Friday, October 2nd 2026

Steam Hardware Survey: 32 GB Overtakes 16 GB as the Most Common System RAM Configuration

Valve's September 2026 Steam Hardware & Software Survey shows that 32 GB of RAM is now the most common memory capacity among participating gamers. The 32 GB tier gained 4.77 points to reach 42.22%, while 16 GB fell 3.38 points to 37.82%. Last month, 16 GB still led with 41.20% against 37.45%, so that is a big swing for a single month. The idea had been around for a while, with 32 GB already being floated as the next standard back in August 2025. Steam only reports capacity, so the survey says nothing about speed or whether those kits are DDR4 or DDR5. It makes sense, though, with several recent games asking for 32 GB RAM in the recommended specs, with 16 GB now often listed in the minimum requirements.

Elsewhere, AMD gained 1.89 points on Windows to reach 47.41% CPU share, and the GeForce RTX 5070 is the most common GPU at 5.86%, up 2.30 points. 1440p primary displays climbed 5.18 points to 27.04%, and Windows 11 keeps growing at 71.34%. Of course, one month doesn't make a trend, so it's reasonable to wait for October's numbers to see whether 32 GB remains at the top. The survey is also voluntary, and the sample changed a lot this month. For example, Simplified Chinese jumped 7.42 points to 31.39% of surveyed systems. That change in sampled systems could explain part of the RAM swing, though it could also simply be 32 GB RAM gaining popularity.
Source: Steam Hardware Survey
Add your own comment

44 Comments on Steam Hardware Survey: 32 GB Overtakes 16 GB as the Most Common System RAM Configuration

#26
Athena
I really wish that they would just include everyone, and anonymize the data, and then and only then can we get a true picture of what people are playing with

This data is just too random to make any meaningful results one way or the other
Posted on Reply
#27
Red Hood
Net cafe pcs in China being surveyed multiple times again?
Posted on Reply
#28
_JP_
AthenaI really wish that they would just include everyone, and anonymize the data, and then and only then can we get a true picture of what people are playing with

This data is just too random to make any meaningful results one way or the other
Not only, the universe of queried systems is not known, or if it is consistent across the time of every survey. How is the distribution across regions, OSes or system types (laptop/desktop/handheld)
The irony is that player count per game can be trusted, as well as the estimated worth of games libraries that are made public and there is valuation of items.
But actually tracking the hardware span of the entire userbase is a complex task.

These two always make me laugh, because there are that many VR headsets that it breaks frame size and the MacVoid that is becoming the most common playing platform, over at Apple camp. :laugh: (I'm aware that it's the inability to read the system's name string)
Posted on Reply
#29
harm9963
I have a 4090 , plus DDR4 64GB Ram, have not seen a game yet go past 20GB ,much less 16GB, but I play at 4K DLLSQ or DLAA ,seen my 4090 at 23gb, back in the day when I had two 470GTX SLI or 290x CFX, did see 16GB of Ram get used ,plus my 1080Ti 11GB get used plus around 12 GB Ram .
Posted on Reply
#30
JohH
There's something seriously wrong with Steam's sampling for monthly variations this large.
Posted on Reply
#31
SiliconSeraph
My own bias for the 5070 aside, I'm really surprised the 5060 Ti isn't higher than it. The 16GB variant is/was such a good deal, especially for 1440p gaming, and until now I kinda assumed most people bought that or the 5070 Ti over the 5070. Weird.

Part of me loves the fact the dang GTX 1650 is still in the top 10! :laugh:
Posted on Reply
#32
mrnagant
BwazeThe most common memory itself is more expensive than the top end gaming consoles. We used to say that about GPUs. What's next, due to severe supply constraints, package substrate shortages midrange CPU prices soar past PS5?
The cheap Crucial memory I want for my laptop (2x16GB DDR5 5600) is currently listed for $500 (sold out), but 3rd party is $565. In 2023 it was $66. 1yr ago it was $120. I got the laptop beginning of 2026 and the memory was $250.

:(
Posted on Reply
#33
AusWolf
RinSenpaiBy the way, I found the reason of this month survey

Considering how unlikely it is that 7% more Chinese gamers bought a new PC than anyone else since the last published survey results, I hope now everybody sees why I've been saying that the Steam survey isn't representative data.
Posted on Reply
#34
TheinsanegamerN
_roman_the fake linux binary, aka full ubuntu installation with steam binary reads out and touches too many sensitive binaries. i did not gave consent on the ubuntu full installation steam fake binary!

