When a company’s greatest differentiator is the tight integration of many wireless elements, it has the challenge of keeping ahead of the game in every one of those. Qualcomm is steadily adding to the percentage of a mobile device system it creates inhouse, so that it can increase its integration, its control and its margin. In addition to its traditional strength in apps processors and modems, the past year has seen significant advances in the RF front end (particularly its multiband RF360); the DSP (with an intensified R&D effort behind Hexagon); the graphics elements of Snapdragon; and of course, since the important acquisition of Atheros, in connectivity (Wi-Fi, Bluetooth, powerline and so on). In addition, Qualcomm has diversified the forms in which it mixes and matches these elements, with a stepped-up impetus behind reference platforms for emerging markets; and it is steadily pushing the Snapdragon processor and full SoC (system-on-chip) beyond the smartphone, into home media equipment, set-top boxes and large-screened ‘post-PC’ gadgets. Add to that the major program to expand into the internet of things with ultra-low power implementations of the Gobi modem, and Qualcomm has an awful lot on its plate.
Despite its huge R&D resources ‘ further justified by the licensing business for many of the inventions it patents ‘ any effort this wide-ranging must have gaps, and the game for smaller players is increasingly to spot them, creating a Qualcomm-free niche for themselves, or even attracting the acquisition attentions of the giant or its key rivals. That worked for 4G software-driven modem specialist Icera, for instance, which was ahead of Qualcomm in some respects, a fact which netted it an acquisition by Nvidia. Whether Nvidia successfully integrates the Icera modem and its own processors and capitalizes on the new all-in-one platform sufficiently to worry Qualcomm remains to be seen, but Icera has certainly found a highly strategic position for its inventive technology.
However, that example points to the area where Qualcomm is so hard to beat. Many mobile chip companies are buying up the elements they lack, particularly modem technology (Intel/Infineon Wireless/Fujitsu Wireless/Comsys Mobile; Broadcom/Beceem/Renesas Mobile), but another couple of years ensues while they converge their technologies into an SoC with the price:power:performance ratios to challenge the market leader in sectors where all-in-one solutions are the norm ‘ and those are expanding too, as even high end devices like tablets come under price pressure and replace separate, high performance parts with SoCs. Sandeep Pandya, senior director of product management, summed up the company’s most critical positioning, boasting: ‘It’s not just having a CPU, it’s not just having bits and pieces. When you take a look at what’s involved in our chipset solutions, we have a leading edge CPU, leading edge graphics, leading edge DSP, of course leading edge modem, leading edge connectivity.’
So as Qualcomm set out its stall for 2014 at its analyst day last week, where were the gaps that a more specialized competitor could still target? With the big two handset makers also increasing their inhouse developments of key components, is the market becoming impossibly squeezed for the challengers? Qualcomm has its nerves mainly trained on the firms that could replicate its all-in-one approach at low cost in the future ‘ emerging Chinese players, MediaTek, probably Intel.
But there are also some key areas where there is still space for vendors to provide an optimized discrete solution. Competitors would argue that the processor itself is one of these, and they would maintain that Qualcomm is not a leader in this area. Apple and Intel have 64-bit mobile processors; Nvidia beat Qualcomm to quad-core and now MediaTek and Samsung have octacore.
Qualcomm argues that the debate is focused on the wrong things. It’s not the number of cores, it’s how cleverly you use them that affects the user experience, the firm maintains. It rolled out that mantra when Nvidia made a huge splash with its first quad-core offerings, and is repeating itself amid the current octacore hype. SVP Anand Chandrasekher put it most graphically at a summit in Taiwan this summer when he said Qualcomm was not interested in ‘core wars’, saying: “You can’t take eight lawnmower engines, put them together and now claim you have an eight-cylinder Ferrari. It just doesn’t make sense.” He went further, accusing MediaTek of “throwing spaghetti against the wall and seeing what sticks’ rather than looking at what consumers really want. “That’s a dumb way to do it, and I think our engineers aren’t dumb,’ he said.
