@raybay
So, I take it that you're not offering any substantive proof regarding GTX 470 failure rates, but rather extrapolating a possible future scenario based upon some previous architecture/s.
As a system builder I would steer a potential customer away from problematic cards, but if I used the criteria that failure rates in a previous generation (or particular series) of cards automatically equates to failure rates in unrelated new/upcoming architecture then I wouldn't be recommending anything outside workstation cards.
Cases in point- and these are both from my experience (around ten years of RMA invoices since the advent of mass-market affordable discrete graphics) and web based sources.
Failure rates in reference HD 2900XT, 3870, 3870x2, 4870, 4870x2 (mainly due to long term heat exposure), RMA rates of reference initial series HD57xx/58xx/5970 due to PowerPlay/BIOS issues, HD 4830 shipped with deficient shader numbers due to wrong BIOS loaded, the very real possibility that some early shipments of GDDR5 vRAM chips used in HD5xxx series cards are defective and the recall of ATI's TV Wonder 650 amongst others.....does this mean I should boycott (or instruct others to) v2 HD5xxx series cards, or the upcoming HD6xxx series and Northern Islands?
Likewise, I've had to replace (more than) my fair share of "Green" 9600GT/9800GT as well as assorted 6600/6800, 7900GS/GT/7950GT/GX2, 8800GT/GTX and first production run of GTX 280....How does that correlate to a card that seems to have both a negligable DOA percentage, and the fact that it's a high-end card (i.e. likely to get hammered in benchmarking/heavy gaming) which also seems to have a negligable failure rate?