Intel laid out its plans to aggressively use its new 22nm silicon process to dramatically lower processor voltages while actually improving performance over the next 30 months, the company told financial analysts at a conference on May 17 at headquarters in Santa Clara, California. What analysts heard was by far the clearest picture for a generation of silicon that I have ever heard the company articulate. The plan is especially attractive for the two-thirds of buyers who choose notebooks.
The plan is simple but audacious: use the industry-leading 22nm silicon process that Intel announced earlier this month to cut, for example, the median notebook power requirements to about 18 watts from the 35 watts on today’s 32nm Sandy Bridge process and architecture; then ramp volumes across the product line so that half the company’s volume entering 2014 uses the lower-power, better-performance silicon.
First, lowering power consumption as measured in watts — cut in half — without cutting performance means twice the battery life. For notebook users, that means a choice between a much thinner and lighter laptop — can you say “Apple MacBook Air for the masses at mass market prices?” — without losing battery life, or alternatively, twice the battery life. Intel is calling the next generation of thin-and-light notebooks “ultrabooks”, a trademark.
Second, these relative benefits will also accrue across the computing spectrum to desktops and servers, and especially enhancing Intel’s ability to deliver Systems on a Chip (SOCs) with Atom processors running at below a watt in tablets and smartphones. Lower electrical consumption is a high-priority in data centers and a growing priority and mandate in office desktops.
Why would Intel paint such a clear roadmap for the Ivy Bridge (2012) and Haswell (2013) generations of processors in a highly competitive market? In spite of record sales and profits, Intel’s stock is stuck in the low $20’s as some on Wall Street anticipate an Intel near-death experience from the late delivery of competitive smartphone and tablet chips. That won’t happen. The company wants investors to know that it’s not behind the eight-ball. Driven by rapidly growing emerging markets for traditional PCs and notebooks and a robust and continuing data center refresh cycle, the company claims it will grow at double-digit rates over the next three years without any revenue needed from smartphones and tablets.
But those Intel-based market-changing tablets and smartphones are coming, with highly competitive products in 2012 and after. Across a global economy and time, I believe tablets and smartphones are an additive computing market opportunity. Intel’s SOC engine, the Atom processor line, gains from planned process improvements from today’s 45nm to 14nm in 2o14 across three nodes of silicon process in three years — about twice the rate of Core and Xeon microprocessors.
A second reason that Intel paints such a clear roadmap is the company’s total confidence in the recently announced 22nm process. The fact that Intel will produce all of its 2012 Ivy Bridge microprocessors using 100% 3D tri-gate 22nm transistors tells me the company is convinced it has the technology completely in hand. And the company upped the cap-ex ante this year, deliberately building more 22nm chip-making capacity to come online next year.
Is lower voltage and hence lower electrical consumption that big a deal? Yes, it certainly is. The original MacBook Air, the archetypical thin-and-light notebook launched in January 2008, uses a 17 watt Intel Core 2 microprocessor with dual cores at 1.4 GHz. This year, 32nm Sandy Bridge ultra-low voltage chips will clock 1.7 GHz with quad processors at 17 watts — plus Turbo 2.0 temporary speed boosts up to 2.9 GHz. Ivy Bridge (22nm) in 2012 and the new Haswell micro-architecture in 2013 will improve that further.
The 2008 dual-core processor in the MacBook Air was a trade-off in notebook weight versus processor speed — the processor was OK but no speed demon. The quad core Sandy Bridge and subsequent generations are in the middle of notebook performance curves, now offering excellent computation capabilities while still sipping lightly from the battery. Mainstream notebooks will quickly migrate down to the 1.5-3.0 pound level, offer all day battery power, and a no-compromises user computing experience. They will be ultra-thin, ultra-responsive, and, with hardware and DNA from the McAfee acquisition, ultra-secure. As a road warrior, I look forward to taking permanently taking five pounds of carry weight off my shoulder.
The Atom processor will follow a similar but steeper power-to-performance curve as the heart of devices from the 0.1 watt to 10 watt range. As a betting technologist, the cards Intel dealt last week tell me Intel will have an easier job scaling power requirements down going forward than ARM processors will scaling performance up.
In fall 2009, the company said it was scaling from gadgets to mainframe-class serversacross the computing continuum with one Intel Architecture, and a common set of software, tools, and developer experience. Last week, the company said it was confidently using its new 22nm technology to push the computing continuum down the voltage curve, with benefits to all kinds and levels of computing. Having voluntarily laid its cards and reputation on the table, Intel is surely planning on under-promising and over-delivering. I would not bet against it.