WHATEVER HAPPENED TO 10NM?
With Intel stringing its 14nm production node out to at least four generations (that’s generations with a small “g”) of Core processors, and having officially delayed the 10nm node, is Moore’s Law dead? Moore’s Law, of course, is the observation that computer chips double in complexity or halve in cost—or some mix of the two—every couple of years.
The simple answer is yes. If you define Moore’s Law by that rigid twoyear time frame, then it’s probably done and dusted. Progress with conventional computer chips will continue, but at the slower pace we’re already seeing. After all, Intel’s first 14nm processors went on sale in late 2014. Nearly three years later, we’ve yet to see any 10nm chips, even if Intel is promising them by the end of the year.
You could argue some new paradigm is needed, be that quantum computing or some other fundamentally new technology, if the next 40 years or so are to be anything like the last 40 years when it comes to the growth in computing performance. In the shorter term, however, the fact that Intel is no longer shrinking its transistors on a regular biannual cadence isn’t the end of the world. Intel has shown with the 14nm process that significant gains can be had in terms of power consumption, courtesy of revising an existing process.
Of course, smaller circuit features aren’t just about power consumption. They also allow more transistors to be made available, or the same number of transistors at a lower cost. But even limited access to larger transistor counts might not be a total disaster.
One effect of Moore’s Law over the years has arguably been a certain inefficiency in chip design. Put simply, engineers have always been able to hurl more transistors at a given computational problem. But with Moore’s Law coming to an end, chip designers will be much more incentivized to hone and polish their designs. Odds are, quite a bit of performance has been left on the table over the years. Clawing some of that back should buy several generations’ worth of performance improvements while the industry searches for an alternative to the conventional silicon-based integrated circuit.