Jump to content

Talk:Moore's law/Archive 5

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia
Archive 1←Archive 3Archive 4Archive 5

major enabling factors and future trends -- update

section "Major enabling factors and future trends", bullet point #5:

it states that in 2012, the minimum size of features is 22nm.

Would be best to mention that in 2014, chips built with 14nm features now exist. http://www.chipworks.com/en/technical-competitive-analysis/resources/blog/intels-14-nm-parts-are-finally-here/ What Intel calls "14 nm" is not 14 nm. Is the old processing. Intel managed to make more transistors with the old processing, by tweaking it. It complies with Moore's law, but is not a smaller (14 nm) size. — Preceding unsigned comment added by 186.59.44.223 (talk) 14:45, 26 January 2015 (UTC)

i'd add it myself, but I'm not sure how — Preceding unsigned comment added by Cerewa (talk • contribs) 18:50, 11 December 2014 (UTC)

Double every year?

The article says "... the number of transistors in a dense integrated circuit doubles approximately every two years. The observation is named after Gordon E. Moore, co-founder of the Intel Corporation, who described the trend in his 1965 paper." His 1965 paper predicts doubling time of ONE year. Bubba73 You talkin' to me? 04:18, 10 February 2015 (UTC)

As noted in the History section Moore in 1975 restated his observation to doubling approximately every two years which remains to this date the conventional definition. It is the conventional definition that is appropriate for the lede. I suppose we should take the 1965 date out of the lede Tom94022 (talk) 06:52, 10 February 2015 (UTC)
Moore's law was formulated in 1965, and revised in 1975. The correct date is 1965.71.128.35.13 (talk) 18:47, 20 March 2015 (UTC)

Moore's Law as we know it today was both stated and named in 1975, not 1965 and this is clearly described in the History section. Moore's 1965 observation of exponential growth at the rate of doubling each year was to the best of my knowledge never known as "Moore's Law" and therefore does not belong in the lede since in is time inapposite. I suppose if someone can find a 1965 usage of the term "Moore's Law" then the 1965 date should go into the lede Tom94022 (talk) 05:45, 23 March 2015 (UTC)

Moore's law was formulated in 1965, and revised in 1975. Gordon Moore wrote an entire chapter of of a book documenting the 1965 origin date, under the title "Moore's law at 40." As a matter of fact, Moore wrote this in 2006.[1] 71.128.35.13 (talk) 23:24, 23 March 2015 (UTC)71.128.35.13 (talk) 03:45, 24 March 2015 (UTC)

References

  1. ^ Moore, Gordon (2006). "Chapter 7: Moore's law at 40". In Brock, David (ed.). Understanding Moore’s Law: Four Decades of Innovation (PDF). Chemical Heritage Foundation. pp. 67–84. ISBN 0-941901-41-6. Retrieved March 15, 2015. Following a paper that I wrote in 1965 and a speech that I gave in 1975, the term "Moore's law" was coined as a name for a type of prediction that I had made. {{cite book}}: line feed character in |quote= at position 67 (help)
Thanks for the book reference, it too states Moore's Law as a term of the art was not formulated until 1975 at which time the associated rate was then and now doubling every two years. Since we now have two references that clearly establish this I am reverting original state before yr first revision. I believe that in accordance with WP:BRD guideline the page should remain in this original state until discussion herein results in a consensus. Tom94022 (talk) 16:29, 24 March 2015 (UTC)
According to Intel Corp., "1965: Moore’s Law is born when Gordon Moore predicts that the number of transistors on a chip will double roughly every year (a decade later, revised to every 2 years)"[1]-71.128.35.13 (talk) 21:08, 24 March 2015 (UTC)
A 2011 press release is not sufficient to establish that Moore's Law existed as a term of the art earlier than the first mention in 1975. Accordingly I am again reverting. U really need to discuss and gain consensus before reverting again. Tom94022 (talk) 22:02, 24 March 2015 (UTC)

References

  1. ^ "Over 6 Decades of Continued Transistor Shrinkage, Innovation" (Press release). Santa Clara, California: Intel Corporation. Intel Corporation. 2011-05-01. Retrieved 2015-03-15. 1965: Moore's Law is born when Gordon Moore predicts that the number of transistors on a chip will double roughly every year (a decade later, revised to every 2 years)
Reliable references support 50 years and 1965, but not 40 years and 1975. Here are just a few: Intel CEO Brian Krzanich (2014): "Moore's Law will reach its 50th anniversary next year"[http://forwardthinking.pcmag.com/none/329835-intel-sees-path-to-extend-moore-s-law-to-7nm
Forbes magazine (March 2015): 50 years ago next month ... an "idea that would later become known as Moore’s Law." [3]
Technical paper titled "Fifty Years of Moore’s Law" [4]-71.128.35.13 (talk) 22:37, 24 March 2015 (UTC)
None of which predate 1975. Tom94022 (talk) 19:29, 25 March 2015 (UTC)

Not everyone in the world needs to adhere to the same rigid definition of what is "Moore's Law" or of when it began. Even if that phrase was not used until 1975, it describes a phenomenon that was (at least roughly) described by Moore a decade earlier, and any term such as "Smith's Law" is almost never expressed using that precise phrase until long after the publication by the "Smith" who described the concept that will later bear the name. (Obviously, naming something after yourself when you first publish it would require some bravado or a strong sense of humor.) Whether the doubling period is one year, two years, or 18 months is a relatively minor quibble. So is the question of whether it strictly refers to the density of transistors per square cm, the data processing speed, data storage capacity, or some other measure of the capability of technology. To many people, and probably to most people, "Moore's Law" is merely the idea that technology progress is relentless and rapid, regardless of whether some curve has an exponential shape or not. Moreover, some people might reference the date at which the phenomenon started occurring rather that the date when someone noticed it, described it, and suggested that it might continue to happen. Moore himself commented about the broadness and vagueness of the definition in practice, and so have others. The Wikipedia article should (and to a large extent already does) acknowledge that the concept known as "Moore's Law" is not a single precise string of words that was given to mankind as etchings on stone tablets that emerged from a burning bush at a single point in time. To me, it seems really strange to see someone want to remove a citation to a book chapter about Moore's Law that was written by Moore himself. —BarrelProof (talk) 23:10, 24 March 2015 (UTC)

