Are manufacturers throttling your computer’s (or phone’s) performance?

Q. Is computer performance throttling a common problem, and is there anything that can be done about it?

A. Computer performance throttling refers to the idea that manufacturers purposely build their products to perform more slowly over time as an incentive for customers to upgrade. I’m not sure if it’s real, but it often seems like a common problem, as every computer I’ve owned over the past 30 years seemed to have performed slower after a few years (based on the increasing amount of time required for both the operating system and applications to launch). I used to think the declining performance was due to a combination of hard disk fragmentation and added direct link library (DLL) files associated with the installation of applications, but now I wonder whether computer throttling has also contributed to this problem all along.

In 1987, my office colleagues gathered around my new IBM PS/2 computer, where I remember one of them remarking, “It’s a screamer!” as we watched WordPerfect and Lotus 1-2-3 launch almost instantly. Within three years, it took about five seconds for those same applications to launch on that same computer. Every computer I’ve owned since then has demonstrated similar declines in performance, which makes me wonder why.

In December 2017, the first of many lawsuits was filed against Apple Inc. for allegedly throttling the performance of its older iPhones, presumably (as argued by the plaintiffs) as an incentive for users to upgrade to newer and faster devices. In late 2017, the company acknowledged that certain operating system updates released for the iPhone 6, iPhone 6s, iPhone SE, and iPhone 7 included a feature “to smooth out” power supply from batteries that are cold, old, or low on charge, and that these updates might slow the device’s performance to some degree. This revelation has some experts in the computer industry asking whether this has been a consistent tactic employed by hardware companies to encourage users to upgrade to newer computers and devices. (A Bing search for the phrase “performance throttling” produces more than 500,000 hits on this topic.)

Hardware companies have explained that performance throttling, sometimes referred to as “dynamic frequency scaling,” is a common strategy for preserving energy or preventing devices from overheating, so there may be valid reasons why a device’s performance is automatically ratcheted downward. For now, the controversy is unsettled, so I’ve developed a crude test in an attempt to detect performance throttling based on the computer’s date, as follows:

1. Using my 2017 Dell XPS 8900 Intel Core, x64-based PC, i7 CPU @ 4.00GHz, 4 cores, 8 logical processors, 16 GB RAM, Windows 10 Home as a test subject, I installed all outstanding updates on my computer and then rebooted.
2. I rebooted a second time to make sure any outstanding updates were completely installed and properly configured.
3. I then rebooted a third time and measured how long it took. (I considered my computer to be finished rebooting when my last autoloading application appeared on screen, which was Snagit Editor in my case. You could use any milestone you choose, provided you are consistent across all measurements.)
4. I launched the following applications: Excel, Excel again, Word, Outlook, and Photoshop to measure how much time it took for each application to launch.
5. Next, I disabled the setting that adjusts the computer’s date and time automatically, and then adjusted the date back one year to when I first purchased the computer. I then rebooted and repeated steps 3 and 4 above, again recording how long it took to reboot the system and launch the applications.
6. My intentions were to compare the results to see if my computer performed faster when it was reset to the older date, but the results were inconclusive because the Office applications returned user license error messages due to the improper date, and Photoshop failed to launch altogether.
7. However, when I restored my computer to the current date, and then rebooted and relaunched my applications, I found that it performed about 32% faster overall, as indicated by the crude test results shown on the next page.

These results, which are based on just one computer, are too small a sample size to draw any definitive conclusions, but these time measurements do seem tantalizing enough to warrant further inquiry. If any readers would like to test their computer system following the steps I’ve outlined above and then forward me the results (Carlton@asaresearch.com), I’ll tabulate the data and report back to you once a larger sample size of data has been collected. Who knows? Maybe these results will provide insights into performance throttling. Or perhaps these efforts may lead to potential solutions for working around performance throttling problems, if they do indeed exist.

About the author

J. Carlton Collins (carlton@asaresearch.com) is a technology consultant, a conference presenter, and a JofA contributing editor.

Submit a question

Do you have technology questions for this column? Or, after reading an answer, do you have a better solution? Send them to jofatech@aicpa.org. We regret being unable to individually answer all submitted questions.