Choosing a minimum graphics (card) requirement?

I never understood this about games. How do developers decide what graphics card is required for minimum functionality?

Every single Steam game has a "Graphics" section under "Minimum Requirements", and each one has a different graphics card name written down. I honestly don't get it. It's so specific. With RAM you just have numbers, same with a CPU and hard drive memory. But with a graphics card, they just put down the actual name of the card, not it's speed or anything like that.

I read somewhere that the the name of a graphics card consists of the bran name, and a number. And the number is apparently a combination of the clock speed, and the bandwidth. But I checked one graphics card and the numbers didn't add up in any way.

And it's not like I can just buy hundreds of graphics cards to find out which is the minimum for being able to run the game. With stuff like RAM and CPU speed, I can just run a virtual OS and keep setting them lower, until problems arise, and determine the minimum that way, but I can't just select what graphics card the computer should think it's using.

Someone please explain this to me.

I looked around, but I either found stuff that wasn't what I searched, or old threads that didn't have any answers.

 

I would imagine that you need to look at the shader requirements for your game and make sure the cards you list have those capabilities.

 

The only solution for people who don't have access to all those graphics cards is to try to get people to help test. Not everybody has has only high end solutions. I for example have my desktop with 2 GTX980s(got it right before the 10xx line came out). Then I have a laptop that has switchable graphics, from 7th gen Intel integrated to an R7 M960(which has 2GB dedicated RAM, but is barely a step above integrated graphics beyond the dedicated RAM). In my case, I'm not doing anything that really taxes a system so what I have is enough for testing my own stuff.

Generally, if you intend to commercialize your games, I recommend an investment. You don't have to purchase 50 cards to figure it out, but maybe just a couple...or a laptop with integrated graphics. You may only need 3 cards/systems to get the idea. If your dev machine has a free PCI-E slot, you may be able to use that with a mid-range card and separate monitor(so you don't have to swap cards, etc...). This assumes that your dev machine is high end, and you need to test lower end stuff.

 

Test out your game on old hardware. Borrow old computers from friends. Use the lowest rated of these graphics cards, which runs the game fine, as your minimum requirement. Go to gpuboss and find the graphics card on the list and look for closely rated cards from other manufacturers (it's nice when you list an NVIDIA, AMD and Intel card in your minimum requirements). You will end up with minimum requirements that are slightly too high, but you can feel pretty confident that you are not deceiving anyone.

 

As to Trexug's comment...I agree completely with this. I should add on that there is nothing wrong with simply stating in your requirements that the lowest end system you tested on was xxx. That it may work on lower end systems than that, but that this was the one you tested. Better honesty than not.

By the same token, if you are monetizing, you could very well provide a 1 level demo of sorts. Make sure this demo is as graphic intensive as any part of the game. Then, a player can use this free demo not only to test the game itself, but to see if it works on their system before they purchase. This way, you avoid most situations where someone purchases the game and finds out it didn't work on their system. Nothing is perfect in the business world, and there will be people wanting refunds just because....but this would give you something to fall back on.

 

Personally, I pick a computer as a minimum target spec, then I tune the game to to run on that computer with some minimally comfortable configuration. The computer's specs are then my games minimum specs.

The other option would be to just test your game on a whole bunch of computers, and list down the specs of the least powerful one that got acceptable performance.

You'd need much more extensive testing to get a broader requirement that's still useful. I guess analytics could be useful there - record framerates at key sequences in your game, record the important performance specs, and find the parts most commonly listed at your minimum acceptable frame rate.

 

Most of the time, the way devs select a minimum graphics card does not have anything to do with whether the card is capable of running the game or not. It is there to signal "If your card is weaker than this card, we will not support you and we cannot guarantee your game will run or run at a good framerate."

So yeah, what everyone suggested seems good. Getting a computer for minimum spec testing or using a friend's crappy PC, depending on how rigourous you want to be.

 

I used to work at a QA outsourcing company that some of the 2nd tier big names would use for exactly this type of thing. We would get short run contracts to test a few levels the devs believed to be the most demanding graphics wise on a list of 30+ hardware configurations, mostly focused on different graphics cards. We would then report numbers such as max, min, and typical FPS, load times, and file bugs for any graphics issues. We would also provide any comments if we believed this hardware configuration was actually realistically playable. The devs would then use that information to nail down a minimum and recommended hardware spec for the game, or to decide if they need further optimization.

I didn't work on the $$$ side of the business, but I believe a week long test run like that would cost between $2500 to $4000. You can send me a private convo if you want the exact company I worked for that would do this work for you.

