Image: https://steamcommunity.com/discussions/forum/10/651619326016...

At some point a platform/store front is going to want you to set minimum requirements to run your game. Let’s dig into what that actually means, and how we might arrive at one.

Some quick examples from random steam pages

System Requirements

Minimum:

OK, So, operating system, CPU, GPU (sometimes with capacity), Ram (capacity), Storage (sometimes), and additional notes of guidance (sometimes).

What do we actually mean by Minimum Requirements?

If a user is below any one of the minimum requirements, and the game doesn’t function properly, that’s not the developers fault. Regardless of the reason. The game is not expected to work properly on anything less powerful or older than what you have specified. That doesn't mean that the game won’t run at all, it might. But the user is in an untested place and there’s no guarantee or expectation that it should work. In times gone by, if you were below minimum requirements, the game would just tell you that and close itself. Better to simply say, “Hey, this isn’t going to work”, than to let the player try and have it crash, or fail, or be crummy.

Secondly, if a user is at or above minimum requirements, they should be able to play the entirety of the game. Ideally your min spec runs the game at the quality and performance level of the intended experience for all players. If you’re only listing one set of specs, then this ideally is what you mean. At or above this watermark, and you are getting the experience the developer intended.

But often that isn’t the case. Instead, it means at some low but reasonable performance level. What a low reasonable performance level is, is a bit fuzzy, will vary from person to person, and from genre to genre. So if your minimum isn’t the intended experience, having a concrete statement on what players should expect at minimum is a good idea e.g.720p@30fps. Letting players know it will work, but at a lower than you might otherwise expect resolution and a lower than you might expect frame rate. You’ll note the phrasing, minimum requirements vs recommended. Minimum in this case being much more aligned with “well it won’t crash to desktop, but I wouldn’t do it”, and recommended being “Should run just fine, if you made me play it on that, I wouldn’t be mad about it”.

What is the impact of our minimum requirements?

The higher we put our minimum, the less players might even attempt to buy the game as they don’t meet them. Making the addressable market smaller. We’ll look at some rough data on that shortly.

In contrast, players will judge higher requirements against what the game ‘looks’ like. If they think the art style/look or type of game shouldn’t need that much power, players might be put off, or convinced to avoid the game under the assumption that it must be poorly optimised and thus poorly made. The same way that you’ve investigated your peers/competition in the market for its gameplay and price point, you’ll want to pay a bit of attention to what they set as minimums. If your minimum is way higher, you’ll want to make it clear why, in terms of value proposition, not an apology.

The minimum requirements.

RAM is hopefully straightforward, if you can get by with 8GB then you cover over 98% of steam hardware survey responses. 16GB that’s still over 85% of survey responses. Speed and configuration of RAM is going to impact performance of the game but it’s not commonly listed.

Remember, this is if the user is running steam, your game and nothing else. Ideally you want to fit entirely in ram and not use the swap. If the game does end up using swap, expect frame times to drop and become spikey.

OS this should be easy; Steam needs Windows 10 now anyway. With over 95% share on windows, ~3 on linux (which will include steam deck) and the rest on mac. This also gives you a floor on other aspects but they aren’t really useful. https://www.microsoft.com/en-us/windows/windows-10-specifications

https://learn.microsoft.com/en-us/windows-hardware/design/minimum/windows-processor-requirements

Processor (meaning CPU); a bit less clear from our randomly grabbed examples. Some talk about cores/threads, GHz, or model numbers. Part numbers are the way to go here. If you can specify the model, users can determine if they are newer and if they are in the correct performance class. It’s fairly easy for a user to search for “how does my ryzen 5600 compare to a i7 8700k”, and get back useful information. Specifying a part also comes with it a swath of required features without you having to list them all and users try to decipher what they have.

If you aren’t concerned with the performance of the chip, then list a generation/series i.e. a core i7 6700k. Not just an i7, as i7 has been stamped on high and mid level chips since 2009

You’ll also note these almost never refer to mobile chips, always desktop parts. We’ll leave diving into the details of that for another time.

