Which version of OpenGL/Direct3D should I target for optimum compatability?

Question

When we develop web pages we can broadly work out which browsers to support based on market share.

When we develop in .NET we can broadly work out which .NET version to develop for based on which Windows versions have it installed.

But when developing OpenGL or Direct3D applications, how do we know which video cards people (I mean "people" as opposed to "hard core gamers" or "companies using CAD" :P) are using? Are there statistics on such things? Is there some common logic that people use to work out what version to support? Just as most companies have supported (perhaps until just recently) a minimum of IE6 in web pages, is there a general consensus to support a minimum of, say, OpenGL 1.5, or DirectX 8 or something?

I note that we can find out which specific video cards support which versions of these APIs, but how do we know which video cards people are actually using, is there any kind of research on this?

N.B. I'm more interested in OpenGL because that's what I'm using, but I mention Direct3D because I assume the same problem applies.

Answer 1

OpenGL 2.1 is a good bet. The newer OpenGL 3 doesn't offer that much more functionality. You have to check for the availability of all of the OpenGL extension anyway,so you don't loose much by sticking with 2.1.

For DirectX: Use DirectX 9c. That is the latest version that still runs on WindowsXP. Drivers are stable and very mature. DirectX 10 offers more functionality but you will lock out the user-base that still runs WindowsXP.

About compatibility for non-gamers: Graphic cards that don't support these APIs (at least to a usable degree) have died out more than five years ago. Given the typical life-cycle of a PC you can be almost sure noone will have problems.

If any user complains that the software doesn't run on his 10 year old matrox-parhelia card he should just buy the cheapest graphic card he can get. It will run much faster and will cost a fraction of the software.

Answer 2

It depends on the timeframe of the project on the DirectX side.

If the project is to be ready in months, then follow the advice from Nils. IF the time is over a year, I would reconsider limiting to DirectX9 as the XP machines are getting lesser and lesser with the time. DirectX11 would be a good idea, especially with retrocompatibility till feature level 9.0.

Answer 3

Market fragmentation for 3d hardware has always been a huge problem. There is no simple answer.

You need to define who your uses are. Is this a casual-oriented game that you intend to sell to people who haven't updated their computer in years? Is this a business application that is aimed at workstation-grade hardware? Are you just looking for an average middle-of-the-road game-buyer? Is this something simpler than a game like a screensaver that you plan to sell to people who don't buy games at all? Do you need to support laptops? Netbooks?

The market fragmentation is considerably worse with OpenGL than it is with DirectX. The official standard requirements from the Khronos Group are all well and good. But the hardware vendors tend to be very slow to update their drivers to match the standard. Many features are required by the GL spec, but are only implemented in a fallback software path that is absolutely unusable in commercial software. The new OpenGL 3.1 spec tries to improve this situation by removing support for most older, poorly supported features. But if you need to support hardware more than a couple years old (or most modern Intel integrated GPUs) then GL 3.1 would be aiming too high.

A good place to start for general hardware usage stas among game purchasers is the Steam Hardware Survey ( http://store.steampowered.com/hwsurvey ). Steam is the most popular digital game distribution service that covers a variety of games from casual to core. They reset the numbers periodically to keep it current. Last year they reported that they had 25 million active users, so the sample population is pretty good.

So you probably need to narrow your target customer group down more. I would recommend picking some recent competing applications that you consider to have a similar customer base to yours and basing your target hardware around what they require.