by Leonard "Viking1" Hjalmarson

  Titus Software, Graham Rudd (F18 Super Hornet)

To be able to answer your question fully it would have to be tailored specifically to a particular piece of software. In general though, this is how I stand on the Transform and Lighting versus fill rate.

The one thing that you can guarantee all 3D pipelines do is a LOT of transformation and lighting calculations. This is the core of any engine, be it a flight sim or a first person shooter. Make no mistakes about it, today's processors can blast through these calculations incredibly quickly, but wouldn't it be nice if they didn't have to? Imagine all of the free CPU cycles that will be available to use for things such as AI.

The Rendering sequence in a game is typically around 60%-80% of the total time taken per frame, most of this is taken up with ( you guessed it ) the transformation and lighting.

F/A 18E Super Hornet

Fill rates on today's cards are already very high. If the fill rate is continually increased then we have to up the polygon count to keep the card busy and this of course leads to more complex models, more vertices and more calculations for the CPU to perform to keep feeding the ever hungrier cards, thus taking even more CPU cycles from the physics and AI etc.

I am personally all in favour of transformation and lighting on the cards. It will hopefully lead to a competitive drive in this direction between the card manufacturers.

Click to continue

3d Hardware Survey

Now you know how some of the major simulation developers feel about the fill rate vs. t&l issues. How do you feel?

Two weeks ago we created a simulation survey that allowed our readers to answer a few questions. The answers were revealing.

The majority of respondents believed that hardware T&L or hardware anti-aliasing and texture compression are the most important new technologies. The split between these two was nearly 50/50, with almost two-thirds of the respondents choosing one of these categories.

When asked which video chip would be in their PC this fall, the split was nearly 50/50 between those who chose Nvidia (GF 256) and those who chose 3dfx. Unfortunately, at almost the same time as we released the survey 3dfx revealed that their next chip will be delayed to February or March of 2000.

Interesting, since the GeForce has also been delayed, at least slightly, for unknown reasons. It seems that the major chip makers are re-examining their strategies and market position in response to one another. Will 3dfx add a t&l engine to Napalm? Personally, I think they will.

Almost one third of the respondents to our survey have no plans to upgrade this fall. Not a bad position, given the uncertainty in the hardware arena and the fact that there will be only a few simulations that can benefit from t&l hardware this year. For information on the few that we know will benefit to some extent, see our earlier T&L NOW. For a discussion on some of the recent hardware developments, see our September 3d Hardware Update. For a primer on Transformation and Lighting, see S3's White Paper.

Join a discussion forum on this article by clicking HERE.