Bear this in mind, the graphics industry has nowhere else to go, but up the 3D pipeline. Again, it's really that simple. I distinctly remember being in a meeting at Siggraph some time ago, and there was someone there from Computer Graphics World, a venerable magazine, and she refused to say any of the PC graphics chips had 3D rendering, although that's the way they were marketed. She called them rasterizers, as if to make a point. Well, that point has been made, rasterization was really the first stage of the development of consumer 3D. It's been done to death, and it will evolve, but there's no revolutionary movement left there. Now, it's the next level up the pipe.

And that's why T&L is a must for the graphics industry. Game developers will eventually get around to trusting the hardware in the market, as they did with the initial 3D rasterizers, some of which were very untrustworthy. OpenGL has supported the full implementation of the 3D pipeline in hardware for some time now. So, there's trust in the higher levels of the 3D pipeline today. However, if you look at the OpenGL graphics market you'll see a disparity between vendors there, too. Intergraph and HP use high-end geometry, as does SGI in its workstations, but their actual beneficial use by applications developers is not overwhelming. Even in the more stable, staid world of engineering workstations, the issue of how much of the 3D hardware pipe you trust with your scene data is subject to debate.

3Dlabs has adopted geometry acceleration as a matter of course, and it fits neatly into their product line, and strategy. Yet, Evans and Sutherland chose not to. Granted, Evans and Sutherland is not doing as well as 3Dlabs these days, or so the rumors go, but that has less to do with the 3D pipeline than Evans and Sutherland's execution of a business strategy. If there's fragmentation at the high-end of the market, there is bound to be even more in the consumer space.

3dfx, ATI, Matrox - no T&L. Ultimately, if something is in DirectX, and if only one company shows it can be done in hardware, everyone else will have to follow. I personally think that S3 and Nvidia are right to integrate T&L at this stage, even with little software support because, by the second or third generation of products, they'll have enough expertise to take on inevitable challenges from the other vendors that will be forced into adopting geometry. At the same time, I don't think companies without T&L will pay an enormous penalty, not for some time. It's all about timing, and I don't actually think that game developers are in a position over the course of the next eighteen months to slip and slide up the 3D pipeline.

Game developers have enough on their plates. There's 3 existing consoles, PlayStation II, Game Boy Color, online gaming, and maybe, just maybe, an X-Box. In addition, development costs continue to go up, and that's forcing game developers to spread their franchises across a greater number of platforms. That means they're going to hold on to their geometry for a while, and only release it on special occasions. There'll be some good stuff on a GeForce, but just as likely, there'll be some good stuff using T-buffer. Otherwise, the majority of game developers are going to have put their big development monies in making the most of franchise title releases on something like PlayStation II, or even, getting their Internet technologies honed.