i have a geForce 256, and i want to know how many triangles per frame i have to restrict myself to in order to get reasonable FPS ( say 60FPS). i''ve also noticed that when rendering on a certian resolution, doubling the triangle count does not decrease the FPS by two, but by rather less. on 1024x768, 968 tris are rendered in 340 FPS, and 3524 tris are rendered at 290 FPS.

Well, what you have to do here is something called optimizing.

The Way I know how to do it involves quadratic regression using a TI-83 calculator. I''m sure you know someone who has one.

What you do is take a large buffer filled with about 10,000 trianles. Turn backface culling off. Render the triangle buffer with about 100 triangles and set your program timer to take a framerate count everyminute. Every two minutes your program averages the last framerate and write it and the triangle count to a file. It will also increase the number of Triangles by 100.

When you are done, you should have a large file with values that you can use to speculate where you should be.

You can stop there, or you can graph the data. I would only graph 5 values before the 60 FPS and 5 after. Then you can use the Quadratic regression capabilities of the calculator to graph a curve. You should be able to find the minimums and maximums and a better value for 60 FPS. Whatever you want to minimze and maximize should be the Y coordinate.

All this (which is very simple) or you can goto www.nvdia.com and chance at them having the specs for your card.

www.nvidia.com ...damn vowels, they fell off the page!

15Mtri/second, what does that suppose to mean? at what res, with/without texture/gourad/lighting/flat/lighting/bumpmapped/envmapped

where could i get that info without calculating (i''ll do that if i have no other choice)

Triangles Per Second:

The rate at which a GPU processes triangles. It is a common industry metric for describing performance. The higher the number of triangles per second, the faster the GPU.

i am not an idiot, i know that

but the information the manufacturer gives is useless, enviroment mapped triangles take more time to render than flat shaded ones, and also rendering at a certian resolution has it''s own impact on the FPS

guess i''ll have to calculate this on my own

the res change causing the fps change is todo with fill rate, with SDR RAM you''ll hit the fill rate earlier than with DDR RAM

Also,I think that most Manufacture quotations are only theoretical, but you can usually get values for flat shaded/textured etc.
Of course then you have game logic to worry about, as well as actually feeding the geometry to the card.

that is why i wish i had a programmable card, i could then write the BezierPatch code on the card, and then only send the control points, for example

right now i try to use as much strips and fans as possible

quote:Original post by Anonymous Poster
Well, what you have to do here is something called optimizing.

actually thats called benchmarking.