Of course, Qualcomm gets there eventually. It did move to quad-core, and of course claimed its timing was correct, and customer-driven, not late. It has a fair point that multicore architectures need to be engineered and programmed cleverly to deliver real benefits to users, a point Intel has echoed for years. But primarily, the noise around eight-core processors is about finding a chink in Qualcomm’s armor. If companies can target high value devices that do need massive horsepower, even if they are a niche (very high end tablets are the obvious example), they can form bonds of trust with customers which may then take their mass market offerings seriously too. They also score profile points in the market and help to reassure it that Qualcomm is not the only game in town, or always the most advanced.
Qualcomm’s real defence here is not that there is no need for more cores yet ‘ that really depends on the target device, and how the cores are implemented ‘ but that the real key in mobile devices is integration. For a growing majority of products, even quite high end models, an integrated SoC will be the norm, for cost, size and power consumption reasons. In other words, it is not enough to have a fancy CPU or GPU, a vendor also needs top class modem technology. This may come from a third party, but will still need to be coupled very tightly to maximize the benefits.
One of Qualcomm’s chief challenges is the increasing tendency for its handset customers to want to control their own chips, at least in flagship models. This is seen in the top two smartphone makers, but HTC has been reported to be considering its own chip programs, while LG is going the same way. The second Korean OEM is rumored to be readying a 2.2GHz mobile processor, codenamed Odin and using ARM’s Cortex-A15 and Cortex-A7 cores in Big:Little power saving formation. This would come in quad-core and octacore versions, the latter targeted at top end smartphones and tablets, the former at smart TVs and handsets. The flagship LG G2 smartphone and G Pad currently use Snapdragon.
In most cases, though, this applies to processors rather than modems ‘ even in the iDevices, for instance, it provides the modem even while Apple uses its own A7 CPU. Samsung still uses modems (and some processors and SoCs) from outside its own group of companies. Modem technology is a far harder area in which to innovate, in the view of Qualcomm and other players, than processors ‘ which is why MediaTek is concerning to the US giant, not so much for its octacore adventure, but for its SoC capabilities, as these start to be targeted at higher end segments.
In modems, Qualcomm has its real power, and a better claim than in processors to be at the cutting edge. Thus it is keen to play up this aspect of its portfolio. ‘You ignore the modem at your own peril,’ said Peter Carson, senior director of marketing, at the US event. The performance of the modem is harder to quantify than that of the app processor, he said, but has arguably a greater effect on user experience, because it governs connectivity rates and quality, and has a big impact on battery life. ‘The modem is the foundation of the mobile experience,’ was a strong theme among the executives on show, and this is an area where there are high bars to new entrants. ‘Processor companies coming from the PC space either haven’t built their own 4G modems or demonstrated any modem leadership, so they tend to take the conversation to the place that they know best, which is the CPU,’ said Carson. ‘Unfortunately, that’s irrelevant, lost in the noise of the equation of what constitutes a really good mobile experience.’
At the analyst day, Qualcomm introduced its latest offering, the Gobi 9×35 LTE/3G chipset, and the WTR3925 RF transceiver chip. Both are the fourth generation of multimode LTE technology and support features of LTE-Advanced. The 9×35 is the first cellular modem designed on the 20nm process node, says its supplier, and supports carrier aggregation up to 40MHz (2x20MHz channels) on uplink and downlink as well as Category 4 150Mbps speeds. With the WTR3925, Qualcomm says it is the only company on its fourth generation of LTE RF transceiver chips, boasting the 28nm node and a single-chip carrier aggregation RF solution. These chips can also pair with the RF360 front end, which spans a large number of LTE’s fragmented bands.
However, it is also looking ahead to operators’ next LTE moves and the new Gobi can support configurations up to Category 6 (part of the upcoming 3GPP Release 10 LTE-A specs) when these are deployed in future. That would enable the 300Mbps peak download speeds ‘ already being discussed by leaders like SK Telecom for 2015 ‘ over two or four MIMO DL antennas and is theoretically future-proofed to support combinations of CA and MIMO that carriers could use to achieve 600Mbps rates in 2016 or beyond. More pragmatically, the new Snapdragon and Gobi combinations will appear in commercial handsets from the first half of next year.