A game client does not need to read out hardware on every bootup, does not need to download 512mib of data on every bootup, does not spam me full with advertisement and has a mess of configuration options. a few ubuntu client options do not even work when you enable or disable them. the communication software and other nonsense also does not belong in a game shop, game drm platform to launch your games.

Download numbers are shown totally wrong, 5GiB off as shown before usage. Disk explorer does not lie. when i install a test game which shows 55mb i expect only 55mib, and not 90mib for the game, plus junk here and there etc. 5gib is a round down number, it was more of course
dude, you claim to be a coder and software guy but the concept of de compressed files and external libraries leaves you reeling?
dgianstefaniI wonder if memory is a bit cheaper sourced for the Chinese cyber cafes.

Or maybe they just understand you can run most games with 16 GB, but not necessarily all the other things with them, e.g. browsers, communication software, streaming etc.
China has access to CXMT first hand, which certainly helps. No tariffs, lower taxes, backdoor deals, government backing, ece ece ece.

Also in the context of a business, you buy what you need. They'll make so much over the years of service the higher price for 32 GB isn't an issue. You pass the costs to the consumer and keep chugging along.
Posted on Reply
#35
PierreJG
It might be based more on DDR 4. DDR 5 just isn't worth it at the moment. Even DDR 4 prices went up here as our retailers are opportunistic; they knew people will go back to DDR 4 and it is almost double the price. I want to get 64GB as my computer needs Memory and Core Count. 64GB DDR 4 cost the same as an RX9070XT/RTX5070 non TI here.
Posted on Reply
#36
Lew Zealand
RinSenpaiBy the way, I found the reason of this month survey

QFE

All the Steam Survey data is again skewed this month and like in previous skewed months, will return next month to expected levels. The fun part is I think I know why this Chinese cybercafe oversampling occasionally occurs:

Steam surveys randomly by user but if that single user (cybercafe owner) owns many dozens of PCs and is a cheapa**, he logs in with the same Steam account on all his mostly identical PCs and gets sampled many dozens of times. Why do I think this?

Because I'm close enough to that same guy as I logged into Steam on 15-20 different machines last month and every single one got the Survey. And I did the damn survey so I hope Steam likes it's Vega 6 and Quadro M4000 and GTX 1650 Mobile and i9-9980HX and Ryzen 5400U and all the other crazy setups I had.

One Steam login, well over a dozen samples. IMO same thing happens at the cybercafes and it's not random because each cafe uses a single Steam login. This is a hypothesis and IMO a reasonable one. I'd love to hear a better one if ya got it!
Posted on Reply
#37
N/A
32 GB sounds good for a gaming PC, but you could also get by with 24 GB if you have a triple-channel setup (like I do) or a 24 GB module, though I’m not sure if those even exist.
Posted on Reply
#38
dgianstefani
TPU Proofreader
N/A32 GB sounds good for a gaming PC, but you could also get by with 24 GB if you have a triple-channel setup (like I do) or a 24 GB module, though I’m not sure if those even exist.
These days with DDR5, you'd have to mix a 8 GB with a 16 GB, or go single channel and get 1x 24 to achieve that. The 8+16 is also not full "dual channel", i.e. it's going to be dual channel, dual rank most likely, at best dual channel triple rank. There's no way to get the optimum dual channel quad rank setup with 1x 8 and 1x 16. There's potentially quite a few unpopulated ranks since the 8 is definitely single side RAM, and the 16 is likely single sided. For max bandwidth on a dual channel DDR5 system, you need a kit of two dual rank sticks with memory on both sides, (2x 32 or 2x 48, rarely 2x 16). Below that is 2x 16 single rank, ideally with low capacity chips so you get both memory sides populated (dual rank), below that is 2x 8, below that is 1x 16, below that is 1x 8. 2x 24 is basically the same as 2x 16, but typically the worse tREFI and other timings of 3 Gb chips compared to 2 Gb chips.

Basically DDR5 with the subchannels doubling over DDR4 adds complexity, but when you understand it, it becomes a bit easier to work out.

