[quote name='grazing' timestamp='1311633365' post='4840230']while the %'s have shifted since that was last updated, I'm sure Vista/Win7 have not taken over in a combined % over XP and if you refuse to create the DirectX 9 rendering .dll (read above) you are in essence cutting out at least 1/2 of your potential buyer base which is crazy. You'll have to learn DirectX 9 anyways to support them, so might as well start there.

According to the latest Steam hardware survey, users with both Vista/Win7 and DirectX 10/11 GPUs are now on 56%. Total Vista/Win7 is 71%/72%. XP is down to 20%. DX11 looks more like the reasonable choice every day.
[/quote]



You have a link to the survey? Wondering if it's a "poll." that was given e.g. the results will not be that accurate by a long shot. Only amongst those who voted.

The OS + hardware data is generated by Steam itself, which queries the OS + drivers. They have the results here: http://store.steampowered.com/hwsurvey

So it's based on those who actually have this 'steam' installed. Gotcha.




Thank you for the link though.

Another statistics form those who has installed Unity Web Player: http://unity3d.com/webplayer/hardware-stats.html
There XP is 51.7% from all Windows'es.

IMHO, if I was putting out a game right now, I would support backends for both DX9 and DX11 (and importantly, feature-level-10 via DX11).

For a commercial endeavour to cut out 10%-50% of it's clients, you'd want to have a pretty good explanation from your engineering team.

If the game was being RELEASED now (or upto maybe a year in the future) and was targetting audiences covered by both the Steam and Unity survey then I would agree.

If the game was being STARTED now and unlikely to be released for 18 to 24 months then I would argue the engineering effort to fully support both APIs and cover the features would depend on your target market. If you are going after the Steam market for example I would argue (and have done at work) that supporting DX9/XP is wasted effort as 18 months from now I expect that number to be sub-10%.

If you were going after the Unity targted market hten 18 months from now it'll probably still be sane to cover both API/OS combinations as it will probably still be +40%, although I'd keep an eye on it as AMD's APUs might shift this value quicker than expected.

The thing to remember about the Unity data and wiki links is that they are sampling different data sets to the Steam survey. The wiki link certainly is going to pick up the thousands upon thousands of XP installs which are being run in offices which will never run games. The Unity data on the other hand gives a better view of the 'casual' gaming market, which people don't upgrade as fast as the 'hard core' Steam market does.

You can't look at just one data set and say 'see! do this!' you have to consider who you want to go after, the amount of effort you want to put in and long term trends over the development of your game.

Also; Battlefield 3 is being released as DX11 only. If nothing else that should be a pretty good guide as to where effort should be pushed. If someone like DICE and EA are willing to cut the remaining XP users free they can't make up a significant amount of projected income for them.

HOWEVER, as the OP is currently LEARNING an API there is no good reason for his to learn DX9; anything he produces in the next 18 months isn't going to be commerically releasable anyway and learning a dead/dying API is a waste of time when the newer one is better and, I would argue, easier to learn as well.

In short;
- pick your market
- watch the trends

Direct3D11 is better in my opinion, the only "problem" is that it won't run on windows XP (if you find that a problem..), people will eventually upgrade for hardware that supports Direct3D11 sooner or later, so better start from the newest release of DX, however, it's up to you, though if you wanna go for Direct3D11, this is a really great place to start from, and you can see the SDK for more sample, and as the others have mentioned, Direct3D10 is garbage, well it can't do anything better than 11