That is probably a little high for what a typical solo indie would want to spend, even though it shouldn't be out of the question for a small indie team. The alternative I would do would be to enlist friends to play test your game, and display the FPS on screen. Get feedback on what graphics settings and graphics hardware they are using, how smooth they thought the game was, was it playable, etc. Based on that info, you could also buy a handful of outdated used graphics cards for cheap (spending maybe $300 for 5 or 6 outdated cards) to nail down what you expect the low end to be.

 

Gamers these days will either know if they are above or below those requirements. Because everyone at some point checks the market for what they can afford and they become familiar with the numerous brands ie only Nvidia/AMD and the performance differences on them. Especially for graphics cards, as they're usually the only part in a computer that gets upgraded out of most pc components , maybe ram upgrades. Intel/Amd both lock users into motherboards where decent cpu update will require entire mobo replacement with all marginal cpu upgrades.

And stick in something like Nvidia GTX660 as minimum gpu, and most AMD gfx card users would know where there own cards performance stands in relation to that card/brand with just a quick benchmark search if they didn't already know.

Do make sure your game has texture quality setting to reach the lowest gfx card ram spec.. 2gb ram on a gfx card would be the lowest option you should be able to fit your games texture S*** into these days. I've seen games released in the past that have F***ed that up. And we still see gtx1060's that are out in the market with only 2gb ram not cool nvidia, laptop market tends to get the gimped out on the gpu's aswel.. so that's another area to make sure you can reach a pretty low spec with graphic options in your game at least look back like 4years to what was average gpu.

If you're indie dev you should be following the hardware market anyway, doesn't take a lot to check out a graphics card review and see the comparative benchmarks every now on and then.. and not be developing on a 1080gtx with 32gb+ system ram and then putting out minimum specs you haven't even bothered testing yourself or getting someone else to is never a good idea.

Should flail the rust developers for thinking there S*** only needs 8gm ram requirement.. maybe like 2years ago, I think most of its users would agree it chews up way more and spills over into page file churning these days with just 8gb system memory.

 

If I can't measure graphics card speed then I don't expect a non-enthusiast to know which is faster than which or where and how to find out.

Technical stuff aside, here's two real-world considerations:

• As an enthusiast who knows this tech well, even I don't usually look at or make decisions based on requirements. (Partly because I'm lucky enough to have recent, mid- to high-end hardware.)
• I don't expect my average customer to know enough to make informed decisions based on tech jargon, because my target audience aren't tech enthusiasts.

An enthusiast may know the difference between a "GTX960" and a "9600GT", but most human beings don't. On top of that, there's loads of stuff that can make a computer behave more slowly than it should, and it's a terrible customer experience to hear "well, your GTX 1080 can handle the game, but all that bloatware on your box is bogging you down".

So, if a potential customer asked me whether my game would run on their machine, especially with platforms like Steam, I think my default answer would boil down to something like "give it a go, and if it's not comfortably playable then grab a refund from Steam". A technical answer might be correct, but that's less important than the user experience.

 

I don't know if Unity has any way to do it, but the way many AAA games will auto-select graphical settings is really nice for an end user.

 

Alternatively you could write a small app (and suddenly I'm curious if it could be and how effective it would be if done through WebGL) that quickly analyses their system and tells them the likelyhood that they can run the game.

A quick benchmark (either with the exact same code written for the above app or something very similar) followed by changing the game's quality setting to the most applicable one.

https://docs.unity3d.com/ScriptReference/QualitySettings.SetQualityLevel.html

 

That's a pretty round-about way to take a guess at something which doesn't take the human factors into account.

Doing it from within the game itself to adjust quality settings is a different proposition. In that case you're not guessing how the game will run, you're measuring it, and you can also measure the outcome of changes made.

 

Just a comment...I don't mind if a game does "auto-settings" as far as quality settings. But if I can't override those settings I will be upset. This is for more than one reason. One, my system might be capable of super high settings, but I may want to lower them at any time, either because I'm going to tax my system(say for recording/streaming), or because I simply don't like all the glitter and rainbows, which sometimes take away from the game if done too much. It could also go the other way. I've seen a game be old enough that for some reason the "auto-settings" sets me at a low setting even though my system is more than capable(Mechwarrior 4 does this in my case), and I'd like to be able to crank up the settings regardless.

 

For sure. Games that do this rarely prevent you from changing it later.

I imagine you could do a check for a settings.ini file or something, and if it doesn't exist run the check.