Actually I pretty much agree with everything u say, which is why IMO the specifics when and what Moore said doesn't belong in the lede. The deletion of the citation was only a consequence of its location, it like most of what the IP is adding belongs in the history section. Tom94022 (talk) 19:29, 25 March 2015 (UTC)
Furthermore, the lede does not have the general definition u like but instead starts out by giving the generally accepted and current version of ML with regard to complexity (doubling of transistors every two years) and then under the current version confuses the issue with the 1965 observation. Tom94022 (talk) 19:43, 25 March 2015 (UTC)
I see. I was misinterpreting the removal of the citations as a desire to keep them out of the article entirely. —BarrelProof (talk) 20:01, 25 March 2015 (UTC)
I agree entirely with your comment, BarrelProof. The article now discusses doubling times correctly and the 50 years of Moore's law (since 1965), supported by reliable references.71.128.35.13 (talk) 20:26, 25 March 2015 (UTC)
If we must split hairs, I think it's important to note that Moore's law describes an historical trend that began with the invention of the integrated circuit in 1958 and has continued ever since. The trend began in 1958, the trend was first described in 1965 (by Moore), and this description was given a name in 1975. So we have three different start dates: the start of the trend, the first description, the naming of the description.
The term "Moore's law" could refer to the trend (starting in 1958), or the description of the trend (first published in 1965). I don't think the 1975 date is relevant at all -- this is the birthdate of the name of the description of the trend. Naming the description of a trend is not a particularly interesting act, thus this is not a significant date. ---- CharlesGillingham (talk) 06:52, 31 March 2015 (UTC)
This article looks through the narrow lens of Moore's law, a 1965 forecast which was revised and christened a decade later. As you say, 1975 is not a significant date. As Note 1 of this article observes, "The trend begins with the invention of the integrated circuit in 1958." Ray Kurzweil says the broader technological trends started earlier than 1958, and are broader than integrated circuits.— 71.128.35.13 (talk) 01:23, 1 April 2015 (UTC)
The first sentance of the lede states:

"Moore's law" is the observation that, over the history of computing hardware, the number of transistors in a dense integrated circuit doubles approximately every two years.

If this sentance is correct, then there is no dispute that Moore formulated this observation in 1975. If this is incorrect and Moore's Law is more generally about exponential growth of semiconductor technology then the next best date is 1958 when the exponential trend began and the lede's lead sentence needs to be changed. There is really no justification for 1965 other than it was 50 years ago. Tom94022 (talk) 21:54, 1 April 2015 (UTC)
Finally, we are talking about lede content, not the article. IMO both 1958 and 1965 belong in the history section and not the lede. Tom94022 (talk) 22:09, 1 April 2015 (UTC)
Perhaps we should just say

"Moore's law" is the observation that, over the history of computing hardware, the number of transistors in a dense integrated circuit has doubled every one or two years.

I any case, I don't think most people use the term "Moore's law" as precisely as you are. The exact rate is not as important as the exponential trend, and I think that we shouldn't exclude the vast improvements that occurred during the 60s on a technicality. ---- CharlesGillingham (talk) 23:10, 1 April 2015 (UTC)

Moore's law is an observation, not a formulation or a trend. It was revised in 1975. Most sources, including the transistor chart in the lede, support "doubles approximately every two years."— 71.128.35.13 (talk) 22:18, 2 April 2015 (UTC)

Moore made many observations and associated forecast with many values ascribed to a law of Moore but I think we agaree that Moore's Law == "doubling every two years" Actually Moore's observations in 1965 and 1975 and associated forecasts were not doubling every two years and I think it wasn't until 1979 that doubling every two years became the "Law" (see The Lives and Death of Moore's Law. Again IMO all of this is interesting historical material that belongs in the History Section and not in the lede. Tom94022 (talk) 16:27, 3 April 2015 (UTC)
This is what Moore said in 1975

... the rate of increase of complexity can te expected to change slope in the next few years as shown in figure 5. The new slope might approximate a doubling every two years, rather than every year, by the end of the decade.

Moore, 1975 IEEE Electron Device Meeting

So Moore's 1975 observation is that complexity has been and doubling every year and will continue at that rate until about 1980 at which point the rate will decrease to doubling every two years. Again interesting history belonging in the History Section. At this point I think we should take most specific dates out of the lede, perhaps only something about ML describing exponential growth in semiconductor complexity since 1959 generally accepted to be doubling every two years since at least the late 1970s Tom94022 (talk) 17:01, 3 April 2015 (UTC)

My personal understanding is that Moore's Law is the notion that technical progress (esp. for transistor counts for ICs) is approximately exponential. The phenomenon started before 1960 and was the subject of a publication by Moore in 1965. Following that publication, people have observed that it applies in various areas and have refined their estimates of the rate of the exponential growth. The estimates of the doubling period since the mid 1970s have been in the range of 18 months to 2 years, and after about 1980 it has seemed like the two-year estimate (an estimate published by Moore in 1975) has been a good one. —BarrelProof (talk) 19:11, 3 April 2015 (UTC)

This is my understanding, too. The late-1970s interpretation is unsupported WP:OR and ambiguous. The article's transistor chart starts in 1971.— 71.128.35.13 (talk) 20:22, 3 April 2015 (UTC)
Based upon these two understandings, the first sentence is incorrect.
Moore's Law comprises both observations AND predictions; Moore's Law according to Moore is shown in Figure 8 of the 2006 MOORE’S LAW AT FORTY and is a piece wise linear graph on a semilog plot having a slope doubling every year from ~1962 until 1980 and then doubling every two years from 1980 until ~2002. Actual industry data in Figure 9 shows Moore's prediction of a kink at 1980 was wrong, the kink had already occured circa 1973. Again interesting historical material but not for the lede.
Actually I now think paraphrasing Moore's own words about his law may make the most sense in the lede, something like:

Following a paper written by Intel's Gordon Moore in 1965[1] and his speech in 1975,[2] the term "Moore's Law" was coined as a name for his prediction of exponential growth in semiconductor complexity. His 1975 speech paper predicted a doubling in complexity every two years which has proven to be accurate, in part ... Over time the term has been used much more broadly, refering to almost any phenomenon related to the semiconductor that exhibits exponential growth.[3]

...

