRoom with a vague promise ‘to blur the lines between on-screen content and the environment we live in, allowing us to combine our virtual and physical worlds.’ The big idea is to extend the TV display beyond the physical dimensions of the screen, and make it blend more harmoniously into the furniture. To do this, Microsoft has combined Kinect, its motion sensing input device for the Xbox games console, with a projector to display suitable images around the TV, such as games or even ‘video wallpaper’. As Microsoft put it, ‘Kinect captures the geometry of a room and adapts projected visuals in real-time without any need to custom pre-process the graphics.’

IllumiRoom itself seems rather woolly and unconvincing, but it did play well into the debate circulating at CES over the likely fate of ultra HD TV, which was certainly the most eye catching and talked about aspect of the show, next perhaps to Qualcomm’s keynote. Samsung came across as the winner here in debuting its 85-inch Ultra HD TV set and taking its first pre-orders in South Korea, at the lofty price tag of $38,000. Samsung even displayed larger 95 inch and 105 inch models that will obviously carry even higher price tags when they are launched later in 2013.

Sony and LG might be feeling a little miffed at being upstaged by their larger and increasingly dominant competitor, having come out earlier with Ultra HD TVs, which stole the show at CES 2012. These were smaller at 64 inch, and cheaper, with LG’s at $20,000 and Sony’s at $25,000.

One problem for ultra HD could readily be identified by asking around at CES among the various technophiles, and even those who could afford an ultra HD TV set said they were unlikely to be buying one in the foreseeable future. This was largely because they had mostly purchased various big screen HD TVs over the last few years and could detect little discernible improvement in quality with ultra HD, at least compared with 1080p HD. This is because ultra HD only really begins to score for screens bigger than 60 inches viewed from distances of 10 feet or less. That is the perception of most people even if a few technology victims claim that once you have watched something in ultra HD you will never want to go back to 1080p, let alone 720p or 1080i. Unless they have super human vision that is patently untrue for displays much under 60 inches, at any rate when compared with 1080p.

Several other issues are conflated here, one being that ultra HD will only become viable once the whole ecosystem has caught up, and that is clearly several years away given that no major broadcast services are yet even in 1080p. It will only be when content is shot in ultra HD that the full benefits will become manifest on the screen. For example the current generation of game consoles cannot handle resolutions higher than 1080p, and upscaled games do not look that good on an ultra HD set at four times the resolution of 2160p.

There is also, to put it mildly, a dearth of Ultra HD content to watch, with no prospect of that changing anytime soon. At present there are just one or two Ultra HD players, with a 4K download service due for launch soon by Sony. But for now consumers who do opt for Ultra HD and buy something like a Redray player or subscribe to Sony’s service are going to be faced with an even greater constraint, lack of bandwidth. Ultra HD files, even with HEVC encoding coming along with twice the compression efficiency of H.264/MPEG4, ultra HD will require at least 10 Mbps to stream and for now more like 20 Mbps, and there is little prospect of that being widely available as a real sustained data rate per home over broadband networks for several years at least. That means that in practice ultra HD will be confined to download services for the foreseeable future. Because of bandwidth constraints cable, satellite and IPTV pay TV operators generally have to be selective over which channels they offer even in 720p or 1080i HD, generating one eighth as much data as 2160p ultra HD.

The other ultra HD issue was the one invoked by Microsoft’s IllumiRoom – cost of manufacture. It is a well-known law of TV that the cost of making a screen increases at a greater rate than the surface area, so that a 60 inch TV will cost more than four times as much as a 30 inch model, even though it has precisely four times the screen area. The actual cost ratio is more like six to one. Therefore a 105 inch model such as Samsung displayed at CES will always cost around 10 or 12 times more than a 40 inch HD TV in the average living room today. This is because the batch reject rate in manufacturing is itself directly proportional to the surface area ‘ double the screen size and twice as many TVs have some fault that means they have to be recycled, as a rule of thumb. The final cost of the TV reflects this rejection rate.

Microsoft’s view of the TV complemented by projected images may be flawed, but the related idea of having walls comprising multi-paneled TVs could well take off, because they could deliver ultra HD to a large area without incurring the high ‘batch reject cost’ of big single screens. This is the idea that is being carried by NDS, now part of Cisco, with its ‘Surfaces’ project first unveiled at IBC 2011, comprising wall sized displays made up of multiple smaller panels perhaps only around 24 inches each. This enables large 105 inch size displays to be made effectively for about 15% of the cost of the forthcoming Samsung ultra HD of that dimension, and also confers flexibility, since content such as a talking head that does not justify a huge image size can be displayed on a subset of the overall screen area more satisfactorily from a visual perspective than on a single big TV. The panels outside the display simply fade into the background wall. While no manufacturer has yet committed to making large multi-panel TVs in production volumes, some very large displays have been demonstrated, and surely the manufacturing economics will prove compelling.

The other aspect of ultra HD that did surface at CES concerned the underlying display technology, which is in principle independent of the resolution. As it happens the ultra HD TVs on show at CES used conventional LCD technology, and at the large screen sizes required to justify the resolution some of its limitations are exposed. This led to mutterings about limited viewing angle, lack of sufficient contrast, and issues with uniformity of the backlighting that can be disconcerting when the screen becomes darker as in a night time scene. All of these deficiencies are more apparent on large screens at the high resolutions that impose greater all round expectations of quality, and may be why some people around the show said they preferred TVs based on the new OLED technology at say 1080p when compared with an ultra HD TV of the same size.

OLED is based on an emissive electroluminescent layer comprising an organic compound film emitting light in response to an electric current and it has greater manufacturing batch reject problems than LCD, which is why the displays are currently so expensive. With normal size screens costing $12,000, OLED displays face the same upgrade fatigue that will afflict ultra HD. But it is generally agreed that OLED is higher quality than LCD, with sharper contrast and greater consistency across the field of vision, highlighting the fact that quality of experience is not purely about resolution. There is no reason why OLED cannot scale to ultra HD anyway, and if the batch reject issue can be resolved, we may well inhabit a world of multi panel ultra HD OLED by the end of the decade.

While that is speculative, there is little doubt that one of the more unsung heroes of CES 2013 will go from strength to strength. This is Cambridge based ARM, which began in the 1980s by developing an architecture for 32 bit RISC (Reduced Instruction Set Computing), designed then to cut the cost and increase performance of high performance graphics workstations by eliminating support for redundant microcode. RISC enabled ARM to reduce the number of transistors in a chip to around 30,000, less than half that of comparable processors at the time from the likes of Motorola. However it was the advent of battery powered mobile devices that really elevated ARM, since then the power saving enabled by RISC made all the difference given the disappointing rate of progress in electrical storage density.

ARM has now become the dominant architecture for mobile devices, shipping 7.9 billion processors in 2011, and 2.2 billion in the third quarter of 2012. This represents 95% of the market for smartphones, and also 90% for anything with a hard disk drive. This has not translated into the multibillion dollar revenues enjoyed by other dominant high tech companies, because ARM merely owns intellectual property and does not manufacture the chips. Its revenues for Q3 2012 were $227.9 million, up from $192.3 million a year earlier. But it does put ARM in a highly stable position because of its dominance, and also well placed to exploit accelerating growth across the whole IP connected device space, or the Internet of Things. This brings the prospect of faster growth, and while rivals, notably Intel, are knocking at the door now, the established ARM ecosystem should protect against significant erosion of its position for the foreseeable future. At the same time ARM is largely immune to competition among System on Chip (SoC) vendors because they virtually all use its architecture. Whether or not Qualcomm beats up Nvidia does not really matter to ARM, which in a way continues to be the architect of the mobile revolution.