A reasonable argument and well put forth ... however
1. The price is determined by what a) each company sees will sell well in the face of competition and b) supply and demand. As long as the 1060 continues to outsell the 480 and until nVidia can keep vendor in stock (which it has not been able to do of late), I don't see this changing.
If building new, if we gonna say cost is an issue, then it should be total cost .
a) the 480 needs a 100 watt bigger PSU (cards draw 200 vs 120 watts in "typical gaming") and that has a cost
b) to keep comparable case temps, the 480 will need an extra case fan (one 120mm per 50-75 watts)
c) the 480 will cost more to own.... I'll use "cheap power" here; where I am it's 2/4 times more... in Europe its 5 times more:
80 watts x 35 hours a week x 52.14 weeks x 3 years x $0.10 per kw hr / (100 watts per kwhr x 0.85 eff) = $51
2. Again, one card OCs 17.7%, the other 7.7%...that's not core speeds that's measured fps improvement. So if both cards do 50 fps outta the box, one will do 54 and the other will do 59 ... a 10% improvement. Should a card that's 10% faster be worth 10% more ? Usually as you move up in performance we see diminishing returns, to come out "even" is a big win.
As for Vulkan and DX12, ther may turn out to be something there but we are still in the infancy of both. With games having a 3 - 5 or even more year development cycle, today's DX12 games have been "adapted" for these new APIs and not developed from the get go with DX 12 and Vulkan in mind, As such, they are still maturing. But, again, we've seen this before. Anyone remember Mantle ? The mantra was "Mantle will change everything" but it became a historical footnote before it was really used by anyone. How about HBM ? AMD put out a lot of hoopla about HBM but HBM1 turned out again to be much ado about nothing. When was the last time we even heard about HBM2 ... last i heard was volume shipments in 3Q .. to who ? With GDDR5 and 5x, HBM2 isn't bringing anything tot he table, is in too short supply to be adopted and just isn't a viable option as yet.
http://www.tweaktown.com/articles/7830/hbm2-graphics-ram-hero-deserve/index.html
AMD was the first to have HBM1 technology on a consumer graphics card, with the release of the Radeon R9 Fury X reaching new levels of excitement - but, really - it fell on its face. Compared to NVIDIA's GeForce GTX 980 Ti which was released weeks earlier than the Fury X, AMD's graphics card lost against NVIDIA's best consumer graphics card at the time, even with the next-gen HBM1 technology.
But yes, AMD is showing better progress here but the Mantle experience leaves me an expectation of "Deja Vu all over again". AMD has often taken the early lead with new APis, but so far they have yet to maintain it for very long.
3. nVidia has a big issue with SLI. It's killing their bottom line. Too many peeps were buying two x70s, x 60s or even x50 Ti's and outperforming the X80s. They make a lot more money selling the x80 than the 2 of whatever. Scaling on the 1070 / 1080 is currently even worse than AMDs. But it would be a conflict of interest for nVidia to improve scaling on the 10xx series. With no competition right now to the 1070 / 1080 from AMD, the sales of only 1 card would be hurt by better scaling ... or adding scaling on the 1060 That would be the 1080 and they make a lot more money selling a single 1080 than to 60s or 70s.
Yes, you can CF a 480. AMD and nVidia back in the day allowed use of this technology on cards that weren't really up to it which is why we continue to see posts today that SLI / CF cause microstutter and a host of other problems. But scaling with the current 10xx and 4xx series is simply dismal. The whole idea behind SLI / CF is that in exchange for the increased power / heat from dual cards, you get more fps than the flagship card at less money:
Two 560 Tis ($400) gave you 40% more fps than the 580 ($500)
Two 650 Tis Boosts for $300 gave you more fps than the 680 ($580)
Two 970s gave you 40% more fps than the 980 for the same price.
So the idea is... SLI / CF can be easily justified if a) the cards are at a level of performance that they don't experience problems and b) the 2 cards must perform better and cost less than a single card alternative. That doesn't happen here with the 1070 or the 480.
At the OPs resolution, you are looking at 39% average scaling across TPUs 16 game test suite. For the 970 it was 70%
But the kicker is this ... relative to the 480:
- a 2nd 480 gets you 39% more fps for $500.
- a 1070 gets you 150%, more fps for $410
All combined, the relative value, using same Brand / Model line (MSI Gaming X in this instance) is
139% x 1.077 OC / $500 = 0.299
150% x 1.177 / $410 = 0.431
That gives the 1070 44% more fps per dollar than the twin 480s.
You've made very good points, ones that should be considered, but when ya dig a bit deeper and consider all factors beyond "out of the box" fps .... I think the 480 holds 'all things considered". However, if you don't OC ya GFX cards, Mom's paying for the electric bill, you already have 8 fans in ya case and an 850 watt PSU, then I really can't make an argument against either card.
Everything below the 480 / 1060 is AMD all the way ... everything above is nVidia all the way. The 480 / 1060 is a battleground in a price / performance niche that they haven't been very competitive in a long time. The 380 was a clear winner over the nVidia offerings in their price / performance range and this time around, you can make a case for both cards. Which way you go will depend on:
a) New Build or old build w/ excess PSU capability and more than adequate case cooling
b) Tolerance for noise
c) Where OCing or not
d) Local power costs