Intel’s new CEO, Brian Krzanich, announced his first major public decisions this week, with a reorganization that put him in direct control of processor design. The changes designed to reassure customers and investors ‘ many disappointed at Intel’s cautious choice of another internal CEO ‘ that he could act quickly and, if necessary, ruthlessly, and that he understood that focus and structure, not just expensive architectures, would be necessary to start the turnaround in the mobile business.
His message is clear ‘ he, together with right hand woman Renee James, is up to the huge challenge ahead of him ‘ to shift the Intel juggernaut into new lanes of business, not just the much discussed mobile transition, but also to extend its processor expertise into foundry services, and into new product areas such as software-defined infrastructure and the internet of things. This requires creative thinking ‘ in particular, to stop defining mobile and post-PC products as smartphones, tablets and ultrabooks, but to consider what will come ‘post-smartphone’ (or at least when the value is sucked out of the smartphone form factor, a day that will come about two decades more quickly than it did for the PC).
All this at a time when the very foundation of the modern Intel, Moore’s Law ‘ the ability to scale chips down in size while increasing performance – is coming under increasing strain. Broadcom CTO Henry Samueli, speaking at a Silicon Valley event to celebrate Ethernet’s fortieth birthday, said: ‘Moore’s Law is coming to an end. In the next decade it will pretty much come to an end so we have 15 years or so. Standard CMOS silicon transistors will stop scaling around 5nm and everything will plateau. I am comfortable we will get to terabit networking speeds, but I’m not sure I see a path to petabit speeds.’
Even Intel itself acknowledges the issues, recently calling for more global research into future processes. William Holt, general manager of Intel’s technology manufacturing group, said in a recent speech: ‘Are we closer to an end than we were five years ago? Of course. But are we to the point where we can realistically predict that end, we don’t think so.’ He added: “The fact that now as we look at the future, the economics of Moore’s Law … are under considerable stress is probably appropriate because that is fundamentally what you are delivering. You are delivering a cost benefit each generation. As we make things smaller, the effort that it takes to make them actually work is increasingly difficult.’
The debate seems theoretical and long term, but it highlights the sheer size of the challenge facing Krzanich and James. The mobile industry in particular demands ever-smaller, faster chips and ever-shorter product cycles, and Intel is responding with its traditional weapons ‘ excellence in pushing the boundaries of Moore’s Law in terms of process technology. It has the most advanced manufacturing in the industry today with its new 3D transistor structure on the 22nm process ‘ innovations it is also starting to sell externally to foundry customers like Altera and Achronix.
Even an Apple deal has been mooted, and such talk, along with Krzanich’s background on the manufacturing side of Intel, suggested that expanding the foundry activity would be his greatest personal focus. Instead, in his new reorganization, he has placed James in charge of that business and is taking the lead himself in processor design, with a particular focus on mobile strategies. Again, this reflects his recognition that the mobile processor is the weakest part of Intel’s future plan, but also the most critical to get right, in order to compensate for the decline in PC sales and drive a short term rebalancing of the device chip market which is responsible for so much volume in the industry.
While foundries and SDN-enabled infrastructure represent brand new businesses where Intel has the chance to be at the cutting edge, in the post-PC device world, it has been firmly on the back foot for years, and this may well be its last real chance to take a leadership role rather than forever playing a bit part ‘ and for a company of Intel’s scale, that would be worse than useless to its revenue mix or its ability to define industry goalposts, the latter a hallmark of x86’s success.
In PCs and servers Intel has usually made the rules and forced others to follow or rail against them. In the mobile world, the opposite is true, and Intel’s recent Atom developments have been clearly within the framework set by ARM and its key partners. The new Atom platform, Silvermont, glistening with innovations such as the 22nm process, must not just seek sales, but to steal some of the initiative from ARM, especially in new and fluid device categories such as tablets, where there is still room for disruption. However, technology alone does not deliver radical change ‘ Intel needs the laser focus that Qualcomm has on its core mobile segments, and that thinking seems to lie behind Krzanich’s first big public decisions in his new role.
