AMD Passes 30% CPU Market Share in Steam Hardware Survey, Loses in GPUs

Steam Survey = does not equal GPU share world wide--not even close. Why people like to make something out of the Steam survey--a 100% opt-in survey, that doesn't even include all Steam accounts because X number of Steam accounts never opt in--and we don't know how many Steam accounts there are because Valve never says--beats me. I guess it's always good filler material on a slow news day, imo.

waltc3 said:

Steam Survey = does not equal GPU share world wide--not even close. Why people like to make something out of the Steam survey--a 100% opt-in survey, that doesn't even include all Steam accounts because X number of Steam accounts never opt in--and we don't know how many Steam accounts there are because Valve never says--beats me. I guess it's always good filler material on a slow news day, imo.

100%

The steam survey is used because all the data is collected in one area for you doesn't require much work so makes it easy for articles like this. However it in no way speaks for the whole market. So anytime I see reports like this I mentally add an asterisk to it

Not that many Steam users are quick to adopt the latest hardware. The most popular GPU on the platform is the GeForce GTX 1060, according to the survey, and 1% of Steam users are somehow continuing to get by with CPUs featuring clock speeds lower than 1.4 GHz. It takes a while for the survey results change much.

That statistic is likely referring to base clocks, so it doesn't really mean a whole lot for modern CPUs, especially for the "lower TDP" variants. Even a brand new i9-11900T desktop CPU only has a 1.5GHz base clock, but it will boost up to 4.8-4.9GHz for lightly-threaded loads, and has a 3.7GHz all-core boost. The 11700T similarly has a 1.4GHz base clock, and the 11400T a 1.3GHz base clock. On the laptop side of things, the i7-10710U has a 1.1GHz base clock, and the i7-1060G7 a 1.0GHz base clock.

waltc3 said:

Steam Survey = does not equal GPU share world wide--not even close. Why people like to make something out of the Steam survey--a 100% opt-in survey, that doesn't even include all Steam accounts because X number of Steam accounts never opt in--and we don't know how many Steam accounts there are because Valve never says--beats me. I guess it's always good filler material on a slow news day, imo.

The Steam Hardware Survey results may not perfectly represent all systems, and obviously won't represent things like business systems, servers and others without the service installed, but they are likely fairly representative of the systems running Steam, and in turn systems used for gaming in general. Most Steam users are realistically going to click through to submit the survey, as it's completely automated and takes no effort on their part, and while not every user gets surveyed every month, enough should be surveyed to provide a good representation of the userbase as a whole. And while some may choose not to submit the survey, its questionable that there would be any significant correlation between those users and any particular hardware configuration, so it's unlikely that would affect the results much either.

And as for the number of systems represented, Valve recently reported during their yearly review that they had over 120 million monthly active players in 2020, so probably somewhere around that. Again, you don't need to survey every single system every month to obtain meaningful results, especially since a system's hardware typically isn't changing on a monthly basis. And due to the size of Steam's install base, the trends should be fairly similar for systems used for gaming that are not running Steam on a regular basis as well.

Of course, there are some limitations with the data they provide. The results are worldwide, so they are not necessarily representative of the distribution of hardware in any given market. The ratio of hardware used in Asia may vary substantially from the ratio in North America, for example, based on the availability and pricing of various components in those regions, and they don't provide any way to restrict the results to a particular market. The size of certain markets can also grow relative to others over time, which can potentially shift growth trends in ways that are contrary to what's actually happening. Likewise, laptop and desktop data is mixed together, which can muddy the results, and the same goes for older systems compared to modern systems that are capable of running the latest games. Having more ways to filter the results down could make them a lot more meaningful, but Valve doesn't make that functionality publicly available.

There have also been times where things like changes in the software used to manage internet cafes has thrown off the results until corrections get made some months later. So, changes from one month to the next will often not be particularly meaningful. The wider, multi-month trends are probably more meaningful though, like how AMD's processors have been gradually increasing from under 20% up to over 30% of the Steam install-base over the last 18 months.

I'm dead set on getting an AMD GPU even if they're worse on physics and reflections, and have much poor performance with Ray Tracing, just because in raw power and drivers they're much more reasonable.

The thing is that I'm stuck with a 1000 series nvidia GPU because I can't find any AMD GPU besides the 6900 XT and at that price point right now, no thanks.