I understand your frustration. Believe me its VERY frustrating for us too. The PC market is segmented worse than my hardrive.
Today's developers have to do a lot more to make sure the games cover the enormous gap from lowest end to highest end machines that are actually in use for the given type of game.
For a brief period of time Windows made things easier because it was handling certain driver issues that game devs used to have to write into their own code. Now the expanded hardware possibilities have lead to a much greater chasm.
Unfortunately, even with the most scaleable engines out there, there is only so far you can go. Even if our engine could scale to 0 polys there's a lot of stuff to calculate outside of the renderer (AI, pathing, collision ...). There's not much that can be scaled in that part of the code without risking folks on low end machines playing an essentially different game.
The situation is made much worse by the fact that sooo many developers see fit to release amazing looking screenshots of games that will be at least 2 years away. These games are then presented as coming soon and rabid gamers everywhere hold everything that really is coming soon to those products for comparison. We (Shiny) have been guilty of the this too. Its not that devs do it on purpose, we just have a _very_ difficult job that is one part science one part art and all too hard to predict. We have to make experiences that are fun to play and dress them up so richly that they stand up to the highest expectations just to get noticed. Its a lot of fun, but really distressing too because we see how much it hurts the gamers. I think we'll be much closer to the mark with Sacrifice in terms of releasing soon after we've started the hype.
We are working on solutions, but I don't think the problem disappear unless the PC market changes dramatically. For instance: Software rendering had to go away so that we could develop just one set of graphics that took advantage of the subset of cards with texture filtering, alpha modes and on and on. What kinds of things will tomorrow's games have to ditch so that they can fully and efficiently take advantage of T&L, T-buffer and the next big thing? The more we try to use the new tricks the harder it is to use the old tricks too. The same textures that work with bump mapping don't necesarily work without it. Ugg, its just a nasty, nasty problem.
That's my rant for today, sorry.
-joby