These are genuine advances which will set a high bar for MediaTek, Broadcom, Intel and the others, but they remain firmly focused on the huge multimode market. But as operators become interested in LTE-only devices ‘ for new services such as M2M, or as they switch off 2G/3G ‘ this is also a potentially rich seam for the 4G modem specialists to mine, as seen in the latest moves by Altair and Sequans. Verizon has been the foremost driver of the LTE-only market, both with its machine-to-machine (M2M) plans and certification programs, and its aggressive timeline for turning off some CDMA services. For this reason, the US carrier decided to forego any 3G fallback options for voice, and to build out sufficient 4G coverage by the second half of 2014 that VoLTE would be sufficient, and it could start migrating legacy networks to LTE.
Other operators are starting to discuss similar plans, but it is Verizon which has driven the development of an LTE-only device ecosystem, which is now providing opportunities for the specialists in this area. Georges Karam, CEO of one of these companies, Sequans, commented: ‘A year ago the market was about 20% on board with LTE-only, but now almost everyone agrees. That is mainly because Verizon has educated the market.’
The giants are focused on the far larger spoils of multimode, and though they will undoubtedly leap in, the smaller firms have a window in which they can forge alliances with device makers and even with other chipmakers ‘ processor vendors in search of a leading edge modem alliance perhaps. Importantly, their laser-like focus means they are starting to pile up operator certifications, and achieve a technology headstart in a field which, by Qualcomm’s admission, is a tough one to enter. Karam echoed the giant’s sentiments, commenting: ‘Yes, processors move to 28nm and more cores, but they’re kind of stuck ‘ there’s far more innovation in modems with [3GPP] Releases 8 and 9, Category 1 for M2M, peer-to-peer, and then there will be Release 10. It’s not stopping. There’s lots of work to be done but also chances to differentiate. By the time the big guys get to single-mode we’ll be on the third generation.’
The main areas of interest for LTE-only are tablets and other data-dedicated devices which do not require voice or universal roaming ‘ factors which will ensure the majority of handsets, except for a niche category of phones for prepaid, localized use, will remain multimode for many years to come. But as soon as there is significant 4G availability, with Wi-Fi as back-up, the need for 2G/3G in tablets, notebooks and cloudbooks decreases rapidly, and starts to represent an unnecessary extra cost to the OEM, with no benefit to the consumer.
The other product segment where LTE-only will be important right from the start is 4G-based M2M or internet of things (IoT). Though GSM will also remain significant for M2M, since many applications require only low data rates, a rising tide of IoT functions will start to use broadband-class data, and so all kinds of utilities and enterprises will be wanting a mobile network that will last them for a decade or more to come, driving interest in embedded LTE modules.
Sequans this month announced three new customers for its EZLinkLTE modules, which will go into tablets and mobile PCs for the US market. Although the buyers are unnamed as yet, the catalyst for such deals ‘ and several more waiting to sign, says the firm ‘ has been the Verizon certification. The EZLinkLTE modules combine baseband, RF and key interfaces in a single package to reduce cost and time to market for vendors. Two were recently certified by the US carrier for its Band 4 and Band 13 spectrum (700MHz and AWS) ‘ the VZ20Q surface mountable module for CE and M2M devices, and the VZ20M M.2 interface module for tablets and PCs (there is a third model, the VZ21M, which includes GNSS for location services).
The French firm says these modules will contribute to its revenues primarily in the second half of 2014 and beyond, the period in which it sees the LTE-only market starting to gain significant mass. It will be competing with two main rivals, Israel’s Altair and US/Korean GCT, all three companies having started in the WiMAX platform and made a sometimes painful transition to LTE. Karam admits the other two got into the LTE race earlier, but believes the EZLinkLTE products, and wider MontBlanc platform, have leapfrogged the competition in time for the market to open up. In particular, Sequans is trying to offer differentiating features, including eMBMS for mobile broadcast ‘ an example of a technology which enables carriers to deliver value added services that are unique to LTE, and where 3G fallback is not an issue. It has also put significant effort into key LTE-Advanced features such as carrier aggregation ‘ Verizon will soon be aggregating its two LTE bands, as are Korean, Japanese and some European players. Its Cassiopeia LTE-A product sports 150Mbps Category 4 capabilities, carrier aggregation up to 40MHz, eICIC and eMBMS, optimized VoLTE and enhanced MIMO.
