G A M E S P O T
 
By Loyd Case
Designed by Ethan O'Brien


The day before Microsoft's Gamestock event, I sat down with Kevin Bachus, program product manager for DirectX. We spoke about DirectX 7.0, and he shared a little insight into next year's DirectX 8.0. I noted that things had been fairly quiet on the DirectX front. Bachus chuckled and said, "Everyone's been saying that, though I'm not sure why. I think it's because we didn't have a Meltdown event in February." (Meltdown is the event where software and hardware developers get together and test their products on the newest version of DirectX.)

DirectX 7.0 is taking shape very nicely. Bachus noted that Microsoft was still on track for delivering the first beta in the March/April time frame, with the final product shipping by mid to late summer. "It's pretty much the same schedule as any other year," he noted.

Most of the development work for DirectX 7.0 is taking place on beta versions of Windows 2000. In fact, Windows 2000 is the target platform for game developers. "The goal is for developers to use Windows 2000 for the development platform but still target the titles for Windows 95/98."

Direct3D Features
Many of the changes to DirectX revolve, as usual, around Direct3D. The goal is to further improve performance for Direct3D, particularly texture-download performance. In addition, Microsoft is working to significantly improve texture management. "When we asked developers about 3D features back in the early days, they told us they just wanted us to give them access to the hardware, and then get out of the way," Bachus said wryly. "Then, later, they came back and told us, in effect, 'your texture management sucks.'"

There's a lot of effort being made to streamline Direct3D (and all DirectX development). The goal is to "make game development more fun and less work."

Two of the biggest features to come out in DirectX 7.0 will be hardware acceleration of lighting and acceleration of transforms and extensions.

Right now, DirectX immediate mode passes rendering chores to the graphics accelerator. But geometric lighting and transforms are still done by the host CPU. Bachus suggested that most of the benefit would come from accelerating transforms. The hardware lighting implementations that Bachus has seen seem limited to a handful of hardware lights - eight or less. A game like Messiah might have 32 or more light sources. Vertex transforms, however, would benefit greatly, but care must be taken because moving data off the graphics card and back to the system can cause performance bottlenecks.

DirectX Extensions next