Miniaturizing processor manufacturing is the engine that has kept Moore's Law hot all these years. As we approach the single digit in the nanometer count, enormous technical challenges arise, forcing us to rethink the current mechanics of silicon. IBM, in partnership with the State University of New York, GlobalFoundries and Samsung, has taken a huge step with its new 7-nanometer chip, the first to work correctly under that technology.
Today’s manufacturing processes
The best manufacturing process we can access today for a processor is 14 nanometers. That means getting a system with one of Intel's Broadwell chips, or the latest Galaxy S6 from Samsung which have an Exynos SoC at that scale. The feeling about the latest compressions is that the benefits are increasingly less marked. Intel has focused more on reducing power consumption than increasing performance, a strategy that for many users simply doesn't justify replacing an entire platform. Even so, improvements don't stop. Moore's Law has become synonymous with Highlander, and no matter how many times its death is declared, it always finds a way to rise again.
A new manufacturing breakthrough
The new evolutionary leap in chip manufacturing comes thanks to IBM and the State University of New York, along with the Common Platform alliance involving the folks at GlobalFoundries (a company that “acquired” IBM’s chip division with $1.5 billion on top), and Samsung. It's a prototype chip manufactured at 7 nanometers with operational transistors. To break the ten-nanometer barrier, IBM had to implement so-called EUV lithography (Extreme UltraViolet), plus a channel formed by an alloy of silicon and germanium, replacing pure silicon. IBM anticipates that moving from ten to seven nanometers should lead to a surface area reduction close to 50 percent, a parameter that excites Big Blue and its partners. The idea is to achieve minimum 50 percent improvements in the power-performance ratio, once all the sharp edges of the process have been smoothed out.
Of course, this last point will be quite complicated. Using germanium helps increase electron mobility below ten nanometers (meaning, its involvement is non-negotiable, so to speak), and EUV lithography is still very expensive. In fact, considering the words of Mukesh V. Khare, IBM Research's head of sub-10nm processes, we basically have to expect processors with a slightly higher price, which will justify the increase through their performance. Anyway, patience. We'll have to wait at least until the second half of 2016 to see 10 nanometers in action, a detail that will probably put 7 nanometers in 2017 or 2018 if no setbacks arise.
