There are 2 main and independent issues that slow video boards down, the geometry processing and the pixel fill. Each board has only so much capacity of either one.

Of course the geometry has to do with how many objects and how complex they are, are in the scene. If you look in a direction with say, more trees and buildings in the distance, and the scene slows down it's probably a geometry limitation. Geometry processing is split between the CPU and the GF2 board. So, a CPU can help, up to the point where the GF2 geometry pipeline is full - then you are stuck at that speed. I would be surprised if an FS could have too much geomtery in the scene.

The pixel fill rate of the card is basically fixed - and that's exactly what it sounds like - the rate that the card can compute and output pixels. What uses more pixels? Higher resolutions use more pixels, FSAA uses more pixels, alot of things stacked up one in front of the other on-screen use alot of pixels. The way to test if it is a pixel-fill limiting your speed is to make sure FSAA is turned off, then compare speed at 2 different resolutions. If the lower resolution gets faster - its a pixel-fill limit. The only thing you can do to help this is to increase the rate of the on-board clock, which genereally isn't a good idea. Or, just run at the lower resolution. Upgrading CPU will do nothing if your problem is pixel fill.

What's a GF4 board going for these days?? $50 bucks off the shelf? That's the real solution.