You know Crysis doesn't come out until September right?
Do you know how many graphics cards are going to come out before that happens?
There's the R600, 8600 Ultra, 8600GT, 8850 possibilities, 8800GT?
And you don't even know how well the 8800GTX will run Crysis. What if DX10 on Crysis is so demanding, people with 8800s have to run it with DX9 shaders to get good frame-rates?
I understand what you're saying but I think you're a little misguided, even though you may be correct in a small way.
Outside of what I just wrote, I have a sneaky feeling Crysis may well get here just a little quicker than we think, as Microsoft are now putting some serious weight behind the game.
Anyway, DX10.
You have to really understand what DX10 does in comparison to DX9 and previous to see why you may be incorrect. DX10 has the potential to bring enormous performance enhancements with no loss (indeed an increase) of visual quality. Developers will however not take advantage of these properly at this stage - it will be mostly crudely implemented (like FS-X) to introduce some stunning new visuals at a high performance cost.
Once properly implemented and developed, DX10 will be far more efficient than anything previous - more importantly the hardware will be in a better position to take advantage of it. Programming for both DX9 and DX10 at the same time will certainly be an annoyance for a couple of years yet.
I strongly suspect that Crysis (as cited in the media) is suffering heavily from being a cross API game. They cannot properly implement DX10 (no time) and they cannot properly implement DX9 (doesn't look improved enough) so the hardware will suffer.
This should not at all detract however from what true DX10 hardware with proper programming can do - it WILL be a huge step forward, nothing like previous steps like 6 to 8, 8 to 9 etc. I strongly believe it will be a good 15 - 18 months before we really start to see what DX10 can do, and more importantly, see how little system overhead it can actually cost.