Originally Posted by
anselhelm
"Not full implementation of GP102"
Thank you!
I posted something about that on Ars and got downvoted to oblivion, probably by zealous Nvidia fans or people who didn't properly read my post.
I simply do not believe that yields can be so low on GP102 now that Nvidia cannot release a "full-fat" GP102 implementation with 3840 shaders (30 blocks enabled instead of just 28) along with the max number of texture units (I think that'd be 240) and ROPs (96 I presume, unless they go beyond a 384-bit memory bus).
As such, once again it feels like Nvidia is deliberately holding back just so it can release a new Titan X (or whatever) for the same >$1000+ price tag and claim it's still the enthusiast's hero. What's worse is that people still defend Nvidia for these kind of practices, just like how people defend Intel!
I really, really hope that AMD isn't going to drop the ball with Vega or price it above Nvidia, just so that Nvidia has to actually compete a bit again. When there's no competition, we'll still get faster products, but they won't be as fast or as cheap as they should be.