AI Impacts Wiki ai_timelines:hardware_and_ai_timelines

2019 recent trends in Geekbench score per CPU price

2022-09-21T07:37:41+00:00

@@ -1 +1,81 @@
+ ====== 2019 recent trends in Geekbench score per CPU price ======
+ 
+ // Published 14 April, 2020; last updated 26 October, 2020 //
+ 
+ <HTML>
+ <p>From 2006 – 2020, Geekbench score per CPU price has grown by around 16% a year, for rates that would yield an order of magnitude over roughly 16 years.</p>
+ </HTML>
+ 
+ 
+ ===== Details =====
+ 
+ 
+ <HTML>
+ <p>We looked at Geekbench 5,<span class="easy-footnote-margin-adjust" id="easy-footnote-1-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-1-2413" title='“Introducing Geekbench 5.” Geekbench 5 &amp;#8211; Cross-Platform Benchmark. Accessed April 2, 2020. &lt;a href="https://www.geekbench.com/"&gt;https://www.geekbench.com/&lt;/a&gt;.'><sup>1</sup></a></span> a benchmark for CPU performance. We combined Geekbench’s multi-core scores on its ‘Processor Benchmarks’ page<span class="easy-footnote-margin-adjust" id="easy-footnote-2-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-2-2413" title='“Processor Benchmarks.” Processor Benchmarks &amp;#8211; Geekbench Browser. Accessed April 14, 2020. &lt;a href="https://browser.geekbench.com/processor-benchmarks"&gt;https://browser.geekbench.com/processor-benchmarks&lt;/a&gt;.'><sup>2</sup></a></span> with release dates and prices that we scraped from Wikichip and Wikipedia.<span class="easy-footnote-margin-adjust" id="easy-footnote-3-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-3-2413" title='Starting with Geekbench’s list of CPUs, we Googled ‘&amp;lt;CPU&amp;gt; Wikichip’ and ‘&amp;lt;CPU&amp;gt; Wikipedia’ to find lists of processor release dates and prices. We then copied Wikichip tables into &lt;a href="https://docs.google.com/spreadsheets/d/1xP2ndDQYfrtC4IQQ35ndsf-Wam1078GC9wzFONE8jAk/edit?usp=sharing"&gt;this spreadsheet&lt;/a&gt;, tab ‘Wikichip / Wikipedia Information’, and used &lt;a href="https://drive.google.com/open?id=1Wn6aIo0qbVLrzI8sQRB9qRO6pyhQhmHf"&gt;this script&lt;/a&gt; to parse CPU data from tables in individual Wikipedia pages before copying them into the same spreadsheet.'><sup>3</sup></a></span> All our data and plots can be found <a href="https://docs.google.com/spreadsheets/d/1xP2ndDQYfrtC4IQQ35ndsf-Wam1078GC9wzFONE8jAk/edit?usp=sharing">here</a>.<span class="easy-footnote-margin-adjust" id="easy-footnote-4-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-4-2413" title="The ‘Geekbench Scores’ tab lists all the Geekbench CPU scores, while ‘Wikichip / Wikipedia Information’ stores all our scraped release dates and prices."><sup>4</sup></a></span> We then calculated score per dollar and adjusted for inflation using the consumer price index.<span class="easy-footnote-margin-adjust" id="easy-footnote-5-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-5-2413" title="“CPI Home.” U.S. Bureau of Labor Statistics. U.S. Bureau of Labor Statistics. Accessed April 14, 2020. https://www.bls.gov/cpi/."><sup>5</sup></a></span> For every year, we calculated the 95th percentile score per dollar. We then fit linear and exponential trendlines to those scores.<br/></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>Figure 1 shows all our data for Geekbench score per CPU price.<br/></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <figure class="wp-block-image is-resized">
+ <img alt="" height="359" src="https://lh3.googleusercontent.com/MQPtCMAM8oJRHbvS9UnCplssaHIG-JV8nLt0B1eoenHs_nBSs1dSremaUUmP41U8WX1Z296jG90O3TmEYG-DoIr-0RWoun-5VGjhKwmafK0SeIuyEdjZ93fAhbwpSa0QaqcQB5mr" width="581"/>
+ <figcaption>
+                   Figure 1: Geekbench scores per CPU price, in 2019 dollars. Red dots denote the 95th percentile values in each year from 2006 – 2019 (we start at 2006 since we have &lt;= 2 data points a year prior to then). The exponential trendline through the 95th percentiles is marked in red, while the linear trendline is marked in green. The vertical axis is log-scale.
+                 </figcaption>
