Better, Faster, Cheaper: Measuring the Speed of Science

Are we better off now than we were 10 years ago? Often times this question is answered subjectively and will vary from person to person. We can empirically show how life expectancy has increased over the centuries thanks to advances in the fields of agriculture and medicine, but what about quality of life? Science affects our lives every day, and the general notion is that better science will (eventually) translate into better lives. There is a burning curiosity shared by myself and others to quantify how we have progressed in science over the years:

Click for full article. Source: Bornmann, L. & Mutz, R. (2015). Growth rates of modern science: a bibliometric analysis based on the number of publications and cited references. Journal of the Association for Information Science and Technology, 66(11), 2215–2222.

Bornmann and Mutz demonstrate in the image shown above how we have been doubling scientific output every nine years since the 1940s. That is not to say that we have become twice as smart or efficient; this phenomenon could be partially fueled by a desire to gain prestige through a high number of publications. To better assess the topic of efficiency, we can measure how long it takes to perform specific procedures and how much they cost. This article compares the rate of improvement for DNA sequencing, PCR, GC-MS and general automation to the rate of improvement for supercomputers and video game consoles.

Moore’s Law: 41.4% better per year

An observation by Gordon Moore in the 70s predicted that the number of transistors per circuit would continue to double every two years, and indeed the trend held true for five decades. The 41.4% rate is calculated from the exponential growth formula: V2 = V1(1 + r)t where V2 = final value, V1 = initial value, r = growth rate (decimal) and t = time. If the rate is known, the doubling time can be calculated: t = log(2) / log(1 + r) where V2/V1 = 2. If the doubling time is known, the rate can be calculated: r = 10(log(2)/t) – 1.

GC-MS: 5.4% faster per year

A gas chromatography-mass spectrometry (GC-MS) analysis of forensic samples required at least 16 minutes with first generation instruments (1952) but was decreased to less than 90 seconds in 1996, as determined by JeromeJeyakumar et al. This means that the speed of GC-MS analyses doubled every 13.2 years.

General automation: 6.6% faster per year

Sarkozi, Simson and Ramanathan analyzed 36 years of data at the Mount Sinai School of Medicine to quantify the effect of automation in the lab. They found that each employee completed 10,600 tests per year in 1965, which increased to 104,558 tests per year in 2000. Assuming this is characteristic for most labs, this data suggests that automation has been doubling the average lab’s productivity every 10.9 years. The authors also looked at cost per test and saw a drop from $0.79 per test to $0.15 per test over that same time period.

Video game consoles: 53.8% faster per year

Floating operations per second (FLOPS) is an established measurement for comparing computer performance. Worth noting is that there are factors besides the number of transistors per circuit that contribute to net FLOPS. The Nintendo Entertainment System (NES) of 1985 was estimated at 7 million FLOPS, and the PlayStation 4 was measured at 1.84 trillion FLOPS in 2013 (FLOPS for the graphical processing unit, GPU, was used since it’s a better indicator of performance for a video game console). This means video game consoles have been doubling in performance every 1.6 years.

Supercomputers: 62.6% faster per year

The IBM 704 supercomputer was capable of 12,000 FLOPS in 1955. The current reigning champion is the Tianhe-2 supercomputer, capable of 33.86 quadrillion FLOPS since 2013 (that’s the equivalent of 18,400 PS4s). This means supercomputers have been doubling in performance every 1.4 years.

DNA sequencing: 81.6% faster per year

The Human Genome Project ran from 1990 to 2003 and cost $2.7 billion to sequence a single complete human genome. Fourteen years later, scientists can now do the equivalent in 26 hours for $1,000. Using a value of 12.58 years (or 110,276 hours) as the time for the first whole genome, we can say that scientists are doubling the speed of sequencing every 1.2 years.

The NHGRI noted that $2.7 billion is not an appropriate amount to compare costs since their facility performs other activities that are not “production-oriented DNA sequencing” and instead use a cost of $100 million per genome in 2001. This means that DNA sequencing has been 105.4% cheaper per year, becoming half as expensive every year.

PCR: weeks to hours

Calling polymerase chain reaction (PCR) a staple for biology labs does not do justice for this now fundamental process. Amplifying sections of DNA would take a couple weeks before the advent of PCR by Kary Mullis. Now this process takes a couple hours. The enabling discovery for PCR was thermostable DNA polymerase—Taq polymerase—isolated by Chien et al. in 1976 from the Thermus aquaticus bacterium living in hot springs and hydrothermal vents. Calculating the percent increase or doubling time for PCR is difficult because there is no clear-cut amount of time for the improvement. Do we use the value of one year? Or eight years when we think about the time between the discovery of Taq and PCR? PCR would be better classified as a discontinuous innovation, and many scientists pay homage daily to PCR to enable their own discoveries.



While somewhat challenging to measure, quantifying technological improvements demonstrates in a very convincing manner how far mankind has come. Moore’s Law itself has reached an end due to restrictions from physical laws such as problems with heat generation and the unpredictable behavior of electrons below the nanometer scale. The next step with computer processors—like with amplifying DNA using PCR—will require discontinuous innovation. While we currently may not be able to conceive of such “next generation” computer processors or the next laboratory breakthrough, we surely won’t be able to imagine our lives without them after they come to be.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.