Ed's Threads

Moore’s Law is Dead – (Part 1) What?

Featured Products

Previous Article
Three fundamental shifts
Comments (13)
  1. Juliusz says:

    It may be that our thinking will shift to calculations per second, and incorporate Trinary, crystals and harness light in new ways to keep Mr. Moore happy.

    1. edkor says:

      Industry has already shifted to discussing caluculations per second, flash memory multi-level cells already use 3 levels, silicon and all alternate channel materials are crystals, and people are already working on integrated photonics (all of which should allow innovation to continue…without doubling the number of components per chip).

  2. Sanders says:

    Well Moore’s law was declared dead 25 and 5 years ago, 20 and 10 years ago, and all the years in between. There was no shortage of prophets who declare the law dead.

    I guess will keep watching the show going forward, I bet 10 years and 20 years from now new prophets will start declaring Moore’s law dead.

    1. edkor says:

      Well, prior pundits lacked the current perspective on the economic limits imposed by lithography, as well as the scaling limits imposed by atoms…as future posts in this series will detail.

  3. Sam Beal says:

    “We” know the scope of Moore’s Law is semiconducting device technology. The pundits and futurists, e.g. Kurtzweil, use the term to define “technology evolution” which is geometric and by some calculations, accelerating. And is not restricted to semiconductors.

    1. edkor says:

      Everyone knows that the smallest device that can be built must be composed of atoms, and in Part-3 of this blog series I’ve shown that the smallest nanotube array we can theoretically build and possibly interconnect would be ~4nm pitch. Electronic circuit speeds are already limited by the lack of room-temperature superconductors and the lag in photo-electric conversion. No one has shown anything close to 4nm pitch optical circuits nor magnetic circuits that can compete with let alone surpass electronic circuits (except for niche applications). What else might we use for digital logic and memory?

      Kurtzweil was smart enough in the mid-1970s to rely on Moore’s Law acceleration of IC processing power through the end of the century to create first OCR and text-to-speach then decent music synthesis, and acquired deserved money and fame; however, Kurtzweil and comparible “futurists” lack experience in IC manufacturing technology, and their main business seems to be selling science fantasy (an honorable entertainment business IMHO but not to be confused with engineering or technology).

  4. Jack Ogawa says:

    I would say: New geometry nodes are very hard won, so the economic ROI becomes tougher to rationalize. This should force the industry to innovate more on circuits and architecture rather than simply ride the Moore’s Law wave. I think the semicon industry overall has become complacent. The over-emphasis on process geometry has put companies like TSMC in control of the industry’s profits (ask any small semi company how they are paying 2-3x for the same wafers that Broadcom uses), and has caused VCs to walk away from semis. Reset in value prop is required!

    1. edkor says:

      Good points. A relative shift in the IC value-add from manufacturing to design could indeed trigger more investment in fabless start-ups.

X