Tom94022 (talk) 07:13, 4 April 2015 (UTC)

References

  1. ^ suitable reference
  2. ^ suitable reference
  3. ^ MOORE’S LAW AT FORTY
This focuses wrongly on the coining of Moore's law, not the observation itself. It's over-complicated, makes little sense, and oxymoronic considering that "coining" already subsumes "naming."— 71.128.35.13 (talk) 18:31, 4 April 2015 (UTC)
It is pretty much Moore's own summary language looking back at 40 years of Moore's Law but I do agree that it could be improved by removing the minor redundancy. It accurately summarizes the current statement of Moore's Law unlike the current lede which glosses over Moore's inaccurate forecast of 1975. It replaces the first sentance which at least three editors and Moore agree is wrong. What's wrong about that? Tom94022 (talk) 17:58, 5 April 2015 (UTC)
The first sentence of the article is accurate. The proposed change is complicated, and makes little sense. It would be wrong to shift the focus away from the observation itself, to the coining of Moore's law.— 71.128.35.13 (talk) 22:43, 5 April 2015 (UTC)
While u are entitled to your POV and opinions, I think we as editors should defer to Moore's 2006 view of Moore's Law. The current first sentance is inaccurate according to the Moore article, among others, in that:
  • ML is a predcition based upon a set of observations
  • In fact from 1959 to about 1973 complexity doubled annualy, not every two years as stated.
Furthermore, 1965 is just a date when Moore predicted annual doubling until 1975, as it turns out a correct prediction but not the current prediction and not nearly as relevent in the lede as 1959 (the date at which exponential growth started (according to Moore 1965) or 1973 when doubling every two years actually began (according to Moore) or 1975 when Moore predicted doubling every two years beginning circa 1980. Actually the focus should be on the prediction itself, that is doubling every two years made in 1975. It is actually your insistance on confusing the lede with the earlier 1965 prediction that complicates the lede. Tom94022 (talk) 16:52, 6 April 2015 (UTC)

The proposed changes are unsupported.

  • This is the 50th anniversary of the birth of Moore's law in 1965.[5]
  • "It was an observation and a projection," Moore said in 2015.[6]
  • "A decade later, he [Moore] revised what had become known as Moore’s law: The number of transistors on a chip would double every two years," according to Dean Takahashi writing in the Seattle Times.[7]
  • References support the historical validity of Moore's law, and the first chart in this article shows a two year doubling time during 1971-2011.— 71.128.35.13 (talk) 22:05, 6 April 2015 (UTC)
Does no one like this language "doubling every one or two years"? It succinctly and clearly describes both versions. ---- CharlesGillingham (talk) 14:30, 14 April 2015 (UTC)
Moore wasn't comfortable with just splitting the difference, and calling it 1.5 years: "The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and their being faster). Despite a popular misconception, Moore is adamant that he did not predict a doubling "every 18 months." Rather, David House, an Intel colleague, had factored in the increasing performance of transistors to conclude that integrated circuits would double in performance every 18 months." He clearly said one year, and revised it to two years. 71.128.35.13 (talk) 23:06, 18 April 2015 (UTC)

MansourJE (talk) 04:08, 29 April 2015 (UTC)

Other formulations

Recently, I did some original research into other formulations of Moore's law (the section at the end of the article) and I discovered about 50 mentions of "Moore's law of X" or "Moore's law for Y." My website has a preliminary plot of the data at http://www.tedsanders.com/graphs/moores-law-of-moores-laws/. Do you think this research, once cleaned up and sourced, would be worth mentioning in the 'other formulations' section? Right now the 'other formulations' section is not at all organized or comprehensive - it just seems to be a few examples picked arbitrarily. Thanks for any thoughts. Ted Sanders (talk) 23:01, 26 June 2015 (UTC)

Inclusion of "neurogrid" is a mistake. Can I delete it?

I found it curious that the "neurogrid" of 2014 was listed here as relevant to Moore's law (or future trends). It's not a transistor computer. Worse, it claims that it's 9000 faster than a P.C. That doesn't sound right! Sure enough, I followed the link and found that it's 9000 times faster than a PC *simulating* it. ([1]) Whoops! (Even if you could interpret that as being 9000 times faster at one rather contrived thing, it's still entirely misleading as written). So what's written on this page is quoting a mistake (Stanford News service's summary of its own article), and is also irrelevant. It should be removed. Objections? Theredsprite (talk) 01:00, 10 July 2015 (UTC)

It should be removed; no objections here.71.128.35.13 (talk) 23:58, 16 July 2015 (UTC)
  1. ^ http://news.stanford.edu/pr/2014/pr-neurogrid-boahen-engineering-042814.html

Thermodynamic limit

Could this be summarized briefly, with a link? The exposition is too long. 71.128.35.13 (talk) 00:26, 18 July 2015 (UTC)

More than could it should be substantially reduced in size and complexity! Tom94022 (talk) 05:11, 18 July 2015 (UTC)

The picture in the top right corner only goes to 2011

this image

Shouldn't the picture be updated soon? Why not update in on a yearly basis? Why not make a script to update it automatically, even more often? Zanthius (talk) 19:11, 6 February 2014 (UTC)

