Screw this Steam survey. It's not accurate at all.
I've been tracing those records for three years and found that those models (among GPUs) which take less than 0.15% don't show up at all, and there are almost seasonal huge changes in manufacturer percentages (this month one grows and next month it falls back to normal). I don't think that's explainable or believable. What? All their users unplug or plug together at the same time monthly?
Yeah, this article should be retracted. A under 1% fluctuation in the Steam survey is basically meaningless, as they have huge seasonal swings, based on user behavior. For example people that use Windows 11 going on vacation on certain months more than people that use Windows 10. And that's also affected by geography, as some geographical areas can use Windows 10 more than Windows 11. Or a new game being launched and being most popular in a geographical area where Windows 10 is used more. Also, even ignoring that the Steam data is almost meaningless when trying to determine short term trends, a slight single month "dip" is not a "trend", as this article claims.
Just look at the "dip" for Windows 11 from
40.59% in Aug 2023 to
31.34% in Oct 2023, then up to
43.54% again in Nov 2023. Almost 10% fluctuations, up and down, for both Windows 10 and 11, over just a few months tells you everything you need to know about drawing conclusions based on this data when there is a less than 1% fluctuation, like last month.
If Steam would show a graph over time for the operating systems, like the have for other stuff, this issue with their data would be even more obvious. At the very least they should put a disclaimer about not drawing conclusions about the installed base based on their data, so that tech sites that are not familiar with their data, like TPU, do not get (repeatedly) fooled by it.
Anyway, let's do Steam's job and create a graph with their data about Windows 11 vs other OSes over time (I included Windows only, as it was easier to extract the data from the Wayback Machine):
As a note, the data for May 2024 probably has some errors, since the numbers do not add up to 100%, but they total 100.58% instead for some reason. But compared with the other problems of the data, that's not a big issue.
Other people in this thread said that Steam's data looks very clean. That's not the case at all, at least for the OS data. Just look at those huge spikes in March 2023 and October 2023.