Graphics (meaning GPU); This is a bit messier. But the best answer is again a chip/model. It’s very easy to compare the card they have against the one you’ve listed. Coupling this with the vram requirement if you have one.

Unlike CPU, we often allow the user a lot of control over how much pressure they apply to the GPU. Changing resolution, AA, various other effects that increase the amount of GPU memory and performance required. To put this another way; often running the game at 4k vs running it at 720p does not change the cpu performance required but is the difference between over 8 million pixels or a little under 1 million pixels. So here you want a minimum resolution and quality level in mind.

Storage, not always listed but nice to include, this should just be the total installed size of the game on disk, plus the space required for its minimal number of save files.

How do you choose/arrive at minimum requirements?

In an ideal world, you would set them at the start of production, based on current hardware in the market and how long till the game would release. Or determine what a vast percentage of the games target audience has and set that as the minimum. Or perhaps just the entire market at large.

Similarly if you are targeting a console or handheld, those are fixed and often on the lower end.

During development computer(s) that are at (or below minimum spec) would be used (often by QA if no one else) to continually confirm the game is running correctly. If you don’t have access to the target platform developer kit, you might put together a slightly underpowered pc counterpart.

Let’s assume that didn’t happen though. OK, so the computer you are on, how powerful is it? If it covers a huge portion of the steam survey results then maybe you are good to go, just test on that. If it’s pretty powerful, do you have any older computers or parts you could swap in for testing? Got a friend or family member with a less powerful PC. If none of those are options, then you’ll want to farm for some pc specifications from some of your first playtesters or other developer friends.

Another aspect to consider is the tools/engine/tech you are using. These will put technological requirements on the platform, such as Manual: System requirements for Unity 6.2 .

Storage and RAM requirements aren’t usually going to vary or budge. So you can find those out regardless of your situation. How big is the install size, how big are the saves, how much ram does the game need to run. If your game's ram use is quite varied throughout, you might want to write a bit of code to track usage/peaks. One note on RAM, your engine might be conservative about unloading unused resources, or might require you to instruct it to unload things that are not in use. This might be keeping ram use higher than it needs to be.

Resolution, over 50% of the market is using 1080p primary, another ~20% at 1440p. By aspect ratio, over 80% at 16:9, 4% at ultrawide, and 7% at 16:10. By vertical pixels. Less than 7% of users are under 1080 pixels high.

GPU, there’s a lot of them. So many that it looks like over 15% of them are in the survey results are globbed together into categories like ‘AMD Radeon Graphics’, or make up such a small share (less than 0.15% it seems to be the cut off) that they get dumped in ‘other’. Removing those combined sets, we end up with a huge chunk of the top share being exclusively nvidia cards, and almost entirely mid range cards from the last 4 generations, xx60, xx70, a lone xx80, the 3080. I’ve smashed together some steam data and some aggregate perf numbers from Toms Hardware, (steam hardware data copied july 2025). There’s a 4x perf gap (in frame time) between the slowest and fastest card in that pack, lower struggling to reach 30fps at 1080p (a 1050 ti) and the higher reaching ~140fps at 1080p (a 4070 ti).

So perhaps as a rule of thumb, if you have a mid to high end card that is relatively recent you’ll want to be seeing a 2x-3x frame time at 1080p to have confidence that the higher market share, lower end and older generation cards, can attempt to hold even 30fps. You’ll want to see higher frames at lower resolution, not just ok frames at high resolution.

CPU, this is just a big ol’ mess and we don’t have great data from steam to dissect. But as we said in GPU, if you have a more powerful gpu and see high consistent framerates at lower resolutions, then that’s a good indicator that cpu performance isn’t the overall limiting factor. The other thing to keep in mind, don’t take gpu comparison data too literally if you are attempting to use review scores to arrive at minimum spec. GPU reviews usually try to run everything on high end cpus to remove it as a limiting factor as much as possible, so they get a more apples to apples comparison of the one part that is changing. But very rarely is someone buying the top of the line AMD 9800X3D CPU and pairing it with an nvidia 1050 ti.

