Intel Defines 22nm Innovation with '3D' Tri-Gate Transistors
The IT industry loves the concept of "innovation," but many vendors' hearts largely belong to just the most conventional sorts of wisdom. That adoration takes a number of shapes: Stone Age business models; dusty Neolithic technologies and architectures; fossilized go-to-market strategies. In point of fact, such vendors are more similar to staid industries and companies, where dependability trumps progress, than they might like to think.
I was considering this on May 4th when Intel announced the next "tick" of its continuing "tick-tock" development evolution -- upcoming Ivy Bridge chips that will mark the industry's first 22nm production and the first commercial use of the company's Tri-Gate "3D" technologies.
The Tri-Gate manufacturing process forms conducting channels on three sides of a vertical "fin" structure on the silicon substrate. This is in contrast to traditional "2D" planar bulk and partially depleted transistors and fully depleted silicon on insulator (FD SOI) technologies.
Analysts' initial reactions to Tri-Gate and Ivy Bridge were more muted than I expected, with some questioning Intel's basic assumptions and a few rhapsodizing about the inherent risks of the company's new production process.
To be fair, shifting production to an entirely new transistor design does constitute something of a roll of the dice, even for the world's largest microprocessor producer. Though the notion of 3D transistors has been around for nearly a decade, no vendor has ever demonstrated that it can manufacture chips based on the technology in commercially viable numbers.
So why is Intel investing the necessary billions in existing and new chip fabs to accomplish what many are openly questioning? There are three basic reasons:
- Dramatic performance/energy efficiency improvements
- Enhanced cost effectiveness
- Because it can
Going for the Efficiency Gold
As to the first point, chip performance in markets where Intel actively competes is measured in two distinct areas: high-voltage settings (conventional PCs, laptops and servers); and low-voltage settings (mobile handsets, smartphones, tablets and netbooks).
Intel dominates in high-voltage applications and devices, but in low-voltage markets -- outside of the success of its Atom processors in netbooks -- it has faced serious challenges from Qualcomm, Texas Instruments and other manufacturers of chips based on the ARM architecture.
According Intel, Tri-Gate transistors will provide the tools required for the company to continue leading in traditional markets and make it and its OEMs far more competitive in low-voltage applications.
Intel expects Ivy Bridge processors to deliver up to 37 percnet better performance in high-voltage settings than current generation Core, Xeon and Itanium solutions, meaning that desktops, laptops and servers based on those chips will enjoy dramatic overall boosts.
The new architecture will also bump performance in low-voltage Atom microprocessors by up to 18 percent, also a significant improvement.
More importantly, Intel estimates that the new chips will consume just half as much energy as current-generation 32nm microprocessors. That will certainly be welcomed by OEMs looking for ways to boost server and data-center energy efficiency, as well as the battery life of conventional laptops.
However, it will also allow Intel to field products that can compete far more effectively against ARM-based solutions powering the vast majority of mobile devices. That makes Tri-Gate an essential key to Intel's future success.
Improved Performance Without Premium Pricing
The new chips should also help Intel's efforts against resurgent traditional competitors, especially AMD and its upcoming Fusion processors.
AMD has spent a great deal of money, time and energy talking up Fusion's graphics optimization (via the GPU technologies it acquired from ATI) to the point where the company insists on calling the new chips "APUs," or accelerated processing units. That may be due, at least in part, to the fact that the company's CPU development efforts lag so far behind Intel.
AMD recently brought its first 32nm processor to market -- nearly 18 months behind Intel's 32nm Sandy Bridge. While the company says 22nm development efforts continue apace, we expect at least a similar delay before AMD delivers its own 22nm chips.
Why is that such a big deal? Because if Intel is correct in estimates that Ivy Bridge will be in production by the end of the year, it could mark the end of anything like performance parity between its own and AMD's CPU technologies, resulting in an increasingly insurmountable Intel lead.
That is especially the case if the company's Ivy Bridge cost estimates are accurate. According to Intel, achieving Tri-Gate's remarkable performance and efficiency improvements will add just 2-3 percent more to production costs than current-generation processors (vs. an estimated 10 percent more for 22nm FD SOI planar chips).
In other words, along with falling further behind in the CPU performance continuum, AMD is also likely to feel increasing pricing pressure.
Because It Can
Perhaps the most important issue of all related to the Ivy Bridge advancements is that Intel can accomplish them. The company has achieved its leadership position by carefully and continually expanding the boundaries of microprocessor development and production.
As noted above, the concept of 3D transistors is nothing new, and other vendors (notably IBM) are reportedly developing similar technologies. But by becoming the lone vendor producing Tri-Gate-based chips in commercially sustainable numbers, Intel can capture substantial technological bragging rights and first-mover market advantages.
This is not to say that Intel has fully eliminated all of the challenges related to Ivy Bridge. Refitting its four existing chip fabs to produce Tri-Gate-based products and building an entirely new facility will be costly, complex and time consuming. And the company manufacturing record is not without blemishes, as was proved by the $1 billion recall of faulty Sandy Bridge chips earlier this year.
Still, if Intel succeeds with Tri-Gate, it will have attained the next destination on its product road map and be well on the way toward future 14nm chips. It will have delivered what could be a crippling blow to its traditional competitors, and found an effective entry into new markets and opportunities.
Dependability certainly has its charms, especially in circumstances involving significant risks. But tangible progress defines and is a critical result of true technological innovation, a point that Intel's Ivy Bridge will very likely prove.