It seems in the latest years both AMD and nVidia agree on certain elements regarding model number:
1st number = generation (AMD vs nVidia are 4vs2 and 5vs4, 6vs5). AMD is about to release 7xxx, while I saw no rumour about nVidia's 6xx (if they'll call it that).
2nd and following numbers = relative performance and intended usage. 2nd number explained:
0-4(nv) 0-5(AMD) = budget (light gaming+video, but silent). Not realy for gaming.
5-6(nv) 6-7*(AMD) = mainstream (capable of native resolution gaming with low-to-medium details, some with additional power connector)
7-9(nv) 8-9*(AMD) = high-performance, most with 2-slot cooling
* - AMD seems to use the mainstream (6-7) chip for 68xx, but until know this is the only exception.
The difference between the 3 categories of cards are the cips, these being the 3 major chips architectures for each generation. For cards with the same chip the performance does not differ more than 20-30%, due to frequencies and disabled units. But an overclocked slow-chip will never exceed a faster stock chip. Frequencies are comparable only for same chips (or at least for same categories).
Oh, and about RAM, don't bother with the "more-RAM" models. Currently there are 512MB and 1GB mainstream cards, but buying the 1GB is waste of money, as the extra memory will be virtually unused. Or should I say, when you do need the extra memory is with: large textures, anti-aliasing and high-resolution. But the latter 2 require performance 1st. As for the large textures, they are mostly used in high-resolution (so again performance 1st) or maybe non-gaming usage (3D design). I don't even know why they bother. At those prices better get the high-performance cards.