Valve has published the April 2026 Steam Hardware Survey, and the big story this month is normalization. Last month’s data was a statistical mess, with Linux at an improbable 5.33%, the 16 GB RAM tier jumping wildly, and the RTX 5070 temporarily sitting at an anomalous position. April looks a lot more like what you’d expect from a stable month of organic hardware churn, which ironically makes it easier to spot what’s actually changing.
RTX 3060 Back in Its Natural Habitat
The RTX 3060 retakes the top GPU slot at 3.99%, which is exactly where it’s been for most of the past year. The February and March anomalies are clearly behind us. Rounding out the top five: the RTX 4060 Laptop GPU at 3.78%, the RTX 3050 at 3.04%, the RTX 4060 Ti at 2.45%, and the RTX 5070 at 2.86%. The overall picture is a Steam gaming population still dominated by Ampere mid-range hardware, with Ada Lovelace filling in the upper-mid tier and the RTX 50 series slowly accumulating share from the top down.
RTX 50 Series: Steady, Not Spectacular
The RTX 50 series as a whole is accruing share at a leisurely pace. The 5070 sits at 2.86%, the 5060 at 2.57%, the 5060 Ti at 1.81%, the 5070 Ti at 1.62%, the 5080 at 1.37%, and the 5090 at 0.41%. Two new entries worth noting: the RTX 5050 debuted at 0.16%, and the RTX 5050 Laptop GPU appeared at 0.24%, signaling that NVIDIA’s budget refresh is starting to filter into survey data. None of these numbers suggest a mass-market generational transition is underway. Pricing and availability continue to constrain adoption at the mainstream tier, and the Ampere installed base is just enormous.
AMD’s RDNA 4 Finally Shows a Pulse
The RX 9070 XT is still absent from the listed cards, but the RX 9070 appeared at 0.17% this month. That’s a first appearance in the survey and meaningful proof that RDNA 4 is actually reaching gamers, even if the numbers remain tiny. The generic “AMD Radeon(TM) Graphics” and “AMD Radeon Graphics” buckets continue to absorb a chunk of AMD discrete card share that should probably be attributed to specific models, but the trend is what it is: AMD’s discrete GPU presence on Steam remains historically weak relative to its actual sales numbers.
16 GB VRAM Keeps Climbing
The 16 GB VRAM tier reached 23.51%, up another 1.98 percentage points from March. It’s not yet at the top of the chart (8 GB still holds that at 26.76%), but the gap is closing quickly. This is the clearest quantitative signal in the survey that the RTX 50 series is having a real effect: 16 GB is now the standard configuration on NVIDIA’s mainstream lineup, and it’s showing up in aggregate stats. The 12 GB tier is gradually declining, which tracks as RTX 3060 units age and 50-series replacements accumulate.
RAM: 32 GB Claws Back Some Ground
Last month’s implosion of the 32 GB configuration, which dropped roughly 20 points in what was almost certainly a sampling artifact, is partially reversing. 32 GB sits at 37.55% in April, up 0.93 points from March. Meanwhile, 16 GB holds the top spot at 40.86%, down 0.11 points. The gap is narrowing from its March extreme. A new entry worth flagging: 28 GB appears on the chart at 0.55%, which suggests AMD Strix Halo-based laptops with their 28 GB unified memory configurations are accumulating enough of a Steam presence to register.
Linux Retreats from Its March Peak
Linux pulled back to 4.52% in April, down 0.81 points from March’s 5.33%, which we noted at the time was likely a sampling anomaly rather than a true breakthrough. The underlying trend is still real and still upward, driven by the Steam Deck installed base, but 5% was always going to need more than one month to stick. Arch Linux leads the distribution breakdown at 0.32%, followed by Linux Mint 22.3 at 0.27%.
