Fury was the same price/perf as other video cards. Fury X was the same price/perf as other water cooled video cards which had a huge $100 or so premium.
They have been targeting Volta not Pascal so why do you think the price/perf will be worse than Pascal? They knew they weren't competing with Pascal in the high end and that they were going to compete against Volta with their timeline.
I don't understand how people think Nvidia can price Volta at better than 1080 Ti for 1070 price but it would be impossible for AMD to do so as well.
AMD has the biggest change in their arch in a long time and they skipped a high end Polaris so they could get a high end only Vega ("next gen") out faster than Nvidia could (having to do both high/low Pascal and high again on Volta).
Nvidia's R&D budget is massive so they require tons of money in return to make up their ROI so they keep prices as high as they can as long as they can.
AMD just destroyed Intel CPU side. Only CPUs worth buying from Intel are the Pentium and i7 7700k if you only game and nothing else. It looks like Threadripper and Eypc (please AMD find someone better to name these!!) are going to do it again for high end servers as well.
They have been targeting Volta not Pascal with Vega. They are releasing top end first to help make up the R&D costs since server markets are the big $$. They'll spend what, $50 in parts for $400+ more in end markup? Same deal with Titan / Quadro vs xx80 Ti. Very similar arch yet massive price increase and thus profit margin.
They got the increase in market share and gains with developers / game engines they needed with Polaris and now are going high end with their new Arch.
Its going to be very interesting!
There's a lot of unfounded speculation here. While I agree that AMD must be aware that they'll be competing with Volta in a relatively short timeframe, does that necessarily mean that they can simply "choose" to make a faster card? And while the Vega NCU is the biggest architectural change for AMD GPUs since GCN replaced VLIW, that doesn't really tell us anything about actual gaming performance.
I don't understand how people think Nvidia can price Volta at better than 1080 Ti for 1070 price but it would be impossible for AMD to do so as well.
As of now, we have no evidence at all that Vega is more than roughly competitive with Pascal. Simply because Volta will be arriving soon, doesn't somehow magically increase the performance of Vega, nor lower production costs. Of course, the demos we have seen have been early hardware and alpha drivers, so again, we don't know where this will end up, and we have no idea about pricing either. But you're making assumptions that Vega can compete with Volta seemingly based
only on them being on the market at the same time. Which makes no sense. As for price/perf, Vega has a bigger die than GP102, and uses (somewhat) more expensive memory and an interposer. I.e. selling them at lower prices would probably mean noticeably lower margins.
Then again, if AMD has balanced the architecture better than Fiji (which they
must have, come on!) this would gain them a lot of "free" performance, and if they save R&D and fab costs by having fewer die designs (two currently rumored (probably more down the line, but I'd be surprised to see more than 4) vs. the 6 currently existing Pascal chips), this might balance out. The point is: we have to wait and see.