Really? It seems to me that if you bought a 7970 back in early 2012 when it was AMD's new flagship, it would have held up pretty well in the long term. The 680 beat it by a few percent, but it wasn't until the Titan in Feb. 2013 that anything opened up a double-digit lead. Even then, the 7970 was still a viable enough card that AMD rebranded it as the R9 280X in mid-2013 with little criticism. It is still easily a playable card today. That's without taking into account the fact that you could have sold it during the cryptomining craze for almost as much as the original ($549) purchase price.
Nope. The better strategy would have been to buy dual HD7950s for $280-300, or even GTX670s, sell those earlier and with the $ saved from not buying 680s/7970Ghz upgrade to R9 290s/GTX970s, etc. HD7970 OC CF in no way shape or form outlasted HD7950 OC CF. You can check any benchmark you want. The same story with 680s vs. 670s, or 580s vs. 570s.
You can easily imagine a case where a gamer could have bought HD7950 for $280 and then an R9 290 for $375 in the last 3.5 years. That's way better than buying a $550 HD7970 and holding on to it. If we consider resale of 7950, it makes investing into the HD7970 even worse. You know how long it took before HD7950 was $280? By summer 2012.
Kepler wasnt "murdered", it was neglected and just fixed with new driver (353.06).
Proof that Kepler's performance was fixed? All I see are minor
10-12% gains in TW3. That doesn't even make a dent considering how far behind Kepler started in that GW game:
Reference thermal throttling R9 290X is 25% faster than 780Ti
Reference thermal throttling R9 290 is 32% faster than GTX780
GTX780 is 1 fps faster than R9 280X
Things are worse at 1600P.
In FC4, 290X is 20% faster than 780Ti, 780 loses to a 280X.
In fact, we don't need to go far. The latest reviews of GTX980Ti all show Kepler still performs poorly in the very same games it bombed before. Today
reference R9 290 > 780 and
reference R9 290X > 780Ti, while R9 280X is encroaching on 780/Titan.
No doubt. I picked up the 7970 when they went on sale after 200 series was announced (late 2013?) and still run it today. If I'd picked it up a year earlier it probably would still have been a solid choice.
Right but that's the point. You might have gotten 7970 for $300-350 which means you theoretically had $200-250 savings from not buying it on launch date. That's precisely why buying flagship cards and keeping them for 5 years is a waste of money. It's been proven every generation over the years on this forum. A gamer who can't buy a flagship card nearly every gen is better off buying $275-350 cards and reselling them every 2.5 years and getting a new card for $275-350 rather than buying a flagship $550-700 card and keeping it for 5 years.
Case in point: $649 980Ti is
70-75% faster than a $649 GTX780 and today we can buy GTX780 beating performance in only a $240 R9 290. Both of these events (980Ti destroying 780 and R9 290 beating $650 780) only took
2 years for that to happen. In fact, last winter there were deals on R9 290 after-market cards for $200-230, so 6 months ago!!
What is it that you think you're accomplishing with your consistent quote manipulation? I specifically said half way through the life cycle, and mentioned again in the very next sentence when I said "mid-life GPU upgrade" How did you manage to miss it twice?
So you are betting that 4GB of VRAM will become a bottleneck in
2.5 years or less at 1080P-1440P?
No official info is known about Fiji XT because it hasn't been officially confirmed/announced. You still went ahead and bought 980Ti SLI despite trying to come off objective/impartial for months. Ya right. No one
objective does that unless you are specifically tied to a GSync monitor or need some CUDA specific software or NVidia 3D Vision gaming or you have insider info that Fiji is worse for 100% fact. That's the whole point of the discussion that you seem to be evading at all costs. You never intended to buy Fiji even if it had 8 or 16GB of HBM because you didn't even bother waiting for reviews, its price, or a confirmation of its HBM capacity.
I guarantee it if Fiji beats 980Ti in performance at 1440P and 4K, you'll have some excuses lined up how in 2018 it might run into a VRAM bottleneck. If 980Ti OC beats Fiji XT OC and they cost similarly, of course 980Ti is a better choice due to 6GB of VRAM. But we don't know any of that until Fiji PROs and Fiji XT's launch. You didn't wait, which means you never even considered buying it; that's the point.
The card choice now will determine which monitor I go for G-sync/FreeSync, once I have committed to a monitor the likely hood in changing brands again in the next 5 years is very low.
I would think the monitor choice is way more important than the card choice. You'll probably upgrade at least 1 more time in the next 5 years as far as GPUs go but the monitor will stick around. If I were in your shoes I'd read reviews on various GSync vs. FreeSync monitors first.