+ </figure>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>The data is well-described by a linear or an exponential trendline. Assuming an exponential trend,<span class="easy-footnote-margin-adjust" id="easy-footnote-6-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-6-2413" title="Where ambiguous, we assume these trends are exponential rather than linear, because our understanding is that that is much more common historically in computing hardware price trends."><sup>6</sup></a></span> Geekbench score per CPU price grew by around 16% per year between 2006 and 2020, a rate that would yield a factor of ten every 16 years.<span class="easy-footnote-margin-adjust" id="easy-footnote-7-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-7-2413" title='See &lt;a href="https://docs.google.com/spreadsheets/d/1xP2ndDQYfrtC4IQQ35ndsf-Wam1078GC9wzFONE8jAk/edit?usp=sharing"&gt;this spreadsheet&lt;/a&gt;, sheet ‘Geekbench Scores’ for our calculations, which are next to the cell marked ‘Exponential trendline from 2006 &amp;#8211; now’.'><sup>7</sup></a></span></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>This is a markedly slower growth rate than those observed for <a href="/doku.php?id=ai_timelines:trends_in_the_cost_of_computing">CPU price performance trends</a> in the past, however since it is for a different performance metric to any used earlier, it is unclear how similar one should expect them to be– from 1940 to 2008, <a href="/doku.php?id=ai_timelines:trends_in_the_cost_of_computing">Sandberg and Bostrom found</a> that CPU price performance grew by a factor of ten every 5.6 years when measured in MIPS per dollar, and by a factor of ten every 7.7 years when measured in FLOPS per dollar.<span class="easy-footnote-margin-adjust" id="easy-footnote-8-2413"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-8-2413" title='See our &lt;a href="https://aiimpacts.org/trends-in-the-cost-of-computing/"&gt;Trends in the cost of computing&lt;/a&gt; page, section ‘Sandberg and Bostrom’.'><sup>8</sup></a></span></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p><em>Primary author: Asya Bergal</em></p>
+ </HTML>
+ 
+ 
+ ===== Notes =====
+ 
+ 
+ <HTML>
+ <ol class="easy-footnotes-wrapper">
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-1-2413"></span>“Introducing Geekbench 5.” Geekbench 5 – Cross-Platform Benchmark. Accessed April 2, 2020. <a href="https://www.geekbench.com/">https://www.geekbench.com/</a>.<a class="easy-footnote-to-top" href="#easy-footnote-1-2413"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-2-2413"></span>“Processor Benchmarks.” Processor Benchmarks – Geekbench Browser. Accessed April 14, 2020. <a href="https://browser.geekbench.com/processor-benchmarks">https://browser.geekbench.com/processor-benchmarks</a>.<a class="easy-footnote-to-top" href="#easy-footnote-2-2413"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-3-2413"></span>Starting with Geekbench’s list of CPUs, we Googled ‘&lt;CPU&gt; Wikichip’ and ‘&lt;CPU&gt; Wikipedia’ to find lists of processor release dates and prices. We then copied Wikichip tables into <a href="https://docs.google.com/spreadsheets/d/1xP2ndDQYfrtC4IQQ35ndsf-Wam1078GC9wzFONE8jAk/edit?usp=sharing">this spreadsheet</a>, tab ‘Wikichip / Wikipedia Information’, and used <a href="https://drive.google.com/open?id=1Wn6aIo0qbVLrzI8sQRB9qRO6pyhQhmHf">this script</a> to parse CPU data from tables in individual Wikipedia pages before copying them into the same spreadsheet.<a class="easy-footnote-to-top" href="#easy-footnote-3-2413"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-4-2413"></span>The ‘Geekbench Scores’ tab lists all the Geekbench CPU scores, while ‘Wikichip / Wikipedia Information’ stores all our scraped release dates and prices.<a class="easy-footnote-to-top" href="#easy-footnote-4-2413"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-5-2413"></span>“CPI Home.” U.S. Bureau of Labor Statistics. U.S. Bureau of Labor Statistics. Accessed April 14, 2020. https://www.bls.gov/cpi/.<a class="easy-footnote-to-top" href="#easy-footnote-5-2413"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-6-2413"></span>Where ambiguous, we assume these trends are exponential rather than linear, because our understanding is that that is much more common historically in computing hardware price trends.<a class="easy-footnote-to-top" href="#easy-footnote-6-2413"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-7-2413"></span>See <a href="https://docs.google.com/spreadsheets/d/1xP2ndDQYfrtC4IQQ35ndsf-Wam1078GC9wzFONE8jAk/edit?usp=sharing">this spreadsheet</a>, sheet ‘Geekbench Scores’ for our calculations, which are next to the cell marked ‘Exponential trendline from 2006 – now’.<a class="easy-footnote-to-top" href="#easy-footnote-7-2413"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-8-2413"></span>See our <a href="/doku.php?id=ai_timelines:trends_in_the_cost_of_computing">Trends in the cost of computing</a> page, section ‘Sandberg and Bostrom’.<a class="easy-footnote-to-top" href="#easy-footnote-8-2413"></a>
+ </div></li>
+ </ol>
+ </HTML>
+ 
+