Altair too has been significantly boosted by Verizon’s enthusiasm for creating an LTE-only ecosystem. It recently announced that its FourGee-3100/6202 chipset would power a 4G-only Ellipsis 7 tablet, the first device that Verizon will market under its own brand. The Ellipsis 7 highlights an important reason for carriers’ interest in 4G-only data devices ‘ the cost savings. The tablet will retail at just $250, the kind of price tag that can encourage consumers to adopt a cellular, rather than WiFi-only, device, and therefore form a second relationship with the cellco, in addition to the handset contract. In addition, there are savings on patent fees, and the carrier can encourage heavy data users to move off 3G ‘ where the cost of data delivery can be two times or more that on efficient LTE.
Altair believes that, by avoiding a Qualcomm-style multimode chip, an OEM or ODM can save $30 to $40, which is significant as tablet prices spiral downwards. ‘We are proud to have been selected by Verizon Wireless to power its first consumer LTE-only device,’ said Eran Eshed, VP of marketing and business development at the chip provider. ‘This product is a strong endorsement of our strategy to disrupt the market and displace costly 3G+4G multimode with lower cost 4G LTE-only internet connectivity solutions.’
His company may be competing head-on with Sequans, but Karam echoes his sentiments, and agrees that the most fruitful initial territories will be the US, Japan and Korea, where the impetus to turn off 3G at an early stage is highest. He also sees opportunities with greenfield operators which are starting to build out LTE in emerging economies, usually for fixed or fixed/portable services in place of wireline. Indeed, LTE as a fixed line technology is also being adopted by some carriers in developed economies, including AT&T, and this could fuel demand for LTE-only CPE or personal router products. The big prize would always be China, but as usual, there are question marks over the strategy there. In theory, China Mobile would like to move rapidly to TD-LTE and leave its unsatisfactory TD-SCDMA 3G platform behind, but there will be political pressures to keep a 3G/4G system going for some years. However, Sequans ‘ which worked with Mobile in its early TD-LTE trials ‘ is confident the world’s largest cellco will launch some single-mode devices from the second half of next year, which would kickstart the ecosystem in unpaired spectrum. The same timeframe is probably realistic for AT&T, which has trailed Verizon by about a year in terms of LTE coverage, but is starting to engage in discussions with the single-mode sector. Even in Europe, where operators have the biggest motivation to prolong the life of HSPA, some cellcos are starting to achieve the kind of coverage which would justify LTE-only data devices ‘ Karam points to Bouygues, the third mobile operator in his home country, as having achieved 60% LTE coverage already. For many other cellcos, it will be a wait of just 1-2 years before having sufficient 4G availability to make single-mode devices viable, as well as 4G-only services like eMBMS, which rely on having sufficient LTE coverage.
These examples highlight that no company is invulnerable, and no company can do everything in the complex mobile world. There are real opportunities to bypass Qualcomm, or even leapfrog its technologies, for companies which are very intensely focused on the right targets ‘ preferably ones which, while large enough for most players, are slightly too small to be at the center of the San Diego giant’s radar. Very high performance processors, third and fourth generation LTE-only modems ‘ these are two areas at the heart of the mobile space, where there is still room for new challengers, whether specialist technologists, or entrants from other segments such as PCs.