Interesting idea. Is there a source of data that would support automatic updates? You could ask the author of the image where he got data and see if he can easily update. I'd like to see other improvements, too, like something that makes it easy to see a doubling every two years; I can't tell from the line whether it's on that slope, or a fit to the data. Also the styling in the title would be better if it conformed to MOS:CAPS and MOS:DASH. Dicklyon (talk) 21:16, 6 February 2014 (UTC)
The microprocessor chronology here on wikipedia lists this data for least eight newer cpu's, dated 2012-2014.
I tried unsuccessfully to update the hard disk drive chart a little further down in this article with 2014 popular models sold at Newegg (in GB): 300 500 1000 2000 3000 4000. Any suggestions? 71.128.35.13 (talk) 03:41, 21 May 2014 (UTC)

This whole article, but this graph in particular, is so lacking in any mathematical rigour as to beggar belief. There's not even the equation for Moore's law: f(x)=2^(x/2) which this graph patently does not follow. It might be called a best fit logarithmic line, but there's not even any sourcing to show if the data are correct.Myersdtm (talk) 07:26, 11 September 2014 (UTC)--Myersdtm (talk) 07:26, 11 September 2014 (UTC)

The graph has a two-year doubling time, verified by (LN(2) / LN((1.2E9 transistors / 2000 transistors)^(1 / (2010 – 1971)))). As for sourcing, the microprocessor chronology and transistor count articles have references that could be used in this graph.
Even so, I agree that the graph and the article lack rigor. The graph is not updated after 2011; there's no indication that these transistors were minimum cost,[8] nor that these MPUs were produced in high volume; transistor count may not represent performance; and “dark silicon” presently limits thermally the number of usable transistors.[9][10] 71.128.35.13 (talk) 23:35, 11 September 2014 (UTC)
There is a reason the graph stops there, the data after 2010 no longer supports Moore's law. MvH (talk) 20:46, 3 April 2015 (UTC)MvH

I produced this graph and in light of the discussion here will update it. I used the microprocessor list from Transistor count as a source for the current image. Does anyone have a suggestion as to what raw data I should use for the update? As far as the line on the graph goes, the slope is that of the trasistor count doubling every two years, which is consistant with the definition of Moore's Law in this article's introduction. I felt it was better to include the Moore's Law line rather than that of the best fit to the data, which would constitue original research. The only variable in that case is the position of the line on the graph: its intercept rather than its slope. I chose to position it as if it was a best fit to the data to facilitate a rough visual comparison. Do let me know if anyone has an other idea about how to position it. This discussion may be more appropriate on the talk page of the image as it is also used on the Transistor count article. However, probably more people will see the discussion if it is here. -- Wgsimon (talk) 12:55, 12 September 2014 (UTC)

I suggest u position the line at the 1975 date and transistor count when Moore restated his observation and so label it so as to avoid misunderstanding (something like "Doubling of transistor count from Moore's 1975 observation"). For consistency u should continue to use transistor count from microprocessors. You might want to consider adding transistor counts from Flash Memory as a second set of data points starting circa 2000. Finding reliable sources for the additional data shouldn't be hard, you'll probably have to make a choice between several apparently reliable sources so be careful :-) Tom94022 (talk) 19:36, 12 September 2014 (UTC)
Updating post 2011 with Transistor count and/or Microprocessor chronology raw data would be a step forward. The slope (two year doubling time) and position seem reasonable. Changing the 1971 start date when MPUs originated or the title of the graph to conform to a particular prediction (1965, 1975 or 1979) isn't warranted. The focus ought to stay on MPUs: adding flash memory, DRAM or GPU data would complicate the graph. 71.128.35.13 (talk) 20:48, 13 September 2014 (UTC)
(Personal attack removed) 71.128.35.13 (talk) 01:22, 16 September 2014 (UTC)
Would an updated graph still show exponential improvement, or, did that basically stop around 2012? I've been looking at the fastest laptops available over the last 2 years, and there's hardly any speedup, let alone an exponential one. MvH (talk) 14:55, 28 September 2015 (UTC)MvH
PS. For clarity, I'm referring to CPU speed. Obviously a machine with an SSD will feel much faster than one with a regular hard drive, even with the same CPU speed. MvH (talk) 15:00, 28 September 2015 (UTC)MvH

Misplaced

I removed this paragraph from the middle of Moore's law § Other formulations and similar observations. It was a non sequitur there. I'm not sure where it belongs. ~Kvng (talk) 14:18, 13 September 2016 (UTC)

Moore (2003) credits chemical mechanical planarization (chip smoothing) with increasing the connectivity of microprocessors from two or three metal layers in the early 1990s to seven in 2003.[1] This progressed to nine metal layers in 2007 and thirteen in 2014.[2][3][4] Connectivity improves performance, and relieves network congestion. Just as additional floors may not enlarge a building's footprint, nor is connectivity tallied in transistor count. Microprocessors rely more on communications (interconnect) than do DRAM chips, which have three or four metal layers.[5][6][7] Microprocessor prices in the late 1990s improved faster than DRAM prices.[8]

References

  1. ^ Cite error: The named reference Moore_2003 was invoked but never defined (see the help page).
  2. ^ James, Dick. "Intel’s 14-nm Parts are Finally Here!", chipworks.com, October 27, 2014. Retrieved on November 5, 2014.
  3. ^ Bohr, Mark (2009). "The New Era of Scaling in an SoC World" (PDF). UCSD. Intel. Retrieved 2014-06-04.
  4. ^ Bohr, Mark (2012). "Silicon Technology Leadership for the Mobility Era" (PDF). Intel Corporation. Retrieved 2014-06-04.
  5. ^ Saraswat, Krishna (2002). "Scaling of Interconnections (course notes)" (PDF). Stanford University. Retrieved 2014-06-04. Memories ... don't need too many interconnects. Logic chips are more irregular and are dominated by communication requirements...generally have larger number of interconnects and thus need more levels of them.
  6. ^ Bruce Jacob, Spencer Ng, David Wang. "Memory systems: cache, DRAM, disk". 2007. Section 8.10.2. "Comparison of DRAM-optimized process versus a logic-optimized process". Page 376. [1]
  7. ^ Young Choi. "Battle commences in 50nm DRAM arena". 2009. [2]
  8. ^ Cite error: The named reference Aizcorbe01 was invoked but never defined (see the help page).

