right but if the 980 sells with large margins at a gm200 price they have no reason not to keep it at 500 bucks and keep the true high end cards 1000 bucks.
Are you going to claim that the 680 was a high end chip?
the 780 was gk110 with the tesla workstation functionality crippled, that was the high end chip
Do you really think gm200 is not intended to be the high end chip? There will be a gm 200 gtx1080 or w/e they call it next year, either at 20/16nm or still at 28 just like with the 780 and 780 ti.
If the 980 sells at 500 bucks, they don't have to care about how much or how little a (crippled or workstation equivalent) gm 200 titan 2 sells if they can just sell gm 204 at the high end price, the only thing that matters if consumer perception.
They have to convince people that a 980 midrange card is worth 500 bucks, that's it, giving gk110 the cuda performance to be a viable workstation card was a clever way for them to remarket gk110 (instead of people asking hey why are you suddenly selling the high end card for 1000 bucks)
A year later they then marketed and sold full gk110 as the high end , with the benifit of keeping the high end price inflated because 'it performs close to that titan that was basically a consumer version of the super expensive workstation cards'
I think it's pretty baffling if people will accept the 980 as being high end (and pay the price) considering it being barely faster than the 780ti... (with a smaller memory bus probably causing performance to suffer for multiple monitor setups in games)
At least with the 680 I could understand people not realising, it was a good 30 percent faster over the last gen 580 and was a strong overclocker to boot.
Nvidia are going to try the 680->titan->780 formula again for sure, we'll have to see how much of this crap consumers are willing to accept considering it doesn't even come with any noticable performance increase this time around.
@ seanspeed, progress slowed down? are you blind? (no offense

)
You are looking at the same thing I'm looking at right? A new architecture that with a 256bit bus and smaller die and much lower power consumption can match the big die and high end bus performance of last gen.
While TSCM and co are obviously struggling to keep up their schedules, nvidia is more than capable of not 'slowing down' with this new architecture.