Yawn. Did you read what I wrote? I said I have yet to see them... translation: it's never been the case when I measured (sans the few cases where twice the framebuffer mattered.)
Except GTX470 is disappointingly slow, at least here, in my i7-based 8GB WS - maybe you're runnign something faster than i7...?
GTX470 costs the same as a 5870 yet it barely matches a 5850...
...and on top of it every 5850 I've ever seen overclocks better than a GTX470.
Unless you need NV (CUDA) it's the worst possible choice IMO.
Ahh yes, T2k and his magical rig :-/
Every reviewer in the world tested the HD5870 in numerous games (old and new), using several resolutions and both AA on and off and all of them came to the same conclusion (backing it with pages of data) - in most cases the HD5870 is slower than a HD4870x2 - except for games where CrossFire scales bad (like Crysis). So unless you happen to play just those few games with bad scaling or you're playing some very "special" games nobody else knows of, your statement has no value.
I can show you a game where a GTX285 beats a HD5870 - it doesn't mean that's the case in general and that's something you should be using as an argument when recommending cards. I could write "in all the games I have played the GTX285 is actually faster than a HD5870" - I didn't lie (as the only game I played was the one where a GTX285 beats a HD5870), but that would be a worthless statement at the same time and wouldn't add any weight in a discussion or decision at all.
And just a small tip - 2GB on the HD4870x2 is actually 1GB per GPU, so the last gen red card has the same framebuffer size available as the HD5870 (so 1GB in total). Unless you had the 4GB model (if there ever was one?), you're wrong, again. 2GB is physically on the card, hence they advertise it as 2GB, but actual usable size is 1GB and that's just AMD's BS you're being fed (nVidia does the same, so who cares, right?). Don't believe me? Read up on how CrossFire works.
Now, read the next statement carefully:
I told the OP to get a HD5870 and only if that's not possible get a GTX470. Meaning, I know the HD5870 is a better value card. And "barely matches a HD5850" is just flat out wrong. Again, I'll point you to every reviewer out there saying the GTX470 is indeed slower than a HD5870 but it's closer to it than it is to a HD5850. So how is that barely matching the HD5850? It must be your magical rig... red electricity flowing through it or something?
You wrote "no OC" for the GTX470 and I showed you that the card OC's nicely. You used that statement in your argument and I showed it's wrong. Hot and power hungry? The idle and load power numbers for a HD4870x2 are even worse. So it's not a big deal deal for the OP anyway.
Finally, he can select a card from Newegg to get as a replacement for his dead HD4870x2. If he gets a cheaper one, he won't get the money difference anyway. So why would he get a slower HD5850 if he could get a faster GTX470? It's not logical at all. Of course if he can get a HD5870, that's the best solution. But if that's not possible, a GTX470 is the next obvious, logical choice. You want to tell me, when presented with a "free" card, a HD5850 or a GTX470, you'd go with the Radeon? Really? You know what? Maybe don't answer...
EDIT: Okay, the GTX470 is right there in between the HD5850 and HD5870. Doesn't change the fact that choosing a HD5850 over it with no extra money required is a bad decision imo, as the nVidia card is faster and offers at least the same technologies (unless you want 3 monitors).