Unlikely in the extreme. Nvidia at the very least have an internal cadence they need to continue. HPC and workstation - Nvidia's large dies do double duty, and the largest die is primarily a compute chip. Secondly, the same architecture is now applied to SoC's - so reason two to keep arch cadence, as is OEM requirements - OEM's make up the largest buyers of discrete graphics chips - no new bullet points means people don't upgrade - you can only fool some of the people some of the time with rebrands.....and the primary reason? Shareholders.
Using your reasoning, Intel should have shelved Ivy Bridge/-E, Haswell/-E, and Broadwell...and Skylake for all the competition AMD offer.
Because all use 28nm wafers that are finite commodity as far as allocation goes. Why would you spend a second thought about producing old 367mm² GK 104 chips when you need wafers to produce GM 206 and GM 107 ? Likewise, 551mm² GK 110's make the same zero sense when the company need to allocate wafers for 398mm² GM 204
and GM 200 production.
Some people might say that that is rubbish. Intel innovate just enough to keep people upgrading. You really think "best chip regardless" philosophy covers decreasing the PCI-E lanes available to the 5930K ? or deliberately keeping separation between mainstream platforms and HEDT ? The 4790K is the price it is because of the product stack above it, and the number of people who buy $1K processors aren't by any means the majority of users.
Sorry if the logic of the situation and business in general is unpalatable to you....
Meanwhile a $339 GTX 970 basically equals the gaming performance of a $450 R9 290X - does that class as defending my pocket book?
Within Nvidia's GPU hierarchy, GM 204 is indeed a mid range (second tier) chip. GM 200 (or GM 210 as some people are calling it) is the high end. What frdmftr isn't considering is that Tahiti's successor (Bermuda) will also be a midrange chip, since Fiji will be their large die compute chip.
What frdmftr is highlighting about Nvidia's lineup is exactly what AMD are now in the process of doing. Fiji has no analogue in AMD's current lineup. The big chip of each (GM 200 and Fiji) will be heavily compute orientated - IIRC both are likely to feature 1:2 FP64, while the midrange (and lower) starting with GM 204 and Bermuda (
the chip associated with the R9 390X) feature die space area-ruled for gaming workloads. frdmftr can pillory Nvidia to his hearts content, but the fact remains that AMD are following their example because bifurcating the product line is the most effective use of expensive silicon production.