Because discrete sales are built on $1000 cards ? You do realise that the consumer add-in board market is damn small to begin with compared to OEM contracts, and $1000 (or even $400, $500, $600+ for that matter) sales of consumer boards is minute in the greater scheme of things.
Here's a breakdown of Mercury Research's market segment analysis from three years ago. Note the sales of entry level boards
Mainly because there is a general suspicion of cards that have been run hard in mining. If you're happy whatever guarantee offered mitigates any downside then go for it. Personally I purchased a Gigabyte GTX 780 GHz Ed. on Ebay (new, sealed) for $US440, and it is on par with a 290X for my usage scenario. I wouldn't consider "upgrading"to either the 290X or 780 Ti, much like 99.99999% of people shopping for a graphics upgrade
Are you sure you understand the concept of brand? If Nvidia made the bulk of its revenue from consumer GTX 780 Ti/Titan Black then you'd have a point. The fact that most sales are to OEM's on thin margins, and professional cards for the cream, tends to marginalize the narrow focus of your attention.
Like the vast majority of buyers give a shit about architecture? Show me a single person who could differentiate the cache or scheduling structure in a GPU architecture, and I'll show you a thousand who buy based on marketing and "bigger number = better performance".
The combined might of Intel and ATI/AMD were supposed to revolutionize physics with HavokFX as I remember. Still waiting...
So you're singling out PhysX because it's gimmicky, Nvidia, and is...actually used?
Why not rant about the development of Bullet ? or Tokomak ? or ODE ? Havok is used in a lot more games than PhysX and doesn't do appreciably more (less in a lot of simulations) than Nvidia's (also) proprietary engine.
Says the guy who's specs list SLI'ed GTX Titans. You don't think you maybe answered your own question?