In the next issue of Wireless Watch, we will analyze other key outcomes of the Qualcomm analyst day, including its Internet Chip plans, and ambitions for LTE in unlicensed spectrum
Qualcomm’s latest Snapdragon:
The latest Snapdragon processor is the 805, unveiled at the event, and sporting a quad-core Krait 450 CPU combined with the latest Adreno GPU, the 420, which supports Ultra HD 4G resolution for mobile devices and TVs. Each of the cores can run at up to 2.5GHz while the new Adreno promises a 40% graphics processing hike, compared to its predecessor, in a package heavily geared to gaming, TV and other graphics/video applications. Qualcomm said the 805 is the first mobile processor to offer system-level Ultra HD support, 4K video capture and playback and enhanced dual camera Image Signal Processors (ISPs). It integrates with the firm’s Gobi MDM9x25 or Gobi MDM9x35 modems. The former supports LTE-Advanced features including some types of carrier aggregation and 150Mbps Category 4 data rates. The new Snapdragon can also integrate with the two-stream, dual-band VIVE 802.11ac chip, which allows for power efficient Wi-Fi video streaming, according to the suppliers. “Using a smartphone or tablet powered by Snapdragon 805 processor is like having an UltraHD home theater in your pocket, with 4K video, imaging and graphics, all built for mobile,” said Murthy Renduchintala, EVP and co-president of Qualcomm’s QCT unit. ‘Coupled with our industry leading Gobi LTE modems and RF transceivers, streaming and watching content at 4K resolution will finally be possible.”
MediaTek chances its ARM at eight-way smartphone CPUs:
There is always a balancing act when adding more and more processing cores into a design, between the amount of processing power you unleash, and in handsets the amount of battery you may use up in doing it. In non-handset designs there was always a concern that an extra CPU gave you far less than an additional 100% of power, given that some software has to manage the sharing of work between the cpus.
When we were really young and this process was going on for the first time in large servers back circa 1985, it was a land of pioneering in multi-processor architecture ‘ now in the smart phone, it is all about the second time around and offering more granular gobbets of power. There is now established theory in how much overhead additional CPUs take up, and so most companies should reach the same conclusion about just how many CPIs are right to include in a handset design. But breaking away from the great many players offering Quad cores this week is MediaTek, offering an 8 CPU ARM Coretex A7 core design for the first time.
ARM’s great contribution has been to mix and match cores in a single design to support its big.LITTLE concept and last month MediaTek became the first company to license and use this, coming out with a combination of A50 cores and the ARM Mali GPUs. This is all about combining the A57 and A53 ARM designs alongside one another, so that you can either put one small processor on a job or one large one or more than one, so give a granular approach to both power consumption and processing power.
It has followed this up with a GPU architecture that harnesses up to 16 Mali GPU cores at once. But these multi-processor architectures are not limited to handsets or devices which are powered with a battery. Instead this is ARM opening up to set tops and servers, and making sure its architecture scales.
But MediaTek with its new MT6592, is launching an octa-core mobile platform, offering an SoC for top end gaming performance, advanced multi-tasking and enhanced (HTML 5) web browsing aimed at high-end smartphones and tablets
Each of these 8 cores can operate up to 2 MHz clock speed. It is expected to be available in devices running Android Jelly Bean by the end of 2013 and the following generation Android Kit-Kat in early 2014. The chip is cut in 28nm and MediaTek claims it will eat no more power than existing SoCs, because it elects to use an optimal amount of CPU power for any given job. It also comes with 4 GPU cores and an advanced video playback system supporting Ultra-HD 4Kx2K currently using the H.264 codec, with support coming for H.265 and VP9. IT has a 16-megapixel camera built in.
The SoC also features MediaTek ClearMotion technology for automatic frame-rate conversion of standard 24/30fps video to higher quality 60fps video to make videos smoother on devices built with this chip and it has dual-band 801.11n WiFi, and Miracast screen-sharing as well as Bluetooth, GPS and an FM tuner.
This is a serious statement by the Chinese firm that has been struggling to mimic the chip designs of the Qualcomm Snapdragon range at the top end, and clearly feels it has made inroads with this part.
MediaTek is claiming it has its own heterogeneous computing engine and that’s why it feels it can build this octo-core platform ahead of others, such as Qualcomm and Nvidia, adding CPU, GPU, DSP, connectivity and multimedia engines as well as camera drivers, at will. We can expect a similar architectural approach up and down the ARM Cortex range, which include processors more than twice as fast as the A7.