Moore's Law is slowing, and some computer architects report the doubling time of computation on a single chip moving from one year to twenty years.

True if, as of March 2019, Moore's Law was widely documented as slowing and some computer architects (for example Hennessy and Patterson) had publicly reported that the doubling time for single-chip performance or computation had grown from about one or two years to about twenty years. Mostly true if the slowing is documented and architects reported a doubling time near twenty years, but the 'from one year' baseline is slightly off (e.g. 1.5–2 years). Mixed if the slowing is documented but the reported figures differ substantially from twenty years (e.g. 3–10 years). Mostly false if the twenty-year figure comes from a narrow measure stated as general. False if no architect reported anything like this and Moore's Law was not considered to be slowing.

Set before the research began.

1. E1 A New Golden Age for Computer Architecture Communications of the ACM (Hennessy and Patterson Turing Lecture) · 1 February 2019

   “Although Moore’s Law held for many decades (see Figure 2), it began to slow sometime around 2000 and by 2018 showed a roughly 15-fold gap between Moore’s prediction and current capability, an observation Moore made in 2003 that was inevitable.”

2. E2 A New Golden Age for Computer Architecture Communications of the ACM (Hennessy and Patterson Turing Lecture) · 1 February 2019

   “Between 1986 and about 2002, the exploitation of instruction level parallelism (ILP) was the primary architectural method for gaining performance and, along with improvements in speed of transistors, led to an annual performance increase of approximately 50%.”

3. E3 A New Golden Age for Computer Architecture Communications of the ACM (Hennessy and Patterson Turing Lecture) · 1 February 2019

   “The end of Dennard scaling and Moore’s Law and the deceleration of performance gains for standard microprocessors are not problems that must be solved but facts that, recognized, offer breathtaking opportunities.”

4. E4 Measuring Moore's Law: Evidence from Price, Cost, and Quality Indexes National Bureau of Economic Research (Working Paper 24553, Kenneth Flamm) · April 2018

   “It now appears that this two-year cycle for technology nodes definitively ended in 2014, with deployment of the 14nm node.”

5. E5 A New Golden Age for Computer Architecture Communications of the ACM (Hennessy & Patterson Turing Lecture article)

   “Although Moore’s Law held for many decades (see Figure 2), it began to slow sometime around 2000 and by 2018 showed a roughly 15-fold gap between Moore’s prediction and current capability,”

6. E6 A New Golden Age for Computer Architecture Communications of the ACM (Hennessy & Patterson Turing Lecture article)

   “Between 1986 and about 2002, the exploitation of instruction level parallelism (ILP) was the primary architectural method for gaining performance and, along with improvements in speed of transistors, led to an annual performance increase of approximately 50%.”

7. E7 Measuring Moore's Law: Evidence from Price, Cost, and Quality Indexes (NBER Working Paper 24553) National Bureau of Economic Research · April 2018

   “The logarithmic scale used in Figure 4 obscures a fairly dramatic slowdown in improvement in CPU performance after the millennium.”

8. E8 Concepts Introduced in Chapter 1 (computer architecture lecture slides) Florida State University (course slides, CDA5155)

   “The growth rate slowed even more after 2015 due to less increase in the number of transistors on a chip (slowing of Moore’s Law).”

9. E9 January 2018 – Soft Machines, by Richard Jones Soft Machines (Richard Jones blog) · 14 January 2018

   “But after 2011, according to Hennessy and Patterson, the growth rate slowed further – down to 3.5% a year since 2015. In principle, this corresponds to a doubling time of 20 years”

10. E10 January 2018 – Soft Machines, by Richard Jones Soft Machines (Richard Jones blog) · 14 January 2018

    “It was around 1986 that the rate of change really took off – between 1986 and 2003 computer power increased at an astonishing 52% a year, a doubling time of just a year and a half.”

11. E11 January 2018 – Soft Machines, by Richard Jones Soft Machines (Richard Jones blog) · 14 January 2018

    “now that that exponential growth in computing power has come to an end, those people who lived through that atypical period are perhaps the least well equipped to deal with what comes next.”

12. E12 Mundo Chip tenta se aproximar da promessa e do perigo da IA cryptoid.com.br (republishing a ZDNet article) · 14 July 2019

    “(The figure is slightly outdated; an essay by Patterson and Hennessy back in February pegged the slowdown to 3% improvement per year.)”

By early 2019, independent sources widely documented that Moore's Law was slowing. Hennessy and Patterson's Turing Lecture article says it "began to slow sometime around 2000" and refers to "the end of Dennard scaling and Moore's Law" (E1, E3). An NBER paper reports that the two-year technology-node cycle ended in 2014 and describes a sharp slowdown in CPU performance gains (E4, E7). A blog post citing Hennessy and Patterson says single-processor performance growth fell to about 3.5% a year after 2015, which it equates to a doubling time of about 20 years, and a 2019 article cites the same authors at about 3% a year (E9, E12). That supports the twenty-year figure, but the earlier baseline was about 52% a year, a doubling time of about 1.5 years rather than one year (E2, E10), so the claim fits the criteria's "mostly true" case.