Is Moore’s Law Alive, Dead, or Pining for the Fjords? Even Experts Disagree
At Semicon West 2019, a panel of industry experts kicked off a debate on whether Moore’s Law — the great prediction given to us by Gordon Moore, which declared that the number of components per integrated circuit would regularly double over a predictable period of time (originally 12 months, later expanded to 24 months) — was still alive. Over the last decade, discussions of whether Moore’s Law was sustainable in the long term or had already died and been replaced by other methods of scaling have become more common.
Unsurprisingly, there are plenty of views on the topic. I’ve argued in the past that while the text of Gordon Moore’s 1965 prediction hasn’t changed, the ways people have understood it very much have. Almost from the beginning, Moore’s Law — which began as a prediction of transistor density — has been expanded to include predictions about performance. The concept of node scaling has changed over the decades as well; it’s been 20 years since node names were a literal reference to any specific feature size. When TSMC, Samsung, or Intel talk about a new node, what they mean is that they’ve combined a set of new technical approaches, smaller feature sizes (in some cases), material changes, and manufacturing improvements that collectively justify declaring that we have developed a new way of manufacturing transistors.
That kind of definition leaves room for a lot of play. According to Aart de Geus, co-CEO of Synopsys:
The way to think about it is Moore’s Law is the behavior of an exponential that has techonomic feedback on the exponential that drove a revolution of what mankind can do. The reason I’m saying it’s completely alive is that right now we’re facing another decade or two of amazing opportunities that themselves economically will drive the push for technology without a stop. Maybe it’s not exactly the same curve that Moore actually drew, it doesn’t matter.
Victor Peng, CEO of Xilinx, distinguished between traditional Moore’s Law scaling that offered advantages in performance, power, and area (this is Moore’s Law + Dennard Scaling, and the fact that the CEO treated the question this way is its own evidence for how these concepts were blended together over the decades). According to him, where it was once possible to see advantages in all three of those areas over time, companies are now required to focus on one or two of them, and that in that sense, Moore’s Law wasn’t working any longer.
AMD CEO Lisa Su, who recently launched the most revolutionary CPUs her company has kicked out the door in 15 years, doesn’t see Moore’s Law as being dead, but believes it is slowing, and that companies must incorporate more and more technology from outside the realm of traditional node scaling to succeed. Other industry speakers shared their own views (Semi Engineering has more). One point that’s made is that how you think Moore’s Law is evolving depends significantly on what other technologies your industry is tapping to continue pushing silicon advances forward, whether that be 3D NAND stacking, advanced packaging techniques, or new material technologies that yield better performance.
Yet, despite the broad debate over whether Moore’s Law can be properly characterized as “dead,” I think the very fact that we’re arguing the point is itself instructive. If you’ve been a computer enthusiast for long enough, you recall a time when even asking the question “Is Moore’s Law dead?” would have gotten you a funny look. Whether Moore’s Law referred solely to transistor densities or if it included aspects of performance and power consumption as well — these are factors that people have been arguing and confusing for decades. The fact that it existed and functioned, however, was inviolate. Today, we see people openly searching for alternate methods of definition that allow Moore’s Law to continue to be true, mostly because the classic definition factually no longer works.
But once you’ve arrived at the point where the old definition factually no longer works, you’ve also arrived at a transition point. It may be useful to keep the concept of Moore’s Law around as a way of describing equivalent improvement, but in doing so, we’ve changed the fundamental nature of what Moore’s Law was supposed to be. Ultimately, Gordon Moore’s seminal paper becomes a sort of ship of Theseus problem — at what point is Moore’s Law itself no longer Moore’s Law? When have we divorced the engineering reality of what was initially described so thoroughly from what the law purportedly refers to that one no longer has much to do with the other?
It would be downright amusing if, 500 years from now, the AI-based hard light constructs that serve as executives to major human-corporate conglomerates are still blithely chattering on about the advances in Moore’s Law as they describe improvements to the latest quantum peanut butter computing singularity — based on technology Gordon Moore never heard of and circuit designs he couldn’t possibly imagine.
Given the way marketing works, I wouldn’t bet against it.
- Chiplets Are the Future, but They Won’t Replace Moore’s Law
- Moore’s Law at 50: Its past and its future
- The myths of Moore’s law