OFC this doesn't account for die performance, with Hynix A die and new version of M die being best, and everything else being gradually worse. You can make up for lack of ranks/channels/density etc by better timing/subtimings and higher MT, but that is still something you can do with dual channel dual rank, everything is relative. Generally speaking you can get more tuning headroom out of lower capacity sticks, but this isn't typically significant enough when comparing high end low capacity good bin dies to high end high capacity good binned dies, both tuned well, to fully offset the advantages of having quad rank, especially if you have CUDIMM, helping high density memory run as fast as low density memory.

Out of the box, even with XMP/EXPO, dual rank high density kits will have worse timings/subtimings than the same grade of low density kits, but you can typically set the timings to of the high capacity kit to that of the lower (primarily tREFI) and equalize timing performance, while retaining the rank advantage, and thus pulling ahead. This may be offset by max frequency you can achieve, especially on AMD, or cooling of the DIMMs.
Now let's talk about the tradeoff that comes with hitting this price point. Standard DDR5 DIMMs use 8 ICs per module to deliver their full bandwidth potential, but this kit's 8 GB DIMMs are populated with only 4 ICs each per stick, half the normal complement. This directly reduces the raw memory bandwidth available compared to a fully populated 16 GB or 24 GB DIMM running at the same frequency, and it's the primary reason this kit consistently lands at or near the bottom of the relative performance charts across all three platforms tested, even against 5600 JEDEC kits with a full IC count. This isn't a defect or an oversight, it's simply the cost-cutting measure that allows XPG to hit an aggressive price point, and buyers should understand it as a direct hardware performance tradeoff rather than an inconsistency in testing.
Dual rank 32 (2x 16 GB) kit performs better than single rank per channel 32 (2x 16).

At this point I'd consider buying a 1x 32 GB, and if needed, later adding a second.
Or: Trying to find a dual rank per channel 2x 16 GB kit.

Or a compromise, not having dual rank modules, but having slightly more capacity, and similar OC performance as 1x 16 modules, and get 1 or 2 24 GB modules.

@ir_cow can explain more if he has time.
Posted on Reply
#39
melkor_unlimited
dgianstefaniBesides, your list doesn't show this "$749" 9070 XT. So the other user saying it's a false statement, is also correct.
Or "700 euro".

Can you circle the €700 one?
Or the $750 one. Either is fine.

Thanks :laugh:
Do not thank before you check what you are asking for.

$750 is without VAT.
Prices in EU shop are with VAT (20% in Germany)
Converting USD -> EUR and adding VAT you will get
$750 --> 801€

So anything listed bellow 801€ is bellow MSRP of Radeon RX 9070XT.
Posted on Reply
#40
RinSenpai
melkor_unlimitedDo not thank before you check what you are asking for.

$750 is without VAT.
Prices in EU shop are with VAT (20% in Germany)
Converting USD -> EUR and adding VAT you will get
$750 --> 801€

So anything listed bellow 801€ is bellow MSRP of Radeon RX 9070XT.
But €801 isn't the MSRP of the 9070 XT. MSRP is US$600 before sales taxes, and back during the release, it translated into about €570 before sales taxes. This was about €680-690 after VAT. When Nvidia reviewed their MSRP and decreased them when the dollar dropped like an anvil in the water back to 2018/2021/1999 level at €1 = US$1.20, doing the same on AMD's side would have lead a supposedly new MSRP of €620 including VAT.

You're just proving us AMD's MSRP was fake if you really believe what you said at €801 MSRP.
Posted on Reply
#41
harm9963
harm9963I have a 4090 , plus DDR4 64GB Ram, have not seen a game yet go past 20GB ,much less 16GB, but I play at 4K DLLSQ or DLAA ,seen my 4090 at 23gb, back in the day when I had two 470GTX SLI or 290x CFX, did see 16GB of Ram get used ,plus my 1080Ti 11GB get used plus around 12 GB Ram .
latest game Control Resonant ,max Ram is 14GB, GPU 16GB , and that's with recording , 12GB RAM max ,NVIDIA's RTX Mega Geometry technology has a lot to do with that !
Posted on Reply
#42
melkor_unlimited
RinSenpaiBut €801 isn't the MSRP of the 9070 XT. MSRP is US$600 before sales taxes,
You are right.
It sounded too good to be true.
Posted on Reply
#43
Ols-Hol
Rather abrupt move in stats, this could be another of those China spike months or methodology changes/quirks.
Posted on Reply
#44
jdrch
Been on 32+ GB on my main rig since 2012.
Posted on Reply
Add your own comment
Nov 20th, 2026 13:51 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts