Microsoft on DirectX 10

[KimB]

Of course, you're talking about independent developers or developers with huge backing from their publishers. "any time soon" is open for discussion. Games that require DX10 would be different from games which "Microsoft specifically bankrolls those games to evangelize Vista". Timing is the keyword. I also suspect the console industry/situation/market would play a part too.

First of all, with no DX10 hardware available now, and probably not until this fall or later, there just can't realistically be any game in production right now that is slated to require DX10. Also, production time for games has been creeping upward for years now. Most seem to take around three years. And then I'd be willing to bet that any publishers besides Microsoft will specifically avoid bankrolling a DX10-only game until they know that the hardware installbase is there.

So, give two years for the install base of DX10 hardware to improve (side note: the Vista desktop is only DX9, so most off-the-shelf computers will only make use of DX9 hardware). If, by then, the publishers feel it's worth taking the risk of providing a game with great graphics, even if it shuts out some of their customers, then they may bankroll a game that supports only DX10.

Then, production starts, and it would probably take 3 years for the game to actually be finished.

Thus, I predict five years for us to have more than a tiny handfull of DX10-only games (probably none). And even after five years, they'll still be trickling in. Notice that it's only been very recently that we've seen any DX9-only games (I think Oblivion was the first?), and DX9 was released in December of 2002.

 

[Demirug]

You cannot compare D3D9 and D3D10 in this special point. D3D9 is fully backward compatible to older hardware. Even Voodoos work together with D3D9. This allows any game developer out there to port a D3D8 engine without losing hardware compatibility. The porting from D3D8 to D3D9 was very easy too.

This time everything changes. After you have made a full port you lost backward compatibility. Additional the port is not that easy because D3D10 lost functionality compared with D3D9.

This makes it necessary to select one of multiple porting strategies. There are at least 5 main strategies and some have variants. Every one with pros and cons.

One of them allows building a D3D10 only game in the same time a rebuild takes. But I am sure we don’t want this, as the only advantage could be the usages of the maybe better D3D10 driver.

 

[KimB]

I don't see how that has any bearing on my post, Demirug. I'm not talking about games that support both DX9 and DX10, but games that require a minimum of DX10. I don't see how that will change by any amount from previous history.

 

[Demirug]

IMHO we need to clearly divide between API level and hardware compatibilities. This time the API and the hardware have a strong link. These means that stay compatible to older hardware can be much more expensive as in the past. This could motivate some developers to drop the pre D3D10 support earlier than in the past. On the other hand I expect that at least the same amount of developers will not go to D3D10 before they can make a full transition.

 

[KimB]

The separation in time between games that really started to make decent use of SM2 and games that required it was pretty small, though, about a year.

 

[Lux_]

1. Procedural generation. It's not happening. I'm with John Carmack on this one: procedural generation just cannot be applied as a general technique, and thus won't be done to any meaningful extent (in terms of performance) in games, because it adds far too much to the development time to be worth it. It may be useful for content generation tools to make lives easier on developers, but it most likely isn't going to be used for the content itself.

I'm not so sure. It largely depens on tools, I guess, but imagine a detailed and modeled tree. Like the palms in Crysis videos, but oaks. Tens, maybe a hundred (slightly) different. It's quite expensive to store and stream these from disks, when memory capacity of the videocards is growing faster than available bandwith.

I agree that there is no easy general soution for generatig procedural content, but trees, bushes etc. are relatively simple and recursive.

 

[KimB]

Sure, that's one thing you could apply to runtime content generation. And it's already been done (Oblivion). But that content isn't generated in the pixel shader.

 

[Blazkowicz]

what annoys me about games requiring DX10, they'll de facto require Vista as well. Most DX9 games still run on windows 98 (even though some require NT 5.x). and this time there are less incentives to upgrade the OS, the only major new features of Vista are bloat and the WDDM (and sound drivers in user space as well)

I think when you have a perfectly working and stable OS, it sucks to have to give it up only to play upcoming games, and go through the hassle of again paying the microsoft tax or getting and installing a Vista enterprise corporate..

 

[KimB]

Well, games requiring DX10 doesn't really bother so much in that respect.

What bugs me is that if I purchased a brand new shiny DX10 video card next year, I would have to get Windows Vista to make use of the new features.

 

[Parousia]

What bugs me is that if I purchased a brand new shiny DX10 video card next year, I would have to get Windows Vista to make use of the new features.

What would bug me even more would be if Vista required DX10 video cards to simply run.

Don't complain too much I think the game developers are pleased enough with the Vista situation given that it could have been much worse if M$ simply stomped their foot (and probably will get away with it).

 

[krychek]

What? Won't we atleast get opengl drivers that expose the features of these DX10 cards on XP?

 

[Parousia]

and this time there are less incentives to upgrade the OS, the only major new features of Vista are bloat and the WDDM (and sound drivers in user space as well)

Win98 was a bloated Win95 yet almost everyone upgraded. WinXP is not a bloated Win98 yet most folks upgraded as well.

