It's definitely a very interesting card. If AMD ramps up production of the Fiji based cards - it would be a very compelling product at $499-549. Maybe someday..
I think AMD needs positive PR / word of mouth advertisement from excitement of having 1 solid product that punches above its weight class. During HD4800 both the 4850/4870 cards accomplished this. During HD5800 series, again both the 5850 and 5870 cards did. During HD6000 series, HD6950 was the star of AMD's line-up offering performance far above its asking price once unlocked and overclocked. During HD7000 series, HD7850's overclocking and HD7950's overclocking made them stellar values. It was possible to beat HD7970Ghz/GTX680 with an overclocked 7950. With R9 200 series they had R9 290. With this generation, they do not have such a product.
I think AMD should have priced Fury X at $549-579 because GTX980Ti is a better product overall. After that, they should have decided which card would be that HD4870/5850/6950/7950/R9 290 successor and priced it at $449. This could have been the Fury or the Nano. If they had done so, it would have made a lot of people think twice about buying a 980 or spending the extra $ on a 980Ti. As it stands though, both the Nano and the Fury X are going to be niche cards while the Fury is priced too closely to the 980Ti and too far above the 980/390/970 cards to make sense.
I guess AMD cannot afford to sell those cards at lower prices and while supplies are low, they can keep their prices high but I don't know how long this can continue as their market share will keep dropping more and more.
That says A LOT about where HBM can take this industry. Imagine if Maxwell had HBM. Faster than Titan X, while using less than 200W, while still on 28nm. What does this mean for 16nm?
Still not worth the price, though.
Exactly. The Nano is
78% faster than the R9 280X at Computerbase.de at 1440P HQ. And people said HBM did 0 over GDDR5. Since AMD didn't have a completely new architecture, it is because of using HBM that they were able to increase the die size and reduce power usage/die space over GDDR5. With 2016-2018 GPUs we will start to see the true benefits of HBM with 1-1.5TB/sec memory bandwidth GPUs, new GPU architectures, etc. If Maxwell had HBM instead of GDDR5, it would have been insanely impressive and that's why many expect amazing things from Pascal. I am honestly expecting 80-100% increase in performance over GTX980Ti in the fastest Big Daddy Pascal chip because NV will have HBM2 + 16nm node + new architecture. The last time NV had a full node shrink and new architecture, they doubled the performance moving from GTX580->780Ti.
I presume Fiji/Fiji XT will be shrunk & improved to the mid-range segment next generation and replace existing R9 390/390X level of cards so that perf/watt and size of cards like the Nano would travel down to $300-400 segments.
I think the Nano is priced too high but its performance is miles ahead of the fastest miniITX GTX970 level card. Because it's a fully unlocked Fiji XT GPU, with increased powertune and overclocking, near Fury X performance is possible in a 6 inch videocard.