No, the price of the 780 will likely be over $600, and that's a very conservative estimate. Still, I can't imagine that perf/$ will be much worse than the 680 at launch, as TSMC's production woes will hopefully be much less pronounced.
That's exactly my point. If you have a GTX580 or HD6970 and you waited all this time, then you'll argue that the price/performance for GTX780 has improved over the GTX680. Of course it did since you waited almost 2 years to upgrade and GPUs get faster with time. What about the guy who upgraded from HD6950 or GTX560Ti to GTX680/7970 and got 75-100% performance increase this generation?
Your argument is more about wanting a substantial performance increase for $500-600 than die sizes per say. If GTX680 was a 500mm^2 die GPU and cost $500 and was just 35% faster than GTX580 (but the extra die size was used exclusively for GPGPU and compute), would you have upgraded to it then? It sounds like you just think the performance increase is too small and you use the die size to validate the point that GK104 is a mid-range GPU and that NV held back the real flagship. Sure, that's probably true. But at the same time it also means the longer you wait to upgrade, the more price/performance you get. So then when do you finally pull the trigger (when the card is 50%, 75%, 100%, 500% faster)?
You realize when you upgrade to GTX780 for $550-600, the HD7970 OC / GTX680 owners may just skip that and get Maxwell/HD9000 series because to them a 40% increase for $500 may be poor price/performance upgrade? What you are talking about is upgrading when you think the asking price is worth a certain % increase in performance. This varies from 1 gamer to the other. This is not a die size argument but more of a Moore's law / technological performance increase per $ spent argument. You think this generation didn't provide sufficient performance increase from last, which is fair. Actually, I agree with you that this generation is underwhelming compared to the GTX580. But if GTX780 is a 500mm^2 GPU and is only 40% faster than a 1.15ghz HD7970/GTX680 OC, it's just as much of a 'fail' using your train of thought as GTX680/7970 were compared to the 580. To you it might look like an 80-100% faster card for $550 though.
Since I upgraded from an HD6970, I got a ~60-75% increase in the programs I use. I think that's fair considering after reselling the 6900 card, my upgrade cost was a fraction of the asking price. So even if GTX780 is 40% faster than the 7970 OC, it'll actually be a smaller increase for me. That's the thing -- it also depends where you are on the GPU upgrade cycle if you think the new generation is worth it or not.
Alternatively, you could buy AMD GPUs and bitcoin mine on the side, OR resell your GPUs frequently and incur a smaller cost of ownership through depreciation. There a some small steps that can be taken to minimize the upgrade costs. Buying $500+ GPUs and holding onto them for 4-5 years isn't a good idea though.