Appreciate the reply. It made me look into this situation more deeply. For reference, I'm using
https://www.videocardbenchmark.net/high_end_gpus.html for GPU ratings and
https://store.steampowered.com/hwsurvey/videocard/ for info on steam users.
you're not allowing some settings to be scaled down,
Good point about scaling. The VR minimums are pretty much non-negotiable, since it's already as stripped and optimized as possible, but the non-VR quality can be cut down quite a bit. There are options where you can radically cut down the graphics and enemy numbers. I used to run the game on my own old laptop.
I don't think I'd like to sell the game based on that though. Perhaps some compromise there is possible.
or you just don't really know what the minimums are.
Yes, I really just don't really know what the minimums are. The fact that I'm running a beta to find that out and said "For normal, non-VR play, my guess would be..." makes that clear.
I'll mention how I arrived at the ballpark guess for low end specs that could use normal settings at 60fps, 1080p. On my GTX 970 the typical GPU frame takes 5 to 8.5ms. The typical physics frame, which is the lions share of the CPU time, takes 3 to 7 ms. The two happen simultaneously so which ever is longer is the only one that matters. At 60 fps there's 16.6 ms available.
I picked a GTX 770 because it has a passmark of 6,081 compared to 8,593 for my GTX 970. So it's about 70% as fast. Applying that to the frame time, that gives a frame time of 12.2ms. That's fast enough and leaves some slack, so like I said maybe a generation lower than that is okay too.
If I push these numbers to their logical conclusion, then the lowest passmark would be only 4400! That would be a GeForce GTX 570. But that leaves zero slack on the GPU and ignores all the stalls and problems that might cause. The numbers are all approximate and only a real world test can show what actually works.
The CPU has a lot more room for lower quality. A cpu half as fast as mine should be okay, though again only real world tests will show what actually works. Plus a lower power cpu means longer loading times, so again I wouldn't want to push the numbers to their limit.
Non-VR Minimum Specs:
DirectX10 (Windows 7+)
GTX 770/RX570s.
DirectX12 capable cards != GTX 770 specs. Plus you're not using DirectX 12.
I'm not using DX10 either. The point is, most people don't have really old cards running on XP and are at least DX11 capable.
I'd like to clear up another point too. I've had people tell me "You shouldn't make a PC first person shooter. You should make a mobile game because it has a bigger market."
I'm not interested in making a mobile game, or a sim that runs on an ancient laptop. I'm interested in making a hardcore shooter with cool graphics that can even run in VR. So if your argument is I should make a game that runs on the lowest end laptops with ancient integrated graphics cards, that's not my goal.
If making lots of money was my main goal, I wouldn't be an indie game programmer. That said, I'm certainly hoping to make a profit. It's certainly not that I don't care about money. I very much do. But I also care about working on something I'm actually interested in. Otherwise, I may as well stack shelves at the supermarket.
You say you mostly don't use Windows. Given windows is the main PC gaming platform for both first person shooters and VR, it sounds like you aren't aiming at the same gamers I am.
The min specs for PUBG are a GTX 960, the min for Doom is a GTX 670, which is quite close to a GTX 770 (passmark 5,372 versus 6,081, so within about 12%) and I said it could be a step below a GTX 770.
So, with the idea that I should aim for old laptops out of the way, let's look at what cards people are using.
These are the top 10 from the August 2018 Steam survey.
Code: Select all
Card Share Passmark
NVIDIA GeForce GTX 1060 13.76% 8,986
NVIDIA GeForce GTX 1050 Ti 8.74% 5,931
NVIDIA GeForce GTX 1050 4.85% 4,595
NVIDIA GeForce GTX 960 4.32% 5,805
NVIDIA GeForce GTX 1070 4.15% 11,183
NVIDIA GeForce GTX 750 Ti 3.62% 3,731
NVIDIA GeForce GTX 970 3.51% 8,593
NVIDIA GeForce GTX 1080 2.80% 12,308
NVIDIA GeForce GTX 960M 1.96% 2,221
NVIDIA GeForce GTX 1080 Ti 1.50% 14,057
If I just go by my original guess of a GTX 770 then that's a passmark score of 6,081. That's the following from the top 10.
Code: Select all
NVIDIA GeForce GTX 1060 13.76% 8,986
NVIDIA GeForce GTX 1070 4.15% 11,183
NVIDIA GeForce GTX 970 3.51% 8,593
NVIDIA GeForce GTX 1080 2.80% 12,308
NVIDIA GeForce GTX 1080 Ti 1.50% 14,057
Turns out that covers the same number of the top ten than if I'd just set it at a GTX 970 instead of the GTX 770. So I know this group can play the game fine since they all have cards as fast or faster than mine.
That's 25.72% of Steam users, and probably covers the gaming demographic I'm targeting pretty well: People who play first person shooters as opposed to prison simulators, cookie clicker games, MOBAs or whatever.
Also notice several cards only just miss the cut. Allow a little wiggle room of 10%, since it's only a ballpark estimate and I said maybe a generation lower, plus I demonstrated above that I'm being conservative with a GTX 770. That's a min of 5,472, just below a GeForce GTX 680. That means the following out of the top ten.
Code: Select all
NVIDIA GeForce GTX 1060 13.76% 8,986
NVIDIA GeForce GTX 1050 Ti 8.74% 5,931
NVIDIA GeForce GTX 960 4.32% 5,805
NVIDIA GeForce GTX 1070 4.15% 11,183
NVIDIA GeForce GTX 970 3.51% 8,593
NVIDIA GeForce GTX 1080 2.80% 12,308
NVIDIA GeForce GTX 1080 Ti 1.50% 14,057
That's 38.78% of Steam users. Again, probably covers my demographic pretty well, while missing out on the dove dating simulator market.
(10% wiggle room is even unnecessary. 5% leeway will be a passmark of 5777, covering all the above cards, only marginally above a GTX 680 at 5678 and just below a GeForce GTX 960 at 5805.)
So on paper it looks like the lower spec of a GTX 770 is roughly correct, but maybe too conservative. A lower estimate might be a GeForce GTX 580 to still get 60fps 1080p at normal graphics settings. That's maybe pushing it too far, only a test can tell.
I guess it depends on what people expect to get if they're using the minimum spec. If they're happy to play at less than 1080p then a GTX 580 could probably handle 1600x900 fairly easily... maybe. I don't know. That's the point of a beta.
Maybe I'm setting too high a standard for min spec... given that low spec is basically the same as for standard spec (1080p 60fps standard settings). I guess I should allow for lower settings, like 900p and the lower texture quality option. Maybe even 30fps instead of 60fps. Those options dramatically lower requirements, especially 30 fps.
Lowering the spec to that level might increase the number of steam users covered up to 45%-50% so I guess that's worth doing, although of those extra people probably less and less are the kind of gamers I'm targeting.
Basically, if I just assume 900p instead of 1080p for min spec users, the game is accessible to half of Steam users. But the top third is probably my main demographic, and I already cover those.
Thanks for making me look at this more precisely. Makes me think I should keep in mind that low spec users should expect lower quality graphics, it doesn't make sense to for them to expect the same quality as recommended spec users. With that in mind, a GTX 770 becomes more like a recommended spec rather than a minimum spec, and it can probably easily go lower than that and still maintain 60fps 1080p standard settings, probably around a GTX 680 to a GTX 960.
"In theory there is no difference between practice and theory. In practice, there is." - Psychology Textbook.