An internal email, revealed to Reuters, set the tone for his new leadership ‘ decisive, and focused on near term execution in parallel with future strategy, which will please those who criticize Intel for being big on vision and poor on getting actual near term results. He wrote: “As your CEO I am committed to making quick, informed decisions. I am committed to being bolder, moving faster and accepting that this means changes will be made knowing that we will listen, learn and then make adjustments in order to keep pace with a rapidly changing industry. Our business faces significant challenges, and we simply must continue to execute while finalizing our future strategy.’
The most important new move is, according to Reuters reports, to create a new unit dedicated to emerging devices beyond smartphones and tablets. This is vital. Intel needs to identify new segments where it can take the lead and force others to follow, and where it can ensure the particular qualities of its chip shine best. The smartphone market is sewn up. Intel will gain more slots here and there, but it does not have anything so dramatically better than Qualcomm that it can create real disruption here. Tablets are far less mature and more open to new architectures, particularly at the high end where vendors are looking for ways to differentiate and add value/price, and where the original iPad idea is morphing rapidly into all kinds of candidates for the post-PC device. Intel’s own contender has been the ultrabook, which translates many of the x86-oriented norms of the PC into a slim, always-on form factor which will increasingly feature touchscreens. The ultrabook shows the company’s recognition that it needs to define and lead a popular category, and it may still manage to turn its platform into a majority option, if it can get the price right.
Researchers at IHS iSuppli calculate that Intel could enable an ultrabook priced just $200 at retail, an action which would utterly disrupt the tablet and hybrid spaces and put it back in the driving seat ‘ though of course, at the price of its margins. Principal analyst Craig Stice believes Intel could bring ultrabook prices down from their current $1,000 to $200, and that it has near term plans to collapse the cost of its post-PC offering. He said: ‘The small laptops known as netbooks saw their prices reach down into the $200 range at the height of their popularity a few years ago, and a cost analysis of netbooks shows how such a low level of pricing can be used to support a no-frills type of ultrathin PC. Hitting this kind of price point is not impossible for the PC industry, already a cut-throat market accustomed to razor thin margins.’ According to IHS, Intel can control up to one-third of the total bill of materials through the CPU and motherboard and with Bay Trail, could choose to push a major performance boost in parallel with a price reduction ‘ rather than the usual strategy of charging a premium for technology leaps forward. This would see the traditional PC sector adapting to the harsh realities of the mobile world, where vendors are expected to deliver significant upgrades every year without adding too many dollars to the price.
But it is unlikely either the ultrabook or the iPad will be the last word in post-PC form factors. The emergence of all kinds of hybrids and convertibles, and the slow but steady progress of the fully cloud-based Chromebooks, all indicate that the data devices of the future will be a mixed bag, and will never coalesce around one basic design as the PC did. Indeed, the new unit is looking well beyond traditional definitions of post-PC products and examining directions such as wearable and embedded computers too.
The important thing for Intel is to be the ‘obvious choice’ for at least some of the flavors that prove popular, and so to wrest back the agenda from Qualcomm ‘ which has to adapt itself to the coming ‘post-smartphone’ era, when high end handsets become what featurephones are now, commoditized and important to industry volumes, but not a focus for innovation or leadership.
Qualcomm has had its own challenges in getting Snapdragon into devices which are less critically dependent on mobility and low power than smartphones and tablets. Its own attempt to create a post-PC category, the smartbook, failed, and it has made limited progress in set-top boxes, PCs and other targeted areas. Like Intel, it has been setting up groups to focus on new territories, but the experience of both companies shows how hard it is to break down traditional supply chains and design assumptions and move into unfamiliar markets. Hence the crucial importance that Krzanich’s new unit is not just a showpiece to please investors, but delivers new thinking and ideas ‘ indeed, everything that many analysts hoped Intel would gain by hiring a CEO from a non-traditional background.