Single Electron Transistors

The article mentions single electron transistors (SETs) but gets a lot wrong, or at least is very misleading.

1) SETs have been in development since the 1970's. They were not announced for the first time in 2011.

2) This term "microscopic computer" in nonsense. If it means processors, they are already microscopic. If it means an actual computer a) What does THAT even mean? b) SETs won't help you get there. SETs, while quite small, aren't drastically smaller than conventional FETs used in ordinary every-day computer processors.

3) SETs are terrible logic devices, they aren't going to be replacing CMOS for general computing, ever. Nobody actually thinks they will. — Preceding unsigned comment added by 129.74.159.83 (talk) 14:08, 15 May 2018 (UTC)

End of Moore's law: a reliable source

Here's a NYT article exploring the end of Moore's law. Perfect for us if someone wants to put it in:

Markoff, John (September 26, 2015). "Smaller, Faster, Cheaper, Over: The Future of Computer Chips". The New York Times.

No time to do this myself ---- CharlesGillingham (talk) 19:37, 27 September 2015 (UTC)

Would like to add: the formulation in the article on Moore's law that "some" (referring to Lloyd) see the end of Moore's law in the far future is misleading. The arXiv version of his article states near the bottom of page 17 that Moore's law holding up till the computing capacities of the "ultimate laptop" are reached seems "highly unlikely". But perhaps one should let this misleading information stand, so that transhumanists can hang themselves. 84.169.249.66 (talk) 19:36, 11 February 2016 (UTC)

The PassMark CPU Benchmark website (www.cpubenchmark.net/singleThread.html) shows that Single Thread Performance has increased about 4.38% per year over the last 4 years. If this rate could be sustained then the Single Thread Performance doubles in 16 years. MvH (talk) 13:45, 5 June 2018 (UTC)

Stanford Neurocore project misrepresented

This claim (penultimate item under Future Trends), is false:


"In 2014, bioengineers at Stanford University developed a circuit modeled on the human brain. Sixteen "Neurocore" chips simulate one million neurons and billions of synaptic connections, claimed to be 9,000 times faster as well as more energy efficient than a typical PC." (emphasis mine)


The problem lies in the cited source. The first sentence of the three-sentence italicized intro contains the erroneous claim:


"Stanford bioengineers have developed faster, more energy-efficient microchips based on the human brain – 9,000 times faster and using significantly less power than a typical PC."


This intro claim appears to result from a misreading of the second and third sentences of the main text:


"For all their sophistication, computers pale in comparison to the brain. The modest cortex of the mouse, for instance, operates 9,000 times faster than a personal computer simulation of its functions."


This sentence is not comparing the speed of the Stanford circuit to the speed of a PC. It is comparing the speed of the cortex of a mouse (nb: the rodent, not the computer pointing device) to the speed of a PC simulation of its cortical functions.

Nothing else in the article supports the claim that the Stanford circuit is faster than a typical PC (let alone 9,000 times faster). Do note, however, that a claim about the efficiency of the circuit is made in the penultimate paragraph:


"Boahen notes that Neurogrid is about 100,000 times more energy efficient than a personal computer simulation of 1 million neurons."

--Timrooks (talk) 15:35, 3 July 2018 (UTC)

Good catch. The claim "For all their sophistication, computers pale in comparison to the brain" is especially deceptive. If I weld a teaspoon to the front of a bulldozer, thus simulating a human digging a hole with a teaspoon, then find that the human digs a lot faster, would it be fair to say that "for all their power, bulldozers pale in comparison to humans at moving dirt"? --Guy Macon (talk) 15:49, 3 July 2018 (UTC)
...Or I could simply work the comparison backwards. Instead of having the PC simulate a brain, have the brain simulate a PC. you could have a bunch of books with pages full of squares -- as many as there are bits on a hard drive -- and give the human a pencil and an eraser, then have him put ones or zeros in the squares to simulate writing to the hard disk. Add another set of books for the RAM, a whiteboard for the CPU registers, and a rulebook telling the human how to interpret each machine language instruction, and the human can -- slowly -- simulate the PC. Now time him as he boots Windows 10 and then write a paper about how much more power-efficient a PC is. --Guy Macon (talk) 16:12, 3 July 2018 (UTC)

The Kurzweil-sourced graph "120 Years of Moore's Law"

This graph has nowhere near enough relevance to Moore's Law to justify its presence in this article. The primary and sufficient argument here is that the Z axis depicts "calculations per second per constant dollar" which is not congruent to "devices per integrated circuit" in any useful way. The secondary and independently sufficient argument is that the points represent systems, not integrated circuits. The tertiary and also independently sufficient argument is that for 73 of the 120 years depicted, the systems in question don't even use integrated circuits.

Finally, the article makes no mention of this graph, and no statement in the article is usefully illustrated by it. I have removed the graph from the article. 98.247.224.9 (talk) 10:31, 13 December 2017 (UTC)

The article used to refer to Super Moore's Law, which would be relevant to the graph. Why was that deleted? cheers, Michael C. Price talk 13:44, 18 July 2018 (UTC)

A Commons file used on this page has been nominated for deletion

The following Wikimedia Commons file used on this page has been nominated for deletion:

Participate in the deletion discussion at the nomination page. —Community Tech bot (talk) 22:36, 9 December 2018 (UTC)

Moore's Revenge

Pesce, Mark (4 Jun 2018). "'Moore's Revenge' is upon us and will make the world weird". The Register. Retrieved 27 May 2019.

  • IT support consequences
  • Press driven cultural change

Conrad T. Pino (talk) 19:41, 27 May 2019 (UTC)

Taiwanese innovation

The recent addition to the lead claiming a Taiwanese "technological breakthrough" is unsuitable for the article (particularly the lead). I refer to 885335898, 885299445 and associated minor edits.

  • The "breakthrough" refers to a new method for creating diodes. While it may lead to reductions in circuit size, the method is in a 'proof of concept' phase and only just published. It is not used in any production chips and so it is too early to gauge its impact on Moore's law. The citations are original research and press releases about the research.
  • The focus on Taiwan violates neutral point-of-view and may be a case of disruptive editing. It is too overly political for a computing article.
  • The section has already been reverted by multiple people. I think this establishes consensus that the section does not belong here.

--Quantum7 15:41, 27 February 2019 (UTC)

The IP has been reported for edit warring. The lede already has a sentence regarding TMC which is supported by RS in the article, namely:
"In the late 2010s, only two semiconductor manufacturers have been able to produce semiconductor nodes that keep pace with Moore's law, TSMC and Samsung Electronics, with 10 nm, 7 nm and 5 nm nodes in production (and plans for 3 nm nodes), whereas the pace has slowed down for Intel and other semiconductor manufacturers."
The sentence could be better word smithed and maybe some of the IPs references added but that's about it. None of the IPs refs support the claim that TSMC is beating ML, just that TSMC is now ahead of Intel in device geometry. Tom94022 (talk) 17:15, 6 August 2019 (UTC)

Proposed merge with Neven's law

new and not well sourced, so seems better to merge to Moore's law#Other formulations and similar observations Widefox; talk 08:54, 25 October 2019 (UTC)

I support the merge. WP:TOOSOON for a standalone article at present. --mikeu talk 15:48, 25 October 2019 (UTC)
"Neven's law" is a bit of frippery — one guy's sales talk, propagated by fluffy "science journalism". It was made into a redirect back in June for good reason. It's not a scientific hypothesis; it's a meme. XOR'easter (talk) 15:53, 25 October 2019 (UTC)
I should have looked closer. What appears to be multiple mentions is just reprints or quotes of a single Quanta Magazine article. The republications of Quanta are rather copious... --mikeu talk 16:45, 25 October 2019 (UTC)
Yes, Quanta has a deal of some sort to republish regularly in Wired and probably elsewhere. XOR'easter (talk) 17:32, 25 October 2019 (UTC)

I disagree. The whole point of Neven's law is that the processing power of a quantum computer increases at a double-exponential rate, whereas the whole point of Moore's law is that the processing power of a classical computer increases at only an exponential rate. Double-exponential growth has never been seen before in any physical system. Quantumpundit (talk) 01:13, 26 October 2019 (UTC)Quantumpundit

User:Quantumpundit please see WP:COI and WP:DISCLOSE in all places where you may have a COI. Widefox; talk 11:43, 29 October 2019 (UTC)
It's still a "similar observation". Distinct, but similar. (See also Double_exponential_function#Applications.) XOR'easter (talk) 15:55, 26 October 2019 (UTC)

Widefox redirected Neven's Law to Hartmut Neven. The article before the redirect showed coverage in WP:THREE reliable sources. I have restored it. ~Kvng (talk) 22:03, 1 November 2019 (UTC)

User:Kvng wrong redirect - see edit history of Neven's Law - it was merged by me to the correct spelling Neven's law [11], and then Neven's law has been merged by other editors so take it up there - Neven's Law will only ever be a redirect, just like Moore's Law. I reverted back to a redirect tagged as miscapitalisation. Widefox; talk 01:33, 2 November 2019 (UTC)
Redirect to Hartmut Neven and be done with it. WP:TOOSOON, WP:NEO. Barely worth a mention in an article; certainly not worth its own article or with merging here.--Srleffler (talk) 01:02, 2 November 2019 (UTC)
Whatever the merits, it is Neven's law not Neven's Law. The sourcing issues detailed by others do look severe, maybe borderline WP:TOOSOON, and there clearly was promo editing to avoid per WP:NEO. Widefox; talk 01:33, 2 November 2019 (UTC)

Sorry about twiddling the wrong title. I don't see a consensus to redirect this topic. Supporters include Quantumpundit, Chessrat, Lumos3 and me. Detractors include XOR'easter, Widefox, Srleffler and, recently, mikeu. I see that the Quanta article was republished by Scientific American but together that still counts as one reliable source. What's the problem with [12] or [13]? ~Kvng (talk) 13:27, 2 November 2019 (UTC)

