Intel 3-D Transistors: Why and When?
A few days ago, Intel teased: On May 4th, the company would make “its most significant technology announcement of the year.”
Tongues wagged. Will Intel make ARM chips for Apple? The speculation has roots in reality.
We’ll start with the public breakup of the Wintel marriage. At this year’s CES in January, Steve Ballmer made it clear that x86 exclusivity was done for. With an eye on reentering the tablet market, the next release of Microsoft’s legacy OS, Windows 8, would also run on ARM SOCs. This will “fork” Windows: There’ll be two versions, one on x86 processors, another on ARM chips. Tablets, which introduce UI differences, add a couple more tines to the fork. The impact on application development isn’t clear yet (food for a future Monday Note). Surprisingly, there’s been little talk of Intel “going ARM” to repair the Wintel relationship.
Now let’s consider Intel’s complete absence from the mobile scene. Not a single smartphone contains an x86 processor. Not a single tablet, no GPS device, nothing.
For the past four years Intel has told us we’d see x86 mobile devices Real Soon Now. The company developed its own mobile version of Linux, MobLin, and they made a big deal of joining forces with Nokia’s Maemo to create MeeGo. But Nokia’s new CEO, Stephen Elop, kicked Meego to the kerb, wisely deciding to focus on one software platform, his ex-employer’s Windows Phone 7.
(We’ll see how wise this decision turns out to be. Perhaps Elop should have put his money on the front-running Android horse. Perhaps Microsoft should have “gone Apple” — pardon, “vertical.” They could have acquired Nokia, controlled the hardware and the software. They did so, successfully, with the Xbox and Kinect. Again, more food for future Monday Notes.)
The x86 mobile devices never materialized. Each new low-power processor promise from Intel was matched by ever more attractive ARM development. Now that the PC market is in its twilight, with mobile devices proliferating and stealing growth from the PC, surely Intel has to get into the race.
Then there’s the long-standing relationship between Steve Jobs and Intel — or, more specifically, with Intel co-founder Andy Grove. The relationship flourished at NeXT when Jobs moved the platform to Intel processors. After Jobs returned to Apple, efforts got under way to move the Macintosh away from the PowerPC, which was deemed poorly supported by IBM and Motorola, to the more robust x86 line.
It isn’t hard to imagine Intel offering Apple its advanced 22nanometer fabs, along with some kind of exclusivity and price advantage. And there’s a bonus: They’d be kicking Samsung, an annoying combination of supplier, competitor, and adversary in IP lawsuits. In return, Apple would give Intel the kind of volume the company likes, 100 million ARM chips in 2012.
From there, the train of thought continues to the terminus: the Macintosh line switches wholly to ARM, and Intel supplies the processors. It’s not impossible. Intel hedges its bets, secures an inexpensive ARM license and uses its technology and marketing prowess to grab their share of the explosive growth.
As the rumor site says: “This is going to cause meetings.”
Now, the reality.
What Intel announced last week is a new “3-D” transistor technology. 3-D here doesn’t refer to images but to a design and manufacturing technique: Making transistors in three dimensions, as opposed to today’s “planar” technology where the microscopic silicon circuitry is laid out on a flat surface. Just as you can store more cars in a multi-storey garage than in a flat parking lot, more circuitry can be packed in three dimensions.
The new 22nm semiconductor manufacturing process also helps. The circuitry building blocks are smaller, they waste less electrical power through heat dissipation. All of this — cue the cymbals — is ideal for mobile applications. In plain English: This is Intel’s ARM killer. (Cruelly, Google tells us we heard the same story three years ago. And two years ago. And last year.)
Intel’s press release is firmly planted in hyperbole:
“Intel’s scientists and engineers have once again reinvented the transistor, this time utilizing the third dimension,” said Intel President and CEO Paul Otellini. “Amazing, world-shaping devices will be created from this capability as we advance Moore’s Law into new realms.”
The part about “once again” reinventing the transistor is a bit far-fetched. On Intel’s website, you’ll find the company’s own timeline, replete with innovations, and bowdlerization…but nothing about reinventing the transistor. There’s some dispute as to the transistor’s actual invention: when, where, by whom. Most history books credit William Shockley at Bell Labs Research with the first silicon transistor, which was produced in 1954 by Texas Instruments. (At my Breton Roman Catholic boarding school, the Prefect of Discipline was a certified geek. In 1955, instead of looking at religious pictures, we were in his office drooling at this incredible Philips OC 71 germanium transistor…)
We’re meant to be impressed by the promised performance and power dissipation improvements:
The 22nm 3-D Tri-Gate transistors provide up to 37 percent performance increase at low voltage versus Intel’s 32nm planar transistors. This incredible gain means that they are ideal for use in small handheld devices, which operate using less energy to “switch” back and forth. Alternatively, the new transistors consume less than half the power when at the same performance as 2-D planar transistors on 32nm chips.
Note the Alternatively: it’s either more performance or less power dissipation.
We’ll have to wait a year to see how this markitecture translates into actual devices.
Will this be enough to unseat ARM? Most observers doubt it. The big news was received with an equally big yawn. Wall Street didn’t pay much attention. We’ve been here before: The “product” of the announcement is the announcement. (And there’s the suspicion that “breakthrough” revelations are an attempt to mask a lack of spanking new products.)
But let’s return to the rumor, from SemiAccurate, that the Mac and Intel will soon be “arm-in-ARM.” (That bad pun isn’t mine.)
First, let’s consider the name of the website.
Second, what will Apple do at the high-end, for media creation and editing? What about Photoshop, FinalCut, and other applications, including CAD where the Mac is getting back in the game? There’s no roadmap for ARM chips to beat Intel in these computationally intensive areas.
Today, going ARM is technically feasible on entry-level Macs. Tomorrow, newer multicore ARM chips might work for middle-of-the-line Macintosh products. But will Apple abandon the faster x86 processors at the high end just to avoid the kind of forking that awaits Windows in its own move to ARM? If not, we’ll again see Universal applications (a.k.a. fat binaries — two versions inside the same container), just as we did with the PowerPC to x86 transition. Microsoft is doing it because it must; Apple did it because the PowerPC didn’t have a future. But now?
— — — — -
On a related note…and more food for thought: I’d love to know how the iPad line will evolve. For example: will pressure-sensitive stylus input ever happen? Eschewing stylus input in the early days was a thoughtful move. Perhaps it’s time to relax the restriction and thus enable richer media creation applications.
The next iOS and OS X releases will shed more light on the relative roles of Apple’s tablet and PC product lines, how they will coexist, what they’ll have in common and what will keep them apart. We should know in about a month.
