You said "I think this also should silence those who blabber on incessantly about AMD having no R&D budget so they can't compete. They were first out with a next gen graphics API, HBM, and it's looking now like 14nm FF".
I don't know what you think "compete" means, but with microprocessors, it means performance results. You not only attempted to belittle people concerned about AMD's R&D, but attempted to rebut the point of AMD not competing by listing off two things that still result in AMD losing (API, whereas Nvidia expertly implemented superior DX11 multithreading and are currently winning in the only two DX12 benchmarks out there; HBM that, for all the hype, ended up in a product that gets beaten
in pretty much all performance results by Nvidia's product without it).
Then you stated one final claim (first to 14nm) that's both pure speculation on anyone's part, tells us nothing about the actual quality or relative timing of the competitors' products, and is the biggest jump to conclusion in this discussion.
Touting HBM and APIs as if the terms themselves mean anything while the GPU division factually isn't keeping up with their competitor's efficient and powerful designs while having the nerve to insult people who understand the importance of R&D in that predicament (can't afford to tape out new, more efficient designs; can't afford to redesign GCN on 28nm so that Fiji was actually an effective counter to GM200; can't afford to multithread DX11 drivers like Nvidia) is so wrong.
The only thing you can come up with is a massive leap to conclusion about yet another as-of-now meaningless claim based on nothing but your own interpretation of current rumors and limited information. And you treat that interpretation as if it ends the discussion on whether AMD have the R&D to compete even though we still know near nothing and have the events of 2015 demonstrating R&D's effect? Remember how baseless speculation and hype turned out with Fiji and HBM?
You're right, it's not the point because I never mentioned Nvidia selling more. I'm talking about the aftermarket 980 Tis hands-down beating the Fury X in all currently-available measures (4K performance, 1440p performance, VRAM, power efficiency, DX11 draw call efficiency, per-transistor efficiency, DX12 performance, bang for buck, and manufacturing cost).
Nice straw man attempt though.
More logical fallacies? Another straw man attempt and goalpost moving in one post. Excluding your assumption or ridiculous generalization about what people in this thread trashing the Fury X (which I suppose is me; nice of you to quote me directly) thought in the past, we'll start with the obvious: multi-GPU support regardless of vendor tends to suck these days and single-GPU enthusiast setups are far from pointless or for "peasants". I've never thought so, many enthusiasts have never thought so, and strangely enough, Nvidia have also tended to hold the fastest single-GPU title throughout the past decade as well, so I don't know what you're getting at with your imaginary goal post moving about multi-GPU, as if it's even relevant to the here and now.
But let's play on your field, the so-called ultra-resolution, multi-GPU enthusiast Arena:
http://www.techspot.com/review/1033-gtx-980-ti-sli-r9-fury-x-crossfire/page7.html
Techspot make a very fair point. Multi-GPU enthusiasts aren't going to be the types to run reference cooling or clock speeds (default profiles on any air cooler being prone to throttling). And when Techspot didn't, the 980 Ti SLI was faster than the Fury X Xfire solution in both FPS and frame times. It's a logical assumption for a group of people actually chasing all the power they can get.
Furthermore, said people may even get the AIO 980 Tis or put on their own custom waterblocks as well for even further overclocking and/or less thermal throttling. And the final elephant in the room: what if these people go even further beyond and get Titan X SLI under water and overclock those?
For these enthusiasts you refer to, the best option is still a GM200 (either 980 Ti or Titan X for the less value-oriented) SLI machine of some kind with non-reference cooling and overclocking whether you like or not. Basically, you're wrong.