I can't find a single benchmark to back up that claim. They don't even seem to be close. Perhaps you could share one. Or perhaps you have the 5850, 6850 and 6950 confused?
5830 -> 6850
5850 -> 6870
-> 6950 (class by itself)
5870 -> 6970
5970 -> 6990 (as yet unreleased)
In most benchmarks (not all) the 6950 even beats out the 5870. Not by much and not always, but yes. Given the extra benefits of the 6950 I'd happily call it a wash. If that is "Marketing BS", then so be it.
First: I must say "sorry" -- as I was talking about the 5870 vs the 6950 as in that original post I was replying to. I simply mis-typed it (again, sorry for confusion).
I wouldn't believe those benchmarks entirely... The chips are almost the same from what I've read, because the design they intended wasn't ready.
You look at the pure raw computing power of the cards (not the in-game benchmarks) and the 5870 noticably outperforms the 6950:
Card: 5870 | 6950
Core: 850 | 800
Memory: 2400 | 2500
memory bandwidth GB/sec: 153.6 | 160
Shader operations/sec: 1,360,000 | 1,126,400
Pixel fill rate MPix/s: 27,200 | 25,600
Texture fill rate MTex/sec: 68,000 | 70,400
They trade off in a few areas, but the areas the 5870 has over the 6950 are by a much larger lead than vice versa. At least with the 5870 you are getting a tried and true gaming card with tested and known performance capabilities and driver stability.
So you're going to pay more for a new label with a larger number on it, not get any performance on top of the older card, and have to worry about drivers issues the whole time?
Meh... Pass!
P.S. the way I heard it, a better comparison would be:
5850 -> 6950
5870 -> 6970