How much computing capacity exists in GPUs and TPUs in Q1 2023?

2024-12-10T21:33:43+00:00

@@ -4,8 +4,10 @@
  */
  ====== How much computing capacity exists in GPUs and TPUs in Q1 2023? ======
  
  //Published 3 April 2023, last updated 3 April 2023//
+ 
+ (**Dec 10 2024 Update: **This analysis did not consider typical versus maximum performance in computing hardware. The data and figures presented here are likely based on maximum performance.)
  
  A back-of-the-envelope calculation based on market size, price-performance, hardware lifespan estimates, and the sizes of Google’s data centers estimates that there is around 3.98 * 10^21 FLOP/s of computing capacity on GPUs and TPUs as of Q1 2023.
  
  ===== Details =====

Factors that affect the price of GPUs

2023-07-03T15:36:57+00:00

@@ -1,9 +1,9 @@
  ====== Factors that affect the price of GPUs ======
  
  //This page is a work in progress//
  
- //Published 1 July, 2023//
+ //Published 1 July 2023, last updated 3 July 2023//
  
  GPU prices are affected by factors of demand such as the popularity of PC gaming, the profitability of cryptomining, and AI development, factors of supply such as manufacturing limitations, and government regulations.
  
  ===== Details =====

Nordhaus hardware price performance dataset

2022-09-21T07:37:41+00:00

@@ -1 +1,31 @@
+ ====== Nordhaus hardware price performance dataset ======
+ 
+ // Published 16 February, 2018 //
+ 
+ <HTML>
+ <p>This page contains the data from Appendix 2 of William Nordhaus’ <a href="https://web.archive.org/web/20160222082744/http://www.econ.yale.edu/~nordhaus/homepage/prog_083001a.pdf"><em>The progress of computing</em></a> in usable formats.</p>
+ </HTML>
+ 
+ 
+ ===== Notes =====
+ 
+ 
+ <HTML>
+ <p>This data was collected from Appendix 2 of <a href="https://web.archive.org/web/20160222082744/http://www.econ.yale.edu/~nordhaus/homepage/prog_083001a.pdf"><em>The progress of computing</em></a>, using <a href="http://tabula.technology/">Tabula</a> (a program for turning tables in pdfs into other table formats). We have not checked its accuracy beyond a graph of the resulting data looking visually similar to a graph of the original.</p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p><strong>Important:</strong> we <a href="/doku.php?id=ai_timelines:trends_in_the_cost_of_computing#Nordhaus">previously noted</a> that this data appears to be orders of magnitude different from other sources, and haven’t had time to look into this discrepancy.</p>
+ </HTML>
+ 
+ 
+ ===== Data =====
+ 
+ 
+ <HTML>
+ <p><a href="https://docs.google.com/spreadsheets/d/1b6v_AXGjIYEc22MNXWmv0AcbUH6oJnqMHThrFcqUXoo/edit?usp=sharing">Here</a> is a Google sheet of the data. See ‘Nordhaus via Tabula’ page.</p>
+ </HTML>
+ 
+

