If you’re trying to avoid the idea of rendering altogether, I wouldn’t bother. Other than basic opacity and alpha graphics you really need to render everything, and you’ll probably see a quality increase in the alpha graphics when rendered as well.
A better graphics card will give you more real time, but for anything with any meat to it, you’re still going to need or want to render to ensure playback out to tape. But for the convenience of the editing process and making editorial decisions, yeah, the better the graphics card the more real time and the less intermediate rendering neccessary to envision the final rendered product.
I’m running a quad 2 ghz mac pro with the standard card, and the new fxplugs they introduced that work with 5.1 and 6 and insanely fast compared to FXscript. Things like light rays and glows render faster than simpler effects like wipes. Crazy. We just did a project with a bund of sugarfx templates (which show up as a generator in fcp) which work in conjunction with fxfacory. These things are pretty amazing. Imagine a multilayered template (some in 3D space) that you simply type in your text and drop video clips into the video wells. It plays back nearly in real time.
Since I know I’m going to eventually render, I turn off the “safe mode” in the timeline and always always always edit in “unlimited.” You can turn off the dropped frames warnings and I find this is the way to edit. If a layered effect starts to stutter or get to where I can’t envision it, I’ll render a portion or drop the quality of the rendering or the frame rate, but usually I’ll just render it.