In 1965 Gordon Moore, the co-founder of Intel, wrote a paper giving his observation that the number of components on integrated circuits was doubling due to economic considerations roughly every year and that he expected that trend to continue for the foreseeable future. That's the same thing as saying that CPU performance is doubling every year. Sometimes we use the figure of 18 months, instead of one year, but the idea is the same. The thing that annoys me, is that people are still saying this, even though it hasn't been true for a long time.
The false belief that Moore's law still holds today bothers me so much that I decided to plot the data myself for Intel CPU's. To do this, I took the passmark scores of Intel CPU's made roughly between 2001 and 2018. I plotted them against the dates they were first released to the public. The dates came from Wikipedia articles (see references below). I looked at single-core performance, rather than multi-core performance for two reasons. First, the vast majority of commercial software today still only uses one or at most two cores at a time. So, it doesn't matter if you have twenty cores on your computer, you will only be using one or two of them. Second, the true measure of "CPU performance" is the performance of a single CPU core. However, I realize that this can be debated. Regardless, I plotted 245 data points of passmark scores and release dates in the chart below. I did not include any atom processors or any ultra-low power processors, those designated with a "U", because those processors are down-clocked. Thus, they are intentionally not run at their their true performance, and I'm not sure to what degree the passmark scores reflect this.
Single-Core CPU Processing Power versus Time
You'll notice that from about the year 2005 onward, the plot is a straight line, not an exponential curve. So, the time between doublings will no longer be constant. As time goes on, there will more years between each doubling. It looks like CPU performance has doubled this last time over the last eight years, between about 2010 and 2018. If this trend continues, we will not see another doubling until about 2034. If this is correct, it means that if you were to buy a reasonably capable computer today, you should be able to get good use out of it for at least the next sixteen years! Moore's law is dead.
The death of Moore's law is great for those of us who are cheapskates. It means we can wait a very long time before we will be forced to buy our next computer. So, we should have no hesitation about buying a used computer at the knee of the cost versus time curve, or even significantly to the right of the knee. This also means that we should now be buying computers that will last a long time in preference to the latest shinny ultrabooks. We should be looking at thicker, upgradeable laptops in which we can easily replace the hard drives or SSD's when they die and perhaps upgrade RAM if necessary. And fast USB ports are also important, because software is still getting larger in terms of required hard drive space. These facts mean used business laptops are definitely the way to go--at least until businesses start realizing that the death of Moore's law means they don't have to upgrade as often either. So far, that doesn't seem to have happened. I sense big, big bargains ahead--at least until everyone else figures this out.
Copyright © 2018-2019 The Cheapskate's
Guide to Computers and the Internet. All rights reserved.