But to get some real comparison we’re going to try some quick cross checking with https://www.userbenchmark.com/page/developer . This set of user supplied results may be a little higher or lower, but a quick comparison of their stacked perf by share of results for GPU is not entirely out of wack. Still seeing a pit of 1060 or worse when we get anywhere near 70% market share. So at the 70% and above market share, we see a lot of middle tier desktop cpus from over a decade ago among some higher end laptop cpus from 2019. That may sound a bit rough, but it does seem to line up, there’s a bunch of GPUs from around a decade ago, makes sense that they have entire systems that are that old or older. These CPUs are by no means slouches, but they are comparable to the lowest end chip you can get today, an N100 found in those little mini pcs that cost like 200 bucks even when they’re fully fitted out and with a windows license.

Brass tacks

If you aren’t making a complex 3d game with lots of effects and dense dynamic environments, then computers are really pretty darn fast. If your game should kinda run on anything reasonable, then getting numbers on screen and playing on a friend or family member's clunker is going to let you know if you have an issue that needs attention or minimum requirements that need lifting. You could always look for a cheap second hand test machine, or perhaps one of those N100 based mini pcs if it can run well on that it can probably run ok on pretty much anything, including the things like the steamdeck.

Some reasonable breakpoints

OK so let’s try to sus out what some brackets might be.

Biggest Share

On the GPU front, everything above the 70ish percentile is 1060 or worse.

Ram, there’s a huge gulf between 16GB and 32GB, meaning even at the biggest share 16GB is ~85th percentile. If you can run on 8 you’re laughing.

Resolution is a blow out, 1080p is most of the market, its >90th percentile. If you can run at anything that does 16:9 (meaning 720p), you’re golden.

Cross checking our cpu data back to steam, a mid range cpu from a little over decade ago lines up, 4 cores, something like an i5 2500k.

Decent rig with an ok share?

Now we’re talking.

Ram, if you want 32GB of it, you’re down to 42% share, let’s take the 86% 16GB.

1440p puts as at like 33%, so we’re staying with 1080p

GPU, a 3060 and up gives you around 50% of the market

CPU (again messy), but let’s look at something near 50% share by perf, a Ryzen 5 1600 (or 2600 or 3600 ) or a Core i7-8750H (oh or a Core i5-6600K)

Note: we probably have a user here that expects to be able to drive higher framerates than 60.

Sicker rig, how bad is the share?

Let’s call that a 3080ish or better, 16% share.

1440p is 33ish, 1440p 21:9 is 13% share

32GB ram is over 42%, if you want more than that, you’re down to single digits

For CPU, pick a mid to high end cpu from 2020 on, i.e. AMD 5600x

(Coincidentally, this wound up being pretty close to the Borderlands 4 recommended spec)

That’s a surprisingly high share for that much performance, the lower resolution is a tad surprising. This makes me wonder if that’s a chunk of players with higher end gear massively preferring framerate over resolution.

Closing thoughts

We’ve talked about 30 and 60fps as common framerates, but those are increasingly on the lower end. There are more and more monitors at higher refresh rates. The Steam Deck OLED does 90hz. 1080p IPS monitors are routinely at 180hz or higher at little extra cost. 1080p has been way out in front in the steam hardware survey for a long time now, despite much higher resolutions being available in that period. The logical conclusion is that players are preferring to get a 1080p monitor that does higher refresh rate rather than higher resolution ones. Taking this one step further, players are preferring responsiveness over further increases to visual fidelity.

We’ve used an nVidia 1060 as a lower water mark, let’s talk about that. While it is approaching a decade old, with 6GB where most media outlets would recommend at least 16GB today, and 4.4 TFlops which is apparently very close to a PS4 Pro. It’s aging, and you might love to have way more performance to play with. At the same time, when Valve rolled out the Vive, the minimum requirements for VR were only a 970 (more like 3.4 TFlops), a generation older, and trying to push 2160×1200 (about 25% more pixels than 1080p) @ 90Hz. So by some metrics and for some games, a 1060 is crazy fast.