Note that Quantumpundit may have a WP:COI issue here. I don't see any sign that chessrat or lumos3 have expressed their thoughts in this discussion. --Srleffler (talk) 13:42, 2 November 2019 (UTC)
Interestingly, the second article you link to is arguing that declaring "Neven's Law" is premature, and that Neven based his claim on too little data over too short a span of time. That's exactly the problem with the article here. The whole concept is a bit of fluff and self-promotion based on very, very little. If this pattern of growth held for years, it would be worth declaring it a "law" and mentioning it on Wikipedia, but it's only been a few months. It's just too soon.--Srleffler (talk) 13:50, 2 November 2019 (UTC)
The article in Interesting Engineering is not reliable. Example: Exotic subatomic particles behave in ways that are hard to accept. They can blow right through foundational laws of physics without missing a step, as quantum entanglement does when paired particles communicate instantaneously with each other even if they are on opposite sides of the universe. First, obeying quantum mechanics is not violating "foundational laws of physics" — that's an absurd and self-contradictory statement. Second, that's not how entanglement works: a basic fact about entanglement is that it cannot by itself be used for signalling. It also gets wrong the history of quantum computing prior to Shor. The proposals in those early days — Benioff and Feynman come to mind — did have their practical purposes in mind. Feynman's big point was that a quantum computer would be the right way to do simulations of physically and chemically interesting systems that classical computers could not manage efficiently. That is literally the same application that people invoke today and that Hartmut Neven has himself published on [14]. The topic of quantum computation is deluged with terrible journalism, but we should not propagate it. XOR'easter (talk) 14:48, 2 November 2019 (UTC)
Chessrat (talk · contribs) and Lumos3 (talk · contribs) are active editors who contributed significant material to the original articles.
WP:TOOSOON is about not having enough coverage yet. It is not about whether there's scientific consensus. We can certainly mention in the article that some think this law is premature.
If we take Interesting Engineering off the table, it looks like we still have coverage in multiple reliable sources. There's actually a lot of coverage on this. Here's a few more to try to shoot down: [15], [16], [17], [18], [19], [20]. ~Kvng (talk) 16:30, 2 November 2019 (UTC)
Churnalism. "NextBigFuture" is a blog. So is "ScienceHook" (and that blog post was written by a "second-year computer science undergraduate" to boot). HPCwire just recapitulates the Quanta story. The Next Web is a tech website that also gets all its details on this from Quanta. The same appears to be true of Analytics India Magazine (for which Wikipedia doesn't have an article). Live Science does its own reporting sometimes, but here, it's just another echo of the Quanta story. Applying the slightest bit of critical scrutiny to any of the "sources" on this reveals that it's a catchy bit of sales talk that had a flash-in-the-pan moment. Mention it in Hartmut Neven, maybe, or in a subsection of Moore's law perhaps, but that's all it deserves. XOR'easter (talk) 16:49, 2 November 2019 (UTC)
(ec x2) I've changed the two redirects to target the section in the BLP. User:Kvng there seems clear edit consensus that this was a NEO months ago, and these sources are all too close to the event WP:PRIMARYNEWS / have quality issues detailed above, so could be discounted as pure WP:CHURNALISM "Editors need to show critical judgment when reviewing sources in the business press. It's worth the few extra minutes to compare a contentious or implausible article ". How statistically valid is a double exponential fit to so few data points in close proximity? This fails critical judgement at this time. What does the data for Moore's law look like in the initial timescale? With all that in mind, it seems there's no consensus for creating an article. I believe that's the fair appraisal, and one in which I'm sitting on the fence but leaning towards lack of notability currently. While we're here {{Computer laws}} includes the term, is that undue weight? Widefox; talk 17:06, 2 November 2019 (UTC)
Yes, I'd call that undue weight. XOR'easter (talk) 17:22, 2 November 2019 (UTC)
(ec) ...and WP:RECENTISM. Removed from template. Widefox; talk 17:30, 2 November 2019 (UTC)

GA Review

GA toolbox
Reviewing
This review is transcluded from Talk:Moore's law/GA3. The edit link for this section can be used to add comments to the review.

Reviewer: Lee Vilenski (talk · contribs) 18:45, 1 May 2020 (UTC)


Hello, I am planning on reviewing this article for GA Status, over the next couple of days. Thank you for nominating the article for GA status. I hope I will learn some new information, and that my feedback is helpful.

If nominators or editors could refrain from updating the particular section that I am updating until it is complete, I would appreciate it to remove a edit conflict. Please address concerns in the section that has been completed above (If I've raised concerns up to references, feel free to comment on things like the lede.)

I generally provide an overview of things I read through the article on a first glance. Then do a thorough sweep of the article after the feedback is addressed. After this, I will present the pass/failure. I will use strikethrough tags when concerns are met. Even if something is obvious why my concern is met, please leave a message as courtesy.

Best of luck! you can also use the {{done}} tag to state when something is addressed. Lee Vilenski (talk • contribs)

Please let me know after the review is done, if you were happy with the review! Obviously this is regarding the article's quality, however, I want to be happy and civil to all, so let me know if I have done a good job, regardless of the article's outcome.

Immediate Failures

  • It is a long way from meeting any one of the six good article criteria -
  • It contains copyright infringements -
  • It has, or needs, cleanup banners that are unquestionably still valid. These include{{cleanup}}, {{POV}}, {{unreferenced}} or large numbers of {{citation needed}}, {{clarify}}, or similar tags. (See also {{QF-tags}}). -
  • It is not stable due to edit warring on the page. -

Prose

Lede

General

GA Review

GA review (see here for what the criteria are, and here for what they are not)
  1. It is reasonably well written.
    a (prose, spelling, and grammar): b (MoS for lead, layout, word choice, fiction, and lists):
  2. It is factually accurate and verifiable.
    a (reference section): b (citations to reliable sources): c (OR): d (copyvio and plagiarism):
  3. It is broad in its coverage.
    a (major aspects): b (focused):
  4. It follows the neutral point of view policy.
    Fair representation without bias:
  5. It is stable.
    No edit wars, etc.:
  6. It is illustrated by images and other media, where possible and appropriate.
    a (images are tagged and non-free content have fair use rationales): b (appropriate use with suitable captions):
  7. Overall:
    Pass/Fail:

Comments

- I didn't think NODEADLINE applies at GAN? Wikipedia:Good article nominations/Instructions talks about it being usually 7 days. I'm quite happy to keep this open longer, but I'd like to know it was being worked on and not gone stale. Best Wishes, Lee Vilenski (talk • contribs) 17:05, 31 May 2020 (UTC)

Yes probably not and that's probably why I don't do much GA stuff. I'll touch this page whenever I work on your flagged GA issues. If you don't see satisfactory progress, I won't be offended if you shut this down. ~Kvng (talk) 14:30, 3 June 2020 (UTC)

