Written by Lisa Ann Cairns, Ph.D.
Is it the device hardware constraints or the hardware constraints plus the operating and application software that are directing processor roadmaps, or neither? This chicken-egg design question is guiding OEMs' and chip makers' strategic paths. As with any broad-based strategic design changes, a resulting ripple effect impacts the entire supply chain.
Component design is still primarily influenced by user demands for mobility. It is truly phenomenal how one trend that pivots on feature specifications requiring small footprints, power, connectivity, and efficiency has redefined the semiconductor and electronics industry. The end-device switch to smartphones and tablets as the foremost devices supporting ubiquitous computing, to the detriment of the trusty PC, has changed the course of the industry. The mobility switch has directly altered supply chains, companies, bills of materials (BOMs), prominence of industries, market strategies, and, significantly, the architectural and engineering designs for components, most notably processors.
Ubiquitous computing connects markets
Intel's new CEO, Brian Krzanich, stated during the recent Intel Developers Forum (IDF) in San Francisco that Intel plans to cover the growing breadth of diversity and dynamic movements in the industry. As Krzanich argued, "Innovation and industry transformation are happening more rapidly than ever before, which play to Intel's strengths. We have the manufacturing technology leadership and architectural tools in place to push further into lower power regimes. We plan to shape and lead in all areas of computing."
The present device phase that we find ourselves in continues to be transitional. We are still moving from the recent past of desktops and the initial mobility of laptops towards a more ubiquitous computing ecosystem. This new, connected ecosystem links diverse aspects of our daily lives through our increased interactions with the mobile devices we use – work, entertainment, personal life, tasks, social networks, health and fitness, and beyond. As we integrate more facets of our lives into our computing and connected world, the movement towards true ubiquitous computing is pushed forward.
At the core of the fully-interconnected moment are the central features of the new and yet-to-be-created devices enabling this next phase of our computing ecosystem. It is in the parameters of such connectivity and ubiquity that we come to understand many of the choices in processor and device innovation – the choices made by engineers and architects – particularly when ahead of the reach of the present end-device. These next generation design choices give clues to the wider strategic visions for the industry.
Chicken or egg?
If we are designing for specific devices, what about those devices is important, and why? Similarly, to what extent does that weighting of devices open or close the scope of processor design? If we are designing for the sake of processor innovation, does that extend or narrow the scope of devices that OEMs might be able to successfully market? Given the challenges faced by the semiconductor and electronics industry over the past decade, and especially since the 2009 global recession, supporting a fruitful and diverse line of innovation is critical to the continued health of our industry. Economic conditions and conservative spending remain the global norm. To succeed in today's competitive market with tight margins and short lead times, innovative end-product wins are required. So what makes for an innovative product?
Drilling down into the component trajectories themselves, with particular attention to the latest processors and those soon to be released, we come to a chicken-and-egg question for leading-edge processors, and for the companies designing and producing them: are today's leading-edge processors being designed for devices or are leading-edge devices being designed for processors? First we need to ask a more specific question. Namely, what aspects of devices are we considering? Is it the hardware constraints of the device or the hardware constraints plus the operating and application software that are factors in processor design, or neither? The chicken-egg question centered on design is important because, as we move into more complex designs (for both processors and devices), understanding the guiding factors shaping these designs informs us, in turn, about the wider strategic paths that OEMs and chip makers are taking. With any changes in broad strategic paths, the ripple effect will touch the entire supply chain.
When we left the desktop and laptop perched atop a desk and walked out the door with a smartphone or tablet, the hardware device constraints stepped up to a level we'd not yet experienced. While laptops demanded lighter, portable form factors, the size-shrink experienced in the move to smartphones and tablets was transformative, representing a disruptive innovative event in device and component design. In this manner, form factor of the device began to play a greater role in shaping component designs, placing new demands on them for power efficiency, heat dissipation, and smaller footprints, while providing the high-quality user interfaces and experiences that are the hallmark of this device category – and our lives today.
At this point in the shift, favoring smart devices over traditional PC computing devices, questions around which component design innovations are the most important are at the fore of both chip makers' and OEMs' conferences and roadmaps. Both Intel and ARM extol their roadmaps' encompassing of diverse markets, from smartphones to data centers and from entry-level, low-priced devices to the leading-edge, high-priced ones. Similarly, their attention to the Internet of Things (IoT) as a driving innovation factor bridges the chicken-egg question, considering both form and function simultaneously – that is, bridging the form factor and the features demanded of devices.
Taking a step away from the direct tracking of processor innovation to consider the ecosystem in and for which processors are being designed lends a wider understanding of what is happening in our industry. We may begin to see how the design choices shape the highly competitive landscape we experience when we look at both the processor and the end-device markets simultaneously.
