Mobile Processors: Here, There, and Everywhere

sagawa

Follow SecSovTMT on Twitter

 

The three decade run of the PC ensconced Intel at the top of the semiconductor food chain, a position that appears ever more precarious as mobile platforms, largely based on ARM processor designs, proliferate.  In response, Intel will introduce parts exploiting a proprietary manufacturing process advantage to deliver superior performance-to-power-use in an industry best silicon footprint.  At the same time, ARM has announced a new processor concept that would allow its licensees to mix and match high performance and low power use cores within a single chip to achieve a similar effect.  Weighing Intel’s process advantage vs. ARM’s superior design, the combined resources of its licensees, and the benefits of mobile platform incumbency, we believe that the erosion of Intel’s hegemony will continue

Moore’s Law famously dictated that the density of semiconductors would double about every two years.  While the industry has periodically hit technical roadblocks to further process improvements, these have proved to be short lived, a basis for on-going improvement in the performance of semiconductor solutions.  At the same time, chip designers continue to eke more performance from each process generation through more efficient circuit design, and faster clock speeds. The performance and power use demands of the burgeoning mobile device market is impelling extraordinary rivalry amongst chip makers that is driving system performance forward and separating winners from losers

To date, Intel has made little progress with mobile processors, despite substantial investments meant to lever into smartphones and tablets.  The company means to change this record of futility with new processors that exploit Intel’s manufacturing process leadership to deliver superior performance and power usage.  The new 22nm tri-gate process is expected to deliver up to 37% better performance with the same power use OR 50% lower power use at the same performance vs. Intel’s previous 32nm scalar process.  Intel appears to have a 3 year advantage, as its move to 22nm uses a proprietary 3D technique to solve interference problems that are endemic to resolutions better than 28nm.  Moreover, Intel has engaged Google to optimize its Android platform for the new processors, reducing a substantial advantage currently held by Intel’s rivals

ARM holdings has built its dominant position in mobile device processors by licensing its core designs to chip makers who incorporate them into their own System-on-a-Chip (SoC) products.  The standardized cores allow device makers software portability across silicon vendors, while also allowing those vendors to differentiate their products.  The ARM ecosystem has evolved to a vibrant market with several capable licensees – e.g. Qualcomm, Broadcom, NVIDIA, Texas Instrument, Samsung, and others – challenging for leadership with each rapid generation.  ARM earns a royalty for each core in each chip sold by its licensees.  With the newest designs utilizing multiple cores per processor, ARM’s opportunity portends to grow faster than the overall device market, evidenced by the 25% royalty growth experienced in 3Q11

This vitality benefits device makers in pushing down costs but also driving innovation.  To wit, both NVIDIA and Qualcomm have introduced multiple core designs that can shift to low power when high performance operations are not required, thus saving overall system power use.  This concept will be inherent in the ARM architecture with the next generation designs, as the newly introduced big.LITTLE core designs will allow licensees to mix and match high performance/high power cores with low power/low performance cores to achieve up to 70% power savings. ARM’s mastery of low power usage design and the combined engineering capabilities of its licensees are formidable weapons in competing for future platforms

While Google will optimize Android to run on Intel’s architecture as well as on ARM, most of its licensees have written their own alterations that may involve calls directly to the processor.  It is likely that each device maker would have to make additional investments to assure that their Android implementations ran on Intel’s platform.  This is an obvious incentive to remain within the ARM ecosystem and a challenge for Intel

Within the ARM community, competition is fierce.  Because of the intercompatibility of ARM-based solutions, ODMs are able to shift processor vendors from generation to generation.  This allows for significant and rapid share shifts, favoring chip makers with innovative solutions, short design cycles and frequent incremental updates. This mirrors the short product lives inherent to the mobile device market, and contrasts with the history of the semiconductor market, where intense competition for product design-ins were often rewarded by years of component demand.  While all chip makers have coped with this accelerating churn, those that came of age in the mobile market may have better adapted to it

We see ARM and Qualcomm as the biggest beneficiaries of these market conditions, with Samsung, Nvidia, Texas Instrument, Marvell and Broadcom also positively positioned.  On the flip side, we are pessimistic for Intel’s ability to build and sustain momentum in the mobile processor market and have concerns that ARM-based processors could begin to undermine the x86 architecture stronghold in servers as power efficiency becomes an ever more important factor in that market.  AMD, which is also dependent on the x86 architecture, is also threatened by these developments. We are also concerned that Marvell may suffer once products from competitors based on ARM A7 come to market

Print Friendly, PDF & Email