Sure drivers are going to help, just look at the GF8800 series improvement thanks to their many driver revisions, performance went up about 20-25% in many areas (some of the reviews using similar systems for the G80 and R600 launches are testament to this).
However I wouldn't bank on it everywhere. There are somethings that are obviously not working well in reviews, and those will likely be addressed by driver improvements, but then there's other areas that seem to scale and react properly, those are unlikely to see drastic improvements. The thing is that the early tests are all over the map, with the HD2900XT beating the GTX in some 'specific' games and areas. So there's obviously alot more to this than the early handfull of benchmarks of the 5 major games people benchmark. One of big areas of performance problems is the shader-core based AA and mediocre AF performance. It's not known how mcuh this can be improved upon, or if there are fundementally broken parts to it.
ATi's history of driver improvements usually focus on better memory management, but non of the tests seem to be bandwidth limited (not surprising with 512MB), so unless they re truely being held back by a global issue/bug there should only be improvements to ensure full utilization at first, then after that expect a few percentage per driver update like happened with the G80.
I would wait until the first official update driver and see what the next few days brings to see what happens after all the hulla-baloo, but other then that, find the games you like and buy the card that's the best match.
To me the speed of ATi's response to the Frozen Panet demo will also be an interesting test, since unlike the other titles they claim to not have had much exposure to this, so it should give people and idea of how well they trouble shoot a 'performance bug' of noticeable magnitude (still not as bad as if Oblivion, WOW or FEAR, didn't work, but big enough and a good start point to see a fix in action).
Facebook
Twitter
Instagram