Price-performance trend in top supercomputers

2022-09-21T07:37:41+00:00

@@ -1 +1,37 @@
+ ====== Price-performance trend in top supercomputers ======
+ 
+ // Published 08 November, 2017; last updated 13 November, 2017 //
+ 
+ <HTML>
+ <p>A top supercomputer can perform a GFLOP for around $3, in 2017.</p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>The price of performance in top supercomputers continues to fall, as of 2016.</p>
+ </HTML>
+ 
+ 
+ ===== Details =====
+ 
+ 
+ <HTML>
+ <p><a href="https://www.top500.org/lists/2017/06/">TOP500.org</a> maintains a list of top supercomputers and their performance on the Linpack benchmark. The figure below is based on empirical performance figures (‘Rmax’) from <a href="https://www.top500.org/lists/2017/06/">Top500</a> and price figures collected from a variety of less credible sources, for nine of the ten highest performing supercomputers (we couldn’t find a price for the tenth). Our data and sources are <a href="https://docs.google.com/spreadsheets/d/1nV6djZI7csDv_ewElbNKiQZVl36ViGyUEy_MSqk2krI/edit?usp=sharing">here</a>.</p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>Sunway Teihu Light performs the cheapest GFLOPS, at $2.94/GFLOPS. This is around one hundred times more expensive than peak theoretical performance of certain GPUs, but we do not know why there is such a difference (peak performance is generally higher than actual performance, but by closer to a factor of two).</p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>There appears to be a downward trend in price, but it is not consistent, and with so few data points its slope is ambiguous. The best price for performance roughly halved in the last 4-5 years, for a 10x drop in 13-17 years. The K computer in 2011 was much more expensive, but appears to have been substantially more expensive than earlier computers.</p>
+ </HTML>
+ 
+ 
+ 
+ 
+ 
+