Windows 11 Continues Its Slow Grind
Windows 11 reached 67.74%, up 0.89 points from March. Windows 10 held at 25.63% (+0.27%), which is somewhat surprising given Microsoft’s ongoing pressure campaign to get users off it. The doomsday clock for Windows 10 end-of-support in October 2025 has clearly not moved the needle as dramatically as some expected, at least among the Steam gaming population. Windows 7 64-bit is still at 0.07%, which at this point is basically archaeology.
CPU Manufacturer Share: AMD Crosses 45%
AMD’s CPU share hit 45.19% in April, continuing a months-long climb toward parity with Intel’s 54.81%. That’s a story that’s been playing out slowly for a couple of years now, with Ryzen’s competitiveness in both desktop and laptop markets gradually converting the Steam gaming population.
Display Resolutions: 1080p Holds, 1440p Climbs
1920×1080 remains the dominant primary display resolution at 52.21% (+0.28%). 2560×1440 is at 21.41% (+0.71%), continuing a multi-month upward trend as monitor prices for 1440p panels have come down substantially. 4K reached 5.09% (+0.30%), a slow but consistent climb. If the 1440p trend continues at its current pace, it may approach 25% by the end of the year.
VR: Meta’s Consolidation Continues
The Meta Quest 3 leads VR headset share at 28.58% (+0.92%), with the Quest 2 at 22.09% and the Quest 3S at 12.98%. Meta collectively accounts for the majority of the Steam VR user base. The Valve Index HMD slipped to 11.77% (-1.15%), its continued slow decline since the Index received no refresh. Total Steam users with a VR headset sits at around 1%, consistent with prior months.
Overall, April’s survey reads like a return to baseline after a chaotic March. Trends include: RTX 50 series adoption is real but slow, 16 GB VRAM is becoming the new normal, the Ampere installed base is massive and sticky, and AMD is gaining ground in CPUs while remaining a minority player in discrete GPUs on Steam. The 28 GB memory tier appearing for the first time is a small but interesting signal of where laptop hardware is heading.
Discussion (19 replies)
Join Discussion →I know it's a bit of a conspiracy at this point, but I don't think AMD cards get surveyed as often as Nvidia cards.
When I was running my 980ti in one PC and 6800xt in the other, I kept getting the survey popping up on my 980ti machine.
Now that I have a 9070xt and 6800xt in the two machines, I don't get surveyed on anything.
Maybe just my own perception, but when I hear people talking about it I do wonder.
That is a fair point ive been AMD a long while and think ive had.. 1 survey?
This time I got asked to participate in the survey two days in a row. Yesterday when I was in Linux, and today in Win10. Usually I only get asked once maybe a couple/few times a year, if that.
I have definately gotten asked to fill out the survey more while on an intel/nvidia setup than AMD/AMD or AMD/nvidia combos over the last 10-12 years. When I last swapped from an AMD/AMD system to the intel/nvidia one, I got a survey within a month I think yet I hadn't had a survey in close to two years on the all amd system. I'm soon going back to all AMD but will have three systems with steam on it, it will be interesting to see what system gets a survey first.
I don't understand the survey approach. Why not just collect system info during idle periods, keep a unique system fingerprint in their database against each Steam username (store multiple fingerprints if the user has multiple machines) and publish the survey using that data without sharing any personally identifiable information. It seems dumb that they ask permission from the users and then post those skewed results.
I'd rather they ask than just do sh1t without my permission or knowledge.
This is fine if it is opt-out by default, and the user has to manually opt-in.
I dunno if I want them doing all that.
I like the current system cuz they ask first, and then the hardware scan happens.
So are we ncluding the survey is crap?
For me it is. At least they should include a menu option in Steam so the user can participate instead of being invited randomly to take the survey. Radeon users reporting that they aren't presented with the survey is suspicious as hell.
It certainly sounds iffy, question is motive though? Also, it woulnt even be a survey if they cant remove duplicates and such matters.