IoT is ubiquitous computing. It is both the chicken and the egg moment, meaning it is both the means for connecting and the type of connectivity simultaneously. IoT is a state of interconnected devices, to be sure, but it is also the state of a connected life in which the devices are capable of connecting to achieve the tasks that the user demands to proceed through her/his daily life. The latter moment is a far more complex interactive model, and one that more closely matches people's actual behavior patterns.
The industry's roadmap for IoT is made clearer when we consider the very choices being made for the very latest devices and processors that are being released into the market.
Core design strategies
One question in the industry press lately, in the wake of the recent release of Apple's iPhone 5S and 5C, was summed up by PCWorld: "Why did Apple go 64-bit?" The same inquiry into the arrival of new Microsoft Windows tablets was also addressed by PCWorld, "Windows 8.1 tablets with 64-bit Atom chips not coming until Q1." The answer to both of these designs goes back to the challenges facing chip makers and OEMs that we posed above as the chicken-and-egg question: software to meet chip designs or chip designs to meet software?
In these two notable situations for smartphones and tablets, the challenges posed to the operating software to take advantage of 64-bit processors are significant, but they are also critical for near- and long-term strategic market positioning. The decision to follow chip design capabilities over what the existing software can address is seen as a strategic path toward the next-generation, integrated level of devices and achieving the systems that will govern their operation. Enterprise-level applications necessitate thinking about connecting the ecosystem of Bring Your Own Device (BYOD). Bridging current devices with enterprise-level domains necessitates working at the 64-bit addressable level for security, higher performance demands, and, importantly, for achieving a unified operating system (OS) across the many devices held by users. As James Bruce, the lead mobile strategist for ARM, commented in an interview with PCWorld:
And what you're seeing is with all the guys playing in different form factors from smartphones to tablets and other form factors…what they want is one kernel, one set of development tools, and to make it very easy for developers, application providers and OS providers with minimal engineering development.
Presently, we see that Apple's latest A7 chip is a 64-bit, licensed ARM chip. Similarly, Intel's forthcoming Bay Trail Atom chip in 1Q 2014 will also be a 64-bit chip to support Microsoft Windows 8.1 for tablets; presently, the available Bay Trail Atom supports 32-bit OS. The 64-bit Atom server chips, code-named Avoton, are based on the new CPU architecture called Silvermont; these chips have started shipping, according to the PCWorld article
Shifts along the processor battlefield
IDF was the official unveiling for Intel's latest processor lineup, designed specifically for tablets and priced specifically for that competitive market. The strategy for this new lineup is to give Intel's OEM partners the design room to innovate competitive tablet and smart wireless devices that can compete in the tight-pricing with high-feature demand markets. The Bay Trail processors are designed as low-power consuming SoCs dedicated to tablet devices, especially, and to supporting Windows or Android OS environments, according to Intel's IDF presentations summarized by PCWorld excerpts.
The investment into the tablet form factor is notable for Intel, which dominates the PC market with 85.2% share of microprocessors (MPUs). As such, the majority of servers in the ecosystem rely on Intel chips, PCWorld underscored recently. However, the market sales volume has clearly been dominated by mobile devices, putting Intel at a disadvantage with those OEM designers, based especially on delivering chips to meet the demands of price, power, and performance to compete in this hot – and tight – device market.
ARM's advantage lies in Intel's weakness to date, as it provides a very different model to leading OEMs. Specifically, the collaborative opportunity to leverage ARM designs and customize with OEMs' intellectual property (IP) in order to tailor processors for OEMs' unique devices and/or ecosystems, respectively. This model has been the one chosen by Apple, as noted above, as well as other industry titans such as Samsung, Qualcomm, and Nvidia. ARM's latest Cortex-A50 series is based on the ARMv8 architecture supporting 64-bit and 32-bit execution states, according to ARM. This latest series of processors is based on ARM's big.LITTLE configuration, designed to balance the changing demands of performance and energy-efficiency depending on the tasks' demands.
On the PC, notebook, and server side, AMD continues to be the leading competitor to Intel, although, again, with significantly smaller market share. Of course, AMD's processors are also ARM-based designs. In this market, Intel and AMD must compete to provide leading-edge solutions. It is in the solutions for this very tough and possibly waning market that we should carefully monitor the strategic roadmaps and successes that point to who will deliver innovative wins for the IoT, embedded market.
The competition between AMD and Intel is long-standing, and this year we saw Intel's Haswell lineup alongside of AMD's A-series for the desktop processor market. In the processor head-to-head this summer, AMD furthered the embedded system design of combining a GPU and MPU on the same die in their next generation series, "Kaveri." This design route represents a significantly more heterogeneous computing model, as PCWorld reported from Computex in early June.