Trends in DRAM price per gigabyte

2022-09-21T07:37:41+00:00

@@ -1 +1,151 @@
+ ====== Trends in DRAM price per gigabyte ======
+ 
+ // Published 14 April, 2020; last updated 15 September, 2020 //
+ 
+ <HTML>
+ <p>The price of a gigabyte of DRAM has fallen by about a factor of ten every 5 years from 1957 to 2020. Since 2010, the price has fallen much more slowly, at a rate that would yield an order of magnitude over roughly 14 years.<br/></p>
+ </HTML>
+ 
+ 
+ 
+ ====== Details ======
+ 
+ 
+ ===== Background =====
+ 
+ 
+ <HTML>
+ <p>DRAM, “dynamic random-access memory”, is a type of semiconductor memory. It is used as the main memory in modern computers and graphic cards.<span class="easy-footnote-margin-adjust" id="easy-footnote-1-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-1-2408" title='From &lt;a href="https://en.wikipedia.org/wiki/Dynamic_random-access_memory"&gt;Wikipedia&lt;/a&gt;:&lt;br&gt;“Dynamic random-access memory (DRAM) is a type of random access semiconductor memory … One of the largest applications for DRAM is the main memory (colloquially called the &amp;#8220;RAM&amp;#8221;) in modern computers and graphics cards (where the &amp;#8220;main memory&amp;#8221; is called the graphics memory).”&lt;/p&gt; &lt;p&gt;“Dynamic Random-Access Memory.” Wikipedia. Wikimedia Foundation, March 24, 2020. https://en.wikipedia.org/wiki/Dynamic_random-access_memory.'><sup>1</sup></a></span></p>
+ </HTML>
+ 
+ 
+ ===== Data =====
+ 
+ 
+ <HTML>
+ <p>We found two sources for historic pricing of DRAM. One was a dataset of DRAM prices and sizes from 1957 to 2018 collected by technologist and retired Computer Science professor<span class="easy-footnote-margin-adjust" id="easy-footnote-2-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-2-2408" title='From &lt;a href="http://jcmit.net/"&gt;his personal website&lt;/a&gt;:&lt;br&gt;“1998-2001 Associate Professor, National University of Singapore, School of Computing &amp;#8211; Computer Science”.&lt;/p&gt; &lt;p&gt;“Biographical Information for Dr John C McCallum.” Biographical Information for Dr John C McCallum. Accessed April 14, 2020. http://jcmit.net/.'><sup>2</sup></a></span> John C. McCallum.<span class="easy-footnote-margin-adjust" id="easy-footnote-3-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-3-2408" title='“Historical Memory Prices 1957+.” Accessed April 9, 2020.&lt;a href="https://jcmit.net/memoryprice.htm"&gt; https://jcmit.net/memoryprice.htm&lt;/a&gt;.'><sup>3</sup></a></span> The other dataset was extracted from a graph generated by Objective Analysis,<span class="easy-footnote-margin-adjust" id="easy-footnote-4-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-4-2408" title='We extracted data from the third graph in &lt;a href="https://thememoryguy.com/dram-prices-hit-historic-low/"&gt;this blog post&lt;/a&gt; using the popular online tool &lt;a href="https://automeris.io/WebPlotDigitizer/"&gt;WebPlotDigitizer&lt;/a&gt; to extract data. The graph is attributed to Objective Analysis, and appears to have been generated by the writer of the post, Jim Handy. (He explicitly says he generated previous graphs in the post.)&lt;br&gt;&lt;br&gt;&amp;nbsp;“DRAM Prices Hit Historic Low – The Memory Guy.” Accessed April 9, 2020.&lt;a href="https://thememoryguy.com/dram-prices-hit-historic-low/"&gt; https://thememoryguy.com/dram-prices-hit-historic-low/&lt;/a&gt;.'><sup>4</sup></a></span> a group that sells “third-party independent market research and data” to investors in the semiconductor industry.<span class="easy-footnote-margin-adjust" id="easy-footnote-5-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-5-2408" title='See &lt;a href="https://objective-analysis.com/"&gt;here&lt;/a&gt; for their website.&lt;br&gt;“Objective Analysis – Semiconductor Market Research.” Accessed April 9, 2020.&lt;a href="https://objective-analysis.com/"&gt; https://objective-analysis.com/&lt;/a&gt;.'><sup>5</sup></a></span> We have not checked where their data comes from and don’t have evidence about whether they are a trustworthy source.<br/></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>Figure 1 shows McCallum’s data.<span class="easy-footnote-margin-adjust" id="easy-footnote-6-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-6-2408" title='See &lt;a href="https://docs.google.com/spreadsheets/d/1qF5vJ2-jkRygMQQIk3fj6G-WEW0ScDiKjyGseovFGqc/edit?usp=sharing"&gt;this spreadsheet&lt;/a&gt;, tab ‘McCallum Data’, for a copy of the dataset and resulting graph production.'><sup>6</sup></a></span></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <figure class="wp-block-image is-resized">