Whatever valve does, they do, it may be the worse 'survey' of all time for all we know... But thats them, and whatever... however perhaps for people doing tech news and commentary perhaps should consider that if this 'survey' lacks reasonable published methodology, they shouldnt even discuss it, communicate with each other and ban this crap of even talking about it, valve can do whatever, but whats the point of discussing what may be complete garbage (i dont know of course, but yeah)
So... If you ask a mathematician, you don't need a huge sample size (assuming the distribution is homogenous). A few hundred upwards of a thousand results are enough to extrapolate useful results, and more sampling than that, regardless of the population size, are usually just used to try to correct for distribution anomalies.
There are, what, hundreds of millions of active accounts, and probably in excess of a Billion accounts total.
So just because you don't personally get an invitation to the survey very often doesn't mean that it isn't relevant or lacks accuracy. You have a roughly 1,000/100,000,000 chance (give or take) of seeing one in a given month. Plenty of people, even on these forums, report getting them occasionally - and given the sample size here, that roughly tracks with about what I would expect based on those odds.
They are looking into do this. Pretty sure I did a post on it last month but may have just put a link in another thread regarding Steam/Valve.
On the flipside, I've been using NVIDIA GPUs for over a decade, and AMD CPU/NV GPU combo for almost a decade, and don't remember ever getting a survey request. Perhaps I got one back during my 2600K/4930K era but don't remember but can confirm none since moving to AMD CPUs.
I've gotten it, on discrete GPUs and on APUs.
But it's still being reported enough that we're talking about it, which makes it suspicious at a minimum.
However, I would argue that the relatively small marketshare of AMD GPUs might have something to do with it; their users are typically louder than than their balance sheets support...
For something like real world surveys involving random people on the street, that works well due to the time and cost involved but the Steam survey is completely digital. Valve can certainly absorb the storage and compute costs of doing a more extensive survey. If the AMD/Intel hardware percentage being reported by them is inaccurate, they are directly hurting these products as people may decide to go with the more popular hardware if they use the survey results for decision making. The April 2026 survey reports NO 9060 models and no 9070 XT models. Only the 9070 model with a really sad percentage.
Then there are the weird anomalies:
What happened there?? Were all those GPUs, including the weaker 4060/5060, sold on the used market to AI companies? So these people bought the GPUs, jumped on Steam, played long enough to get an invite for a Steam survey and then sold them, all in one month???
Right now Steam is reporting something in excess of 100 different models of GPUs
There is a downfall in that smaller sample sizes need some level of homogenization to be truely representative. You go out and ask 1000 people out of 100,000,000, and if there are only a handful of the more rare cards, it's entirely likely that you will see some anomalies like you point out, and that some of the less popular cards don't see any representation at all due to sample sizing.
Could Valve send out a survey to everyone, every month? Sure, they could. It's digital, wouldn't cost them anything extra. But then you'd probably get just as many people/companies/whatever trying to use that to game the system, and introducing intentional anomalies, rather than seeing a true random selection.
I mean, we are talking differences in the tenths and hundredths of percent - very very small differences here. But it looks bigger than 0.1% because of it's absolute position on a list in numerical order and month-on-month deltas.
Every poll I have ever seen puts AMD graphics at about 1 in 3 users https://www.techpowerup.com/forums/threads/which-generation-is-your-graphics-card.308104/
Some also suspect the iGPU is being counted on many AM5 systems instead of the dGPU. Considering how broken the survey has proven time and again, it would not even be surprising.
Going back to the last decade, whatever the latest 60 series card is, and Chinese language will skyrocket on the charts in a given month. Only for Valve to say, our bad, and "fix it". The whole thing is suspect. Why have they not been able to patch out whatever is causing the massive over count? Some speculate it is the same gaming center PCs being counted over and over as users log-in to Steam. If accurate, why do those same systems keep getting invites? Steam has the telemetry data.
The survey is not audited, has zero transparency, and is opt in. While having the same massive data distortion occur over and over. Seems legit.
As noted; Massive over counts of a single language and a single GPU have happened repeatedly going back to the 2010s. Hanlon's razor or not, the same horrendous misrepresentation happening repeatedly is a whole field of red flags.