I expect Vista to have the same (and probably higher) impact as XP did. It's a bigger step over WinXP than XP was over Win98.

I think when you have a perfectly working and stable OS, it sucks to have to give it up only to play upcoming games, and go through the hassle of again paying the microsoft tax or getting and installing a Vista enterprise corporate..

Then you should have no complaints about XP vs '98 when it comes to the "perfectly working and stable" criterias. I have absolutely no doubt Vista would surpass XP in these criterias.

A "perfectly working and stable" is subjective. I get the feeling you simply don't like paying for a new Microsoft OS because all you do with your machine is to play games. I think Vista is the most significant OS from Microsoft even if games were never created. My opinion, of course

 

[Demirug]

What would bug me even more would be if Vista required DX10 video cards to simply run.

It’s possible that a future version of the desktop will need D3D10 hardware to run.

Don't complain too much I think the game developers are pleased enough with the Vista situation given that it could have been much worse if M$ simply stomped their foot (and probably will get away with it).

I’m not sure how many developers already know about the “Vista situationâ€￾. I have the felling that many of them wait for the final release of D3D10 before they start looking at it.

The generally situation looks for me like everybody (users, developer, IHV) want D3D10 for Windows XP but no one is willing to pay for this “upgradeâ€￾. Let’s say some smart guys offers a D3D10 for Windows XP kit likely Cedega offers DX for Linux. Do you think the can survive?

 

[Demirug]

What? Won't we atleast get opengl drivers that expose the features of these DX10 cards on XP?

This is something that only the IHVs decide. As in the past Microsoft can not stop them from including extensions to their OpenGL drivers.

 

[Anteru]

This is something that only the IHVs decide. As in the past Microsoft can not stop them from including extensions to their OpenGL drivers.

And with ATi and nVidia taking over control of OpenGL, I think it's pretty likely they'll create extensions so that you can make the most out of their DX10 parts under XP using OpenGL - after all, it's good for them when the large XP user base can make use of their shiny new boards. I doubt that they are willing to expose the full functionality of their boards only on Vista.

I hope that this will push OpenGL a bit into the foreground But it might turn out very well that OpenGL 3.0 will require Vista, too, as it seems that OpenGL 3.0 and DX10 will be much closer than ever before. Dunno what Linux/Mac OS version it will require though

 

[KimB]

I hope that this will push OpenGL a bit into the foreground But it might turn out very well that OpenGL 3.0 will require Vista, too, as it seems that OpenGL 3.0 and DX10 will be much closer than ever before. Dunno what Linux/Mac OS version it will require though

There's just no possible way that OpenGL 3.0 would be tied to Vista.

 

[Blazkowicz]

as usual people will get the new windows pre-installed on the PC they just bought. that's how the vast majority of people upgrade their windows.

and no, it's not a bigger than 98 to XP jump, more like 95 to 98
well, not quite cut-clear how to compare with previous upgrades, but really, it's still a NT kernel, built from its best version so far, 5.2.
once you have a modern kernel design well tuned over years there's not much room to improve. you can now maybe survive a graphics card driver crash, but how exactly often does that happen anyway when you don't have faulty hardware. instability is caused by hardware these days, not sure why a PC with bad PSU or RAM or dying vid card should run better on Vista.

 

[Demirug]

And with ATi and nVidia taking over control of OpenGL, I think it's pretty likely they'll create extensions so that you can make the most out of their DX10 parts under XP using OpenGL - after all, it's good for them when the large XP user base can make use of their shiny new boards. I doubt that they are willing to expose the full functionality of their boards only on Vista.

If the games use D3D10 an OpenGL driver will not help.

I hope that this will push OpenGL a bit into the foreground But it might turn out very well that OpenGL 3.0 will require Vista, too, as it seems that OpenGL 3.0 and DX10 will be much closer than ever before. Dunno what Linux/Mac OS version it will require though

In the case the make no clear cut with OpenGL 3.0 I am don’t believe that the will win many of the current D3D developers. Additional one of the reason why most developers use D3D instead of OpenGL on the windows platform is the better general driver support. I don’t think that a new OpenGL version will change this.

 

[stepz]

If the games use D3D10 an OpenGL driver will not help.

It will if someone will make D3D10 to OpenGL translation layer.

Maybe the Wine developers will do it. Theire making great progress with the D3D9 layer so far.

Then we wouldn't need Vista. Or any Windows for that matter.

 

[Rodéric]

...
Additional one of the reason why most developers use D3D instead of OpenGL on the windows platform is the better general driver support. I don’t think that a new OpenGL version will change this.

Except that fewer entry points in the drivers = less pain = easier to make it robust, which is exactly the point of OpenGL|ES 2.0, which I believe will be what OpenGL 3.0 will tend to.