Krzanich and James have the added burden, on top of their other challenges, of disproving that assumption that they represent a safe pair of hands, but not a new direction. There will be no honeymoon period for them, as there would be for an externally sourced CEO, so they need to get Wall Street excited from day one. New devices is clearly the best way to do that, and indeed, their thinking in this area was likely to have been the deciding factor in the board’s decision to appoint Krzanich, and to create a two-headed team ‘ James’ role as president is a first for Intel, and a sign that the firm recognizes the huge scale of the task in hand. Intel chairman and former CEO Andy Bryant told The Wall Street Journal that the new device strategy was ‘absolutely what won them the job. Brian and Ren’e delivered a strategy for Intel that is pretty dramatic.”
“The new devices organization is responsible for rapidly turning brilliant technical and business model innovations into products that shape and lead markets,” Krzanich said in the memo to staff.
The new devices unit will be headed by Mike Bell, currently co-leader of the mobile and communications business, which will now be run entirely by its other VP, Hermann Eul. Bell has experience in the mobile market, having previously worked at Apple and Palm (funnily enough, Qualcomm also just snapped up a former Apple/Palm leading light, appointing John Rubinstein to its board of directors). Bell joined Intel in mid-2010 from Palm, which in turn had poached him from Apple in 2007. He was hired, according to Intel statements at the time, to ‘help build and lead a team to build breakthrough smartphone reference designs to accelerate Intel Architecture into the market’. He has had some success, contributing to the delivery of reference platforms based on the first Atom processor (Medfield) to be suited to handsets, which then won slots with Motorola, ZTE, Lenovo and others.
His new role will be a change of product focus rather than responsibility. At the end of 2011, Intel carried out a restructuring which created the Mobile and Communications group, under Bell and Eul, which brought together the activities of four entities – Intel Mobile Communications (basebands, built around the acquisition of Infineon Wireless); Mobile Wireless Group (wireless LAN); Netbook and Tablet Group; and the Ultra Mobility Group (smartphones and other low power devices). The MCG’s remit in general, and Bell’s in particular, given his design experience at Apple and Palm was not dissimilar to that for Krzanich’s new devices group. It was charged with creating a “compelling user experience by providing optimal hardware, software, and connectivity ingredients as well as complete solutions such as form factor reference designs’.
Bell’s new unit points to the future for Intel, but it will need to be supported by continuing, indeed accelerating, enhancement of the actual chip platform and the manufacturing processes underpinning it. Those are the two areas where the new leadership double act will take direct, hands-on control. In the new structure, Krzanich will have the Intel Architecture Group reporting directly to him. Its former chief, David Perlmutter ‘ another favored CEO candidate and, under Otellini, something of a ‘face of Intel’ ‘ will now lead a ‘management transition effort’, a move that will inevitably be interpreted as a sidelining. Meanwhile, James’ main area of direct responsibility will be manufacturing operations. (To complete the reorg, CFO Stacy Smith has been given additional responsibility for Intel Capital and the company’s internal use of information technology.)
Of course, the heart of Krzanich’s architecture efforts will be the new Atom micro-architecture, Silvermont, unveiled in the same week his appointment was announced. Silvermont will carry most of the burden of delivering on the firm’s hopes for Atom and for the device market. This is the first major reworking of Atom since it made its debut five years ago. Only tweaks have been made to its original compute-core architecture, which was first unveiled in 2004 under the codename Bonnell and shipped in 2008 at 45nm. It was then updated with the Saltwell SoC at 32nm, shipping last year, but that did not involve a major redesign, unlike Silvermont, which takes many elements of the more powerful Core and Xeon approaches, such as using out-of-order execution rather than in-order (boosting horsepower).
