lehpron
In the past, twice in fact, AMD's flagship was aiming at Titan performance at a cheaper price, but nVidia would preempt the move at the last minute (usually a month ahead) to fill the gap themselves.  Such a move always made the x80 card cheaper to let the x80 ti card take the original debut price of x80.

But, this time is different because