Curious about something...

Having a big poly could affect performance negatively, yes. It would eat up fill rate. However, breaking it up into sub meshes would only make things worse because not only would it eat up fill rate, but it would eat up extra vertex processing too. So my advice to you would simply be, use the big poly if you have to, but definitely DONT break it up into smaller quads. (There may be some optimizations I don't know of, that would tip the scales the other way, but i sincerely doubt it)

cheers
-Dan

If you're planning to cover the entire screen, just use a single polygon. Fill rate for the entire screen is expensive, so if you plan to use transparency, only do blending where you absolutely need to.

Bear in mind that if you intend to also do lighting on it as well, four vertices won't give you a very nice effect. In that case you will have to split it up into multiple polygons to increase the detail of the lighting.

Well, here's what I'm doing: I'm using this big quad to render a background image. In fact I can have multiple "backgrounds" (such as a mesh of branches, vines, and leaves acting as a foreground in a forest setting, for instance, or multiple layers of clouds parallaxed together.)

Lighting isn't an issue - it's billboarded and I precalculate the colors to give all of my billboarded actors only once. It's essentially just one big textured billboarded quad.

I render it last, so I would guess the fill rate isn't TOO bad (the map already covers most of the screen) but if I can strip away some quads through an occlusion test, I'm not sure wether I'd see a performance boost or not.