The solution path favoring SoC embedded designs continues to gain wins, as we have seen throughout the roadmaps for 2013 and into 2014. One of the challenges for these embedded designs is how well SoC architectures are able to be extended to meet multiple, unique device and OEM designs. The competing strategies are currently being tested by Intel with their OEM partners in the upcoming release of ultaportables, hybrids, and tablets, alongside ARM and AMD's models with their partner's device releases. The recent roadmaps shared by both leading chip makers, and the end-devices released into the market, will provide answers as to which strategies – if not all – might prove successful.
Embedded roadmap winning
Alongside the increased attention to dedicated processor designs for mobile, we see a clear increase in embedded designs. This roadmap makes sense, of course, as the advantages gained from integrated SoC or System in Package (SiP) architectures are essential to meet the form factors and feature demands of tablets and smartphones today.
Embedded solution designs hold promise, and are likely to be one of the more hotly-contested battlegrounds between Intel and the ARM-AMD camp. While Intel announced at IDF that it is intentionally diversifying to satisfy a wider set of market needs, AMD released its roadmap that pursues semi-customized embedded solutions. AMD's focus is to provide SoC CPUs and accelerated processing unit (APU) solutions to a diverse set of markets and industries:
- The "Hierofalcon" series, a 64-bit ARM platform with four to eight ARM Cortex-A57 cores, running up to 2.0 GHz, including 10GBase-KR Ethernet and PCI Express Gen 3 links in the SoC, is designed for high reliability, data center and storage applications, communications infrastructure, and industrial solutions.
- "Bald Eagle" is based on the x86 chip and will be available as a CPU/APU for high performance embedded applications and will feature up to four "Steamroller" CPU cores within a configurable 35W Thermal Design Power (TDP) to dissipate heat and allow engineers more design flexibility.
- "Steppe Eagle" is a G-Series APU SoC platform designed for low-power applications featuring "an enhanced 'Jaguar' CPU core architecture and AMD Graphics Core Next GPU architecture that include new features for increased CPU and GPU frequency," as well as greater design engineer customization and footprint compatibility.
- The "Adelaar" series is a new graphics core GPU for embedded applications with 2 Gbytes of GDDR5 memory and 76 Gbyte/s memory throughout, according to EETimes, and is designed specifically for 3D graphics and multi-display support for both Windows and Linux OS.
Intel's roadmap, importantly, includes a break from its traditional path – namely, producing new CPU series based on the Silvermont microarchitecture, codenamed "Bay Trail," which has a quad core processor yet possesses the power efficiency of the Atom series. Intel's series provides customizable processors and chips to meet the power and performance specifications dictated by Intel's customers, and to allow engineers to run processors at the various frequencies needed. The intent is to provide SoCs that can be targeted for the requirements and demands of customers with quick turnaround, according to an interview with Intel's Diane Bryant, as reported by PCWorld.
Intel's Haswell series, which recently shipped, consists of dual-core Core i3-4012Y processors that are targeted to meet the needs of tablet and hybrid tablet designs that require better battery life and heat dissipation for fanless systems, as summarized by PCWorld. Looking further into Intel's portable solution series, Intel also announced, during an IDF special media briefing, the release of Avoton, a 22nm Silvermont-based Atom Server SoC that supports Intel's roadmap of releasing new Atom series chips every year. But Atom is not the low end of Intel's roadmap. Intel's new low-end is the "Quark" series, also recently announced, which is roughly 10x more power-efficient and 5x smaller than the Atom series, according to an interview with Intel President Renee James. The Quark market target is the wearable and medical electronics sector; a quickly-rising sector that is significantly expanding the IoT device ecosystem.
Taken together, the Intel and the AMD and ARM lineups present compelling innovative designs that share the important focus of diversified processor series to meet the now wider, more competitive range of end-devices. The critical trend is the emphasis on embedded designs to tackle the challenges and balance the chicken-and-egg question of what drives innovation: the processor or the device.
Hopefully the battleground that hosts the continual processor wars will not only endure but will widen, as well, thereby introducing and supporting healthy competition that is essential to successful innovation and our industry's longevity. In the market today, we see a growing variety of devices: from new laptops, dubbed "Chromebooks," to hybrid tablet-laptop devices, to "phablets" which blur the line between tablets and smartphones, and these devices are accompanied by a diversified pricing structure with the push-down into more affordable, price-conscious smartphones and tablets to increase access to the latest technology and device forms.
The promise of continued healthy competition is only one piece of the puzzle for promoting innovation and growth; the path for new devices and technologies that will, in turn, support healthy growth for the semiconductor and electronics industry is necessary. This growth opportunity is at the core of the design roadmaps for the next generation of processors. These roadmaps are addressing the demands for the smaller and smaller footprints of end-devices, such as the upcoming class of wearable electronics which will certainly further our path into a truly ubiquitous computing state. A fully-realized IoT, a connected life, is going to rely on what is still at the cutting edge of processor and chip architectures: the ability to shrink form factors while expanding connectivity and performance in order to integrate our actions, tasks, and connections to realize full ubiquity.