+ <img alt="" height="357" src="https://lh4.googleusercontent.com/7sOJigQV8ex9P7OnAwJ0KCFxnZhm_emqn8ovnqE6pKRBkdxtSw53yygAoL92MFzq7kwPOD73f10arNlImaoeTyD3oQhkL6vWHYOnZJveBUfV_Re_0AVoALA32O2QScdQce5FQvNt" width="579"/>
+ <figcaption>
+                   Figure 1: Price per gigabyte of DRAM from 1957 to 2018 from John McCallum’s dataset, which we converted to 2020 dollars using the Consumer Price Index.<span class="easy-footnote-margin-adjust" id="easy-footnote-7-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-7-2408" title="“CPI Home.” U.S. Bureau of Labor Statistics. U.S. Bureau of Labor Statistics. Accessed April 23, 2020. https://www.bls.gov/cpi/."><sup>7</sup></a></span>
+ </figcaption>
+ </figure>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>Figure 2 shows the average price per gigabyte of DRAM from 1991 to 2019, according to the Objective Analysis graph.<span class="easy-footnote-margin-adjust" id="easy-footnote-8-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-8-2408" title='See &lt;a href="https://docs.google.com/spreadsheets/d/1qF5vJ2-jkRygMQQIk3fj6G-WEW0ScDiKjyGseovFGqc/edit?usp=sharing"&gt;this spreadsheet&lt;/a&gt;, tab &amp;#8216;Objective Analysis Data’, for a copy of the dataset and resulting graph production.'><sup>8</sup></a></span><br/></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <figure class="wp-block-image is-resized">
+ <img alt="" height="357" loading="lazy" src="https://lh3.googleusercontent.com/D1tZNSvoc5MzwnXxmoq3_cqAU1GnNRN5J2Xh9z931OC29ffEY9r8sz5EpD_lZldnw9T9hgiym9phkeW2ao4I4mqU8vbrOD--2m2JVUHBvWtIimahHP4LAcDJMpa3Dphi_p1Jk0jo" width="579"/>
+ <figcaption>
+                   Figure 2: Average $ / GB of DRAM from 1991 to 2019 according to Objective Analysis. Dollars are 2020 dollars.
+                 </figcaption>
+ </figure>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>The two datasets appear to line up (see Figure 3 below),<span class="easy-footnote-margin-adjust" id="easy-footnote-9-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-9-2408" title='See the ‘Combined Data’ tab in &lt;a href="https://docs.google.com/spreadsheets/d/1qF5vJ2-jkRygMQQIk3fj6G-WEW0ScDiKjyGseovFGqc/edit?usp=sharing"&gt;this spreadsheet&lt;/a&gt; for graph generation.'><sup>9</sup></a></span> though we don’t know where the data in the Objective Analysis report came from– it could itself be referencing the McCallum dataset, or both could share data sources.<br/></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <figure class="wp-block-image is-resized">
+ <img alt="" height="358" loading="lazy" src="https://lh6.googleusercontent.com/FshFEQ0Ctb6hnT90vVF3ReYIPIOz0EyUOYYv2yZRNX4STpyKbALc29BVSNAlI_HScZYmIicb5rq_YOR9WjIVH8w3KKJKhiY85SDwsw1v4No6rW2JRBFGLuYwIeZ9rn7sXSEGdrD3" width="580"/>
+ <figcaption>
+                   Figure 3: $ / GB of DRAM from 1957 to 2018, with McCallum’s dataset in blue and the Objective Analysis dataset in red. Dollars are 2020 dollars.
+                 </figcaption>
+ </figure>
+ </HTML>
+ 
+ 
+ ===== Analysis =====
+ 
+ 
+ <HTML>