Silvermont will be the foundation of a generation of 22nm Atom systems-on-chip to start rolling out later this year, but will also underpin offerings for the widening number of sectors where low power is critical, including the new breed of blade servers targeted by ARM-based contenders like Marvell. The first products will be the Bay Trail SoC, targeting tablets, and Avoton, for servers, which are already sampling, and Perlmutter said that future generations built on Silvermont would be churned out more rapidly, on an annual cycle. The Merrifield SoC, for high end smartphones, will follow by year end, majoring on efficient web streaming and context awareness as well as better security and privacy protection.
And the Rangeley SoC, also due late this year, will be aimed at communications infrastructure such as entry level or midrange routers, switches and security appliances. Though Atom is extending its reach as low power becomes vital to a far broader range of applications, there will also be 22nm advances for the Core and Xeon architectures, the former for high end PCs and the latter for servers, embedded processors and high end infrastructure.
Intel claims Silvermont will support threefold increase in peak performance compared to current Atom chips, or the same performance at five times less power consumption, by harnessing the 22nm, trigate process as well as design advances. These include a multicore and system fabric architecture scalable up to eight cores and new Intel Architecture instructions and technologies to enhance virtualization and security.
Intel has created an implementation of the 3D (tri-gate) transistor technology specifically designed for low power SoC use ‘ and different from the tri-gate process adopted for Core and Xeon. When coupled with power management techniques, tri-gate enables chips with an exceptionally wide dynamic range – meaning they can be turned down to very low power when performance needs are low, and turned up again when that changes. This is an alternative to the big.Little approach from ARM, which runs basic processing on low power cores and hands off heavier duty tasks to more powerful cores only when needed. ‘We do big and little in one shot,’ said Perlmutter.
Design wins for the initial Silvermont SoCs will start to filter through during the summer run-up to the end-of-year device buying season. It is fortunate that the first 22nm SoC targets tablets ‘ this could bring Intel some easy pickings in the Windows 8 sector, if Microsoft’s ‘Blue’ upgrade kickstarts the platform; and, as argued above, is more likely to yield a really bighitting victory ‘ a tier one product at a tier one OEM ‘ than the overcrowded smartphone space, where Atom has mainly succeeded in low end or second-string models (with the exception of prime partner ZTE).
What Krzanich needs to ramp up confidence in his new directions is a major league design win in a new device category. Apple would be the prize, and there have been many reports of possible deals whereby Intel manufactures the iPhone maker’s processors in return for a slot in the iPad or future hybrid. Nearer term, it might actually be Samsung which delivers the breakthrough. Two benchmarking sites (GFXBench and SamMobile) have reported that the Korean giant is testing Atom in the upcoming Galaxy Tab 3 10.1-inch tablet. Of course, Samsung has its own Exynos processors, but its mobile arm has pursued dual-supplier strategies before. Some Galaxy smartphone models use Qualcomm Snapdragon, and given the massive volumes it requires to feed its Galaxy family, it is likely to retain multiple sources of processors, even when one of its suppliers is its own stablemate firm.
If the Samsung reports are true (or even not), they highlight why tablets are so interesting to Intel because the lines are so fluid in the space between PCs and phones, and there are far more complex price/performance/power efficiency trade-offs than in either of those older categories. Firms like Samsung, which make Intel-based PCs as well as mobile and consumer electronics products, will face many dilemmas about the most efficient way to converge their supply chains in future, and could decide to unify all their large-screened products around x86 ‘ as Atom and ARM become more similar ‘ rather than sticking to the current segmentation where slates are treated as stretched handsets. Or it could all go the other way, and see the increasingly powerful ARM architecture, enriched with 64-bit and massive multicore support, storming every Intel citadel and becoming the only SoC game in town. That possibility becomes only marginally less real because Intel got a refreshingly decisive new CEO, or even because it delivered a genuine leap forward with its micro-architecture. But May 2013 will prove to be a hugely significant one in the Intel history, one where it gave its rivals something to worry about again. With ARM, Qualcomm and other power players ranged against it, with a collective R&D spend that tops even that of Intel, Krzanich’s task is a daunting one, but at least he has thrown the ball back into ARM’s court for a while.