As Intel approaches its fortieth birthday, its most publicly visionary technologist, Pat Gelsinger, took the stage to make broad predictions about the future for the company and the chip industry. Gelsinger has been fairly quiet in the past two years, but during his spell earlier in the decade as Intel’s chief technology officer, he was highly vocal, and set out a roadmap for the chip giant’s necessary move beyond computers and into wireless and ubiquitous access.
While he returned to many of his familiar themes at the briefing in San Francisco last week, some of his messages had been changed or subdued by the passing years, and comparison with speeches made around 2004 only emphasized that, so far at least, Intel has failed in many of the wireless ambitions that Gelsinger helped to set out. Perhaps they were over ambitious, but the fact remains that Intel has to be a strong contender in wireless, mobile and low power devices if it is to hold on to its position in the world, and so far its progress and prospects look patchy.
Major R&D programs are in place, and Intel is behind some of the most far reaching developments for future approaches to communication, media and computing, but in the immediate, commercial term, its hopes are bound up with the Atom processor and the x86 architecture, neither of which are dead certs for major success in the mobile world.
The first iterations of Atom are targeted at notebooks and ultra-mobile computers – the latter the successors to PDAs and laptops, and what Intel hopes will be the dominant architecture for the emerging mobile internet device (MID) market, rather than a smartphone-derived approach like Nokia’s internet tablets or Qualcomm’s MID designs. It will not be until at least 2009 that an Atom optimized for truly mobile products like handsets appears, but doubts remain over whether this will be sufficiently power efficient and compact to compete with the established cellphone chipmakers’ new developments, most of which use the main alternative to x86, the ARM architecture.
A major ray of hope was offered up in a research note by analysts at FTN Midwest Securities of Cleveland, Ohio, which reported this week that Apple would incorporate Atom in a future iPhone, and that this device would arrive in late 2009. Analyst JoAnne Feeney said the iPhone would use a 32nm Atom processor in place of its current ARM-based Infineon/Samsung silicon, and claimed the development project is ‘well ahead of schedule.
The 32nm, mobile Atom could even be demonstrated as early as next month at the Intel Developer Forum in San Francisco, where it usually showcases its latest product plans. The key will be the shift to 32nm ‘ ARM-based processors are more power efficient than those using x86 when on the same process manufacturing node, so Intel must pin its hopes not just on the power efficiency improvements it is promising for future Atoms, but on getting to the next process node ahead of competitors such as Samsung.
Apple will never be a mass volume shipper of cellphones like Nokia, but it is very important, in terms of profile, credibility and likely market share, in the high end, web-enabled smartphone sector ‘ in other words, the mobile internet devices that will evolve into the platforms that dominate computing and communications in the next decade and beyond. As such, it would be a very valuable win for Intel, which also lured Apple away from the PowerPC architecture and on to x86 for the Mac two years ago.
But elsewhere, Intel will have a lot of proving to do, to convince the big hitting phonemakers such as Nokia, and other likely heavyweights in MIDs, such as the consumer electronics vendors such as Sony. Gelsinger understands better than anyone how vital it will be to do this in order to stay at the helm as the world shifts from PCs to mobile devices and to embedding processing, internet and communication capabilities into almost every imaginable product.
This was a vision set out five years ago by the then-CTO ‘ who was somewhat sidelined in 2005 as Intel became more conservative on the aggressive wireless roadmap he had devised, though he remains a senior VP and key technologist. Gelsinger, a radio engineer by background, almost single-handedly steered the Intel juggernaut from 2003-5 on to the most ambitious and coherent wireless course in the industry. As head of R&D, he threw dollars, radical roadmaps and a large dose of vision at the notion that in the digital world, every type of device (down to sensors embedded in human beings, at some point) would need to communicate using always-on, power efficient, wireless links ‘ and that Intel could position itself to provide the chips for all those devices if it became pre-eminent in wireless technology.
Opponents of Gelsinger ‘ and his presentation of his pet project, the network-leaping adaptive silicon radio, to the management team, reportedly provoked a high degree of opposition to the CTO ‘ said he was all vision and no execution, and that he was setting expectations that Intel could not realistically meet, with the risk of damaging its reputation with customers and investors. Also, that he was too focused on a long term wireless vision and not enough on the realities of Intel’s current markets.
The same dilemmas remain today, and were clear even in the four-pronged strategy that he outlined for Intel as its enters its fifth decade. This is especially because Intel has not always succeeded with the wireless technologies it could drive itself ‘ Wi-Fi has brought it considerable influence and sales, but UltraWideBand/WiMedia and WiMAX remain unproven in both respects, and Intel will undoubtedly have to work with technologies where its rivals hold the rudder, notably LTE.
Gelsinger’s predictions revolved around four key areas ‘ ‘many-core’ microprocessors for massive processing power; the push of x86 into very low power applications like cellphones; the combination of microprocessors, sensors and embedded technology to make computing ubiquitous; and a move to 450mm wafers that could cut chip costs by 40%.
Processors with hundred of cores will drive new applications such as instant, interactive medical imaging or accurate speech recognition, said Gelsinger, and will make computing ‘immersive, intuitive and interactive’ ‘ though of course, new languages and tools will be required for programmers.
Gelsinger spent some time instilling confidence in the roadmap to get Atom and x86 into cellphones by lowering power and attracting new software developers into the mobile market; and painting a more bluesky picture of ubiquitous computing and internet, embedded even into cars and door locks. “You will be interacting with the computing experience without ever thinking of it,’ he said.
His fourth prediction centered on Moore’s Law with Intel betting heavily that the industry will move to a new generation of the silicon wafers that will support the doubling of capabilities every two years. By moving to 450mm wafers from 300mm, the average cost of producing each chip can be driven down by 40%, Gelsinger noted, and also expects Intel to gain a major headstart over many chipmakers, that will not be able or willing to invest billions of dollars in moving to 450mm.