+ <p>For both sources, the data appears to follow an exponential trendline. In the McCallum dataset, we calculate that the price / GB of DRAM has fallen at around 36% per year, for a factor of ten every 5.1 years and a doubling time of 1.5 years on average. The Objective Analysis data is similar, with the price / GB of DRAM falling around 33% per year, for a factor of ten every 5.8 years and a doubling time of 1.7 years.<br/></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>The 1.5 and 1.7 year doubling times are close to the rate at which Moore’s law observed that transistors in an integrated circuit double.<span class="easy-footnote-margin-adjust" id="easy-footnote-10-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-10-2408" title='From &lt;a href="https://en.wikipedia.org/wiki/Moore%27s_law"&gt;Wikipedia&lt;/a&gt;:&lt;/p&gt; &lt;p&gt;“Moore&amp;#8217;s law is the observation that the number of transistors in a dense integrated circuit doubles about every two years.”&lt;/p&gt; &lt;p&gt;“Moore&amp;#8217;s Law.” Wikipedia. Wikimedia Foundation, April 8, 2020. https://en.wikipedia.org/w/index.php?title=Moore&amp;#8217;s_law&amp;amp;oldid=949708131.'><sup>10</sup></a></span> It seems possible to us that cheaper and denser transistors following this law are what enabled the cheaper prices of DRAM, though we haven’t investigated this theory.<span class="easy-footnote-margin-adjust" id="easy-footnote-11-2408"></span><span class="easy-footnote"><a href="#easy-footnote-bottom-11-2408" title='Moore’s Law only refers to the number of transistors, but our impression is that people also sometimes refer to a Moore’s price performance law, which suggests that the cost per transistor falls at similar rates. From &lt;a href="http://longbets.org/70/"&gt;this post on Longbets&lt;/a&gt;:&lt;/p&gt; &lt;p&gt;“Moore&amp;#8217;s Law, which has defined a doubling of price/performance/value produced by semi-conductors every 12 to 18 months since 1966, will continue to deliver its exponential benefits for at least another five decades, without stopping or slowing.”&lt;/p&gt; &lt;p&gt;Renan, Sheldon. “‘Moore&amp;#8217;s Law, Which Has Defined a Doubling of Price/Performance/Value Produced by Semi-Conductors Every 12 to 18 Months since 1966, Will Continue to Deliver Its Exponential Benefits for at Least Another Five Decades, without Stopping or Slowing.&amp;#8221;.” Long Bets. Accessed April 10, 2020. http://longbets.org/70/'><sup>11</sup></a></span><br/></p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p>Both datasets show slower progress in recent years. From 2010 onwards, the McCallum dataset falls in price by only 15% a year, for a rate that would yield a factor of ten every 14 years, and the Objective Analysis dataset falls by 12% a year, for a rate that would yield a factor of ten every 18.5 years.</p>
+ </HTML>
+ 
+ 
+ <HTML>
+ <p><em>Primary author: Asya Bergal</em></p>
+ </HTML>
+ 
+ 
+ ====== Notes ======
+ 
+ 
+ <HTML>
+ <ol class="easy-footnotes-wrapper">
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-1-2408"></span>From <a href="https://en.wikipedia.org/wiki/Dynamic_random-access_memory">Wikipedia</a>:<br/>
+                   “Dynamic random-access memory (DRAM) is a type of random access semiconductor memory … One of the largest applications for DRAM is the main memory (colloquially called the “RAM”) in modern computers and graphics cards (where the “main memory” is called the graphics memory).”
+                   <p>“Dynamic Random-Access Memory.” Wikipedia. Wikimedia Foundation, March 24, 2020. https://en.wikipedia.org/wiki/Dynamic_random-access_memory.<a class="easy-footnote-to-top" href="#easy-footnote-1-2408"></a></p>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-2-2408"></span>From <a href="http://jcmit.net/">his personal website</a>:<br/>
+                   “1998-2001 Associate Professor, National University of Singapore, School of Computing – Computer Science”.
+                   <p>“Biographical Information for Dr John C McCallum.” Biographical Information for Dr John C McCallum. Accessed April 14, 2020. http://jcmit.net/.<a class="easy-footnote-to-top" href="#easy-footnote-2-2408"></a></p>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-3-2408"></span>“Historical Memory Prices 1957+.” Accessed April 9, 2020. <a href="https://jcmit.net/memoryprice.htm">https://jcmit.net/memoryprice.htm</a>.<a class="easy-footnote-to-top" href="#easy-footnote-3-2408"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-4-2408"></span>We extracted data from the third graph in <a href="https://thememoryguy.com/dram-prices-hit-historic-low/">this blog post</a> using the popular online tool <a href="https://automeris.io/WebPlotDigitizer/">WebPlotDigitizer</a> to extract data. The graph is attributed to Objective Analysis, and appears to have been generated by the writer of the post, Jim Handy. (He explicitly says he generated previous graphs in the post.)<br/>
