Over the past few days I have read numerous articles on the web that compare the Fujitsu Pinolite to the 3Dlabs GLINT Delta. This is a totally illogical comparison to make. Although both chips carry the "geometry accelerator" designation, they have very few similarities. In fact, their similarity ends at the fact that they are both hardware attempts to accelerate 3D rendering.
In the future, however, they may have one additional similarity. The Pinolite may join the GLINT Delta as a product that made big claims but did not provide the performance that gamers need or receive the support it needs.
The Pinolite could be used in a wide variety of ways. In a fast system with a 3D accelerator that is not fill-rate bound, such as a Pentium II 300 with a Voodoo2, the Pinolite could be used as to perform collision detection while the main CPU was feeding graphics information to the 3D accelerator. On a slower system with poor floating point performance, the Pinolite could be used to perform some of the final lighting and clipping operations. If used in conjunction with a 3D accelerator that does not perform triangle setup, the Pinolite could do the same setup operations that the GLINT Delta does.
In fact, this flexability may the Pinolite's downfall. The Pinolite, and similar products that may be available in the future, should be able to provide performace improvments to nearly the whole range of PCs. However, the way that a developer (or driver writter) will need to use the chip changes with the speed of the host CPU. A Pentium II 400 based system would want to use the Pinolite differently than an AMD K6 200. Now instead of having to support many devices in the same way, the developer is forced to support the same device in many ways.
2. 3Dlabs product overview for GLINT Delta.
3. Fujitsu press release 1997-0139, July 2, 1997.