3D is the future! Fortunately, we’re referring to the marketing of Intel’s latest chip design, not Macs or TVs. Rik Myslewski explores…

Here’s a neat trick: Intel’s next-generation chips will require half the power of their current top-of-the-line microprocessors when providing the same performance -- or they will boost performance by around 37 percent when running at the same power levels. Intel achieved this by redesigning the transistor, that infinitesimal on-off switch that makes chips tick. Today’s microprocessors have a slew of those li’l switches—the Xeon X5670 in a fully loaded Mac Pro, for example, has over a billion crowded onto its 240 square millimeters.

Chipmakers fit an increasingly large number of transistors on tiny silicon slices by shrinking a chip’s process size -- a tricky concept to explain precisely, so let’s just say that it defines the size of a chip’s smallest elements. That bleeding-edge Xeon X5670, for example, has a process size of 32 nanometers (nm), which Intel’s recent breakthrough shrinks to 22nm. Smaller-process chips are generally faster and use less power, but Intel’s new 22nm process adds more benefits by rethinking how transistors work.

Stripped to its basics, a microprocessor transistor has four parts: a source, where the current comes from; a drain, where the current is going; a channel, the silicon between the source and the drain through which the current flows; and a gate, which switches the current on and off. The gate has traditionally lain flat on top of the channel between the source and the drain; Intel’s breakthrough is that it has taken the channel, rotated it 90 degrees, and stuck it up into the gate.

Doing this provides three key benefits. First, the surface area of the channel that’s touching the gate -- the part that conducts the current and is called the inversion layer -- now has two sides and a top, which provides more inversion layer surface area and thus more current flow. The more current flowing through a channel’s inversion layer, the faster the transistor can operate.

Second, when the channel is vertical, its base can be thinner, thus allowing the whole transistor to be smaller than a flat-channel—or planar -- transistor with the same-size inversion layer. The smaller each transistor, the smaller the chip; the smaller the chip, the more you can fit on a manufacturing wafer; the more chips per wafer, the less expensive each chip can be.

The third advantage is a bit harder to explain, but let’s give it a shot. Current in a planar transistor isn’t confined to its inversion layer—it leaks into the silicon underneath, especially when the transistor is off. Leaky current is bad. Not only does it waste power, but it also makes it harder to control the transistor’s on and off states.

Over the years, a laundry list of improvements has attempted to tackle this problem -- but I won’t bury you under them now, except to name one quite efficient method called silicon-on-insulator, or SOI. This technique works well, but it’s expensive. According to Intel, their new 22nm vertical-channel technique prevents leakage as effectively as does SOI, but more cheaply. It’s generally accepted that SOI adds about 10 percent to each chip’s cost, but Intel’s new method -- which they call Tri-Gate -- adds only 2 to 3 percent.

Microprocessor engineers have been working on vertical-channel transistor variations for years -- Intel, for example, first demonstrated a simple version back in 2002. But now Intel’s engineers have solved a host of tricky manufacturing problems and have made Tri-Gate the centerpiece of their next-generation 22nm microprocessors, codenamed Ivy Bridge, which will begin high-volume production later this year and will likely appear in your Mac next year. Intel would also love to slip a Tri-Gate chip into your iPhone or iPad -- but don’t hold your breath.

--

Since the late 1980s, Rik Myslewski has paid his rent by keeping an eye on Apple. He was editor-in-chief of MacAddict from 2001 until its transformation into Mac|Life in early 2007, and is now a member of the snarkily sophisticated team at London’s The Register, which is “biting the hand that feeds IT” daily at www.theregister.co.uk.