+ <br/>
+                    “DRAM Prices Hit Historic Low – The Memory Guy.” Accessed April 9, 2020. <a href="https://thememoryguy.com/dram-prices-hit-historic-low/">https://thememoryguy.com/dram-prices-hit-historic-low/</a>.<a class="easy-footnote-to-top" href="#easy-footnote-4-2408"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-5-2408"></span>See <a href="https://objective-analysis.com/">here</a> for their website.<br/>
+                   “Objective Analysis – Semiconductor Market Research.” Accessed April 9, 2020. <a href="https://objective-analysis.com/">https://objective-analysis.com/</a>.<a class="easy-footnote-to-top" href="#easy-footnote-5-2408"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-6-2408"></span>See <a href="https://docs.google.com/spreadsheets/d/1qF5vJ2-jkRygMQQIk3fj6G-WEW0ScDiKjyGseovFGqc/edit?usp=sharing">this spreadsheet</a>, tab ‘McCallum Data’, for a copy of the dataset and resulting graph production.<a class="easy-footnote-to-top" href="#easy-footnote-6-2408"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-7-2408"></span>“CPI Home.” U.S. Bureau of Labor Statistics. U.S. Bureau of Labor Statistics. Accessed April 23, 2020. https://www.bls.gov/cpi/.<a class="easy-footnote-to-top" href="#easy-footnote-7-2408"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-8-2408"></span>See <a href="https://docs.google.com/spreadsheets/d/1qF5vJ2-jkRygMQQIk3fj6G-WEW0ScDiKjyGseovFGqc/edit?usp=sharing">this spreadsheet</a>, tab ‘Objective Analysis Data’, for a copy of the dataset and resulting graph production.<a class="easy-footnote-to-top" href="#easy-footnote-8-2408"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-9-2408"></span>See the ‘Combined Data’ tab in <a href="https://docs.google.com/spreadsheets/d/1qF5vJ2-jkRygMQQIk3fj6G-WEW0ScDiKjyGseovFGqc/edit?usp=sharing">this spreadsheet</a> for graph generation.<a class="easy-footnote-to-top" href="#easy-footnote-9-2408"></a>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-10-2408"></span>From <a href="https://en.wikipedia.org/wiki/Moore%27s_law">Wikipedia</a>:
+                   <p>“Moore’s law is the observation that the number of transistors in a dense integrated circuit doubles about every two years.”</p>
+ <p>“Moore’s Law.” Wikipedia. Wikimedia Foundation, April 8, 2020. https://en.wikipedia.org/w/index.php?title=Moore’s_law&amp;oldid=949708131.<a class="easy-footnote-to-top" href="#easy-footnote-10-2408"></a></p>
+ </div></li>
+ <li><div class="li">
+ <span class="easy-footnote-margin-adjust" id="easy-footnote-bottom-11-2408"></span>Moore’s Law only refers to the number of transistors, but our impression is that people also sometimes refer to a Moore’s price performance law, which suggests that the cost per transistor falls at similar rates. From <a href="http://longbets.org/70/">this post on Longbets</a>:
+                   <p>“Moore’s Law, which has defined a doubling of price/performance/value produced by semi-conductors every 12 to 18 months since 1966, will continue to deliver its exponential benefits for at least another five decades, without stopping or slowing.”</p>
+ <p>Renan, Sheldon. “‘Moore’s Law, Which Has Defined a Doubling of Price/Performance/Value Produced by Semi-Conductors Every 12 to 18 Months since 1966, Will Continue to Deliver Its Exponential Benefits for at Least Another Five Decades, without Stopping or Slowing.”.” Long Bets. Accessed April 10, 2020. http://longbets.org/70/<a class="easy-footnote-to-top" href="#easy-footnote-11-2408"></a></p>
+ </div></li>
+ </ol>
+ </HTML>
+ 
+