Lee Vilenski, I beleive I've addressed all your General concerns now. Please review my comments above and let me know if there is any further work needed to wrap this up. ~Kvng (talk) 20:31, 11 June 2020 (UTC) Lee Vilenski thanks for approving GA for this article. Do you know how to sort out the GA templates at Talk:Moore's law? I think I could fix it up manually but the GA process seems to be bot driven and so maybe there's a better way than manual. ~Kvng (talk) 13:16, 15 June 2020 (UTC)

Legobot usually fixes the "article milestones" bit after a couple days. If not, drop me a ping and I'll do it manually. You may have gotten a "failed" message on your talk (I didn't check), it's a known bug. Best Wishes, Lee Vilenski (talk • contribs) 13:34, 15 June 2020 (UTC)
@Lee Vilenski: Legobot last visited three days ago. This may need some intervention. ~Kvng (talk) 14:33, 18 June 2020 (UTC)
@Lee Vilenski: this did not sort itself out so I made some manual adjustments. Feel free to touch it up if necessary. ~Kvng (talk) 17:29, 13 November 2020 (UTC)

Is "Moorepsilas law" related?

If yes, what is the meaning of that psilas? --Palosirkka (talk) 09:25, 29 October 2020 (UTC)

Palosirkka, that's new to me. I see the term in a few papers and these seem to be referring to Moore's law or at least something similar. ~Kvng (talk) 14:05, 10 November 2020 (UTC)

Biased/unacademic language in the introduction

"Unfortunately, this runaway "growth" leads to environmental impacts, such as Americans throwing 400,000 cell phones out every day, as technology industries use planned obsolescence as a profit centre, with little regard for the Limits to Growth. "

While this is true, I believe it could be rewritten in a more academic way. — Preceding unsigned comment added by 83.99.83.80 (talk) 18:05, 7 February 2021 (UTC)

This article generally uses what one could characterize as unacademic language, and it surprises me that it has been designated as a "good article". —GreenWeasel11 (talk) 05:10, 29 March 2021 (UTC)

Density?

I have read several peopler which define Moore's law as being the density of transistors including a work by MIT. Is that part of the actual Law or is that something added in after the fact?--174.99.238.22 (talk) 03:37, 8 November 2020 (UTC)

Moore's law is about increasing the number of transistors on a die. Increasing the number of transistors on a die is about increasing density and increasing the size of the die. Both are technically challenging. ~Kvng (talk) 13:59, 10 November 2020 (UTC)
Thanks. The reason I asked is the way some people were defining Moore's law effectively resulted in 'make a larger chipand ignore density (one went as far as to say that, in fact).--174.99.238.22 (talk) 17:55, 13 November 2020 (UTC)
Remember, Moore's law is an observation and the observed advancements in IC technology thus far have in been both size and density. I believe density has been the larger driver. ~Kvng (talk) 14:28, 16 November 2020 (UTC)
Both density and chip size are limited by defect density, which has gotten better and Moore knew that early. It seems that there some question about which exponent should be used, and who said it. It seems that Moore left that open, between 1 and 2 years. I suspect someone average that to 18 months. The exact value depends on what you are measuring. Gah4 (talk) 21:46, 7 February 2022 (UTC)

scaling laws

The article mentions Dennard a lot, and for good reason. There are, though, other scaling laws. Dennard scaling shrinks the gate oxide in proportion to the other dimensions, which shrinks the supply voltage. Sometimes die shrink is done without changing the oxide thickness, and keeping the supply voltage the same. (Convenient for users.) Also, Dennard only applies to transistors, not to wiring, where wiring gets increasingly important at higher integration levels. And also, as noted, leakage (tunneling) though thin oxide is limiting Dennard scaling. I am not sure which scaling law they now use. Gah4 (talk) 21:50, 7 February 2022 (UTC)

importance rating

@Kvng: Thanks for helping to do some of the importance assessments. I've basically been trying to prune/edit Category:Top-importance Computer science articles, which was in a pretty inconsistent state and needs some fixing up, and I welcome any help on making the cuts as to what is included! We may also want to start a general discussion at Wikipedia talk:WikiProject Computer science on this.

Any decision about what to include is bound to get a bit political, but just wanted to explain my thinking. Regarding "Mid seems right, primarily a computer hardware topic," I don't agree at all. Moore's law is frequently cited in general-audience CS talks as a primary motivation for research -- it's one of the most relevant topics in computer architecture and as a non-architecture researcher, one of the few things I know about.

Caleb Stanford (talk) 19:31, 23 June 2022 (UTC)

I'm happy to leave it High ~Kvng (talk) 20:20, 23 June 2022 (UTC)

2022 Inventor of the Year: Tahir Ghani Keeps Moore’s Law Alive

Here is Intel's Newsroom article about 2022 Inventor of the Year: Tahir Ghani Keeps Moore’s Law Alive. Rjluna2 (talk) 16:31, 17 August 2022 (UTC)

Conflicts of interest and misleading headlines

Hey folks, I just had to decimate the "forecasts" section, which was quoting stuff from 2005 ("we could have billions of transistors someday!!"). Some of it was about justifying quantum computing hype. I'm not sure what the best thing to replace it with would be; it is trying to predict the future after all. I think there are a few reasons why people are incentivized to be pessimistic of optimistic about the future of Moore's law. Hamishtodd1 (talk) 10:47, 28 November 2022 (UTC)

Intel Research Fuels Moore’s Law and Paves the Way to a Trillion Transistors by 2030

Here is the Intel's Newsroom at Intel Research Fuels Moore’s Law and Paves the Way to a Trillion Transistors by 2030. Rjluna2 (talk) 16:49, 5 December 2022 (UTC)

There are way too many links. If someone is on this page, they know what Observingis. It makes it difficult to read this page without getting distracted (especially with ADHD). Weirdojo1 (talk) 22:01, 10 March 2023 (UTC)

Retrieved from "https://en.wikipedia.org/w/index.php?title=Talk:Moore%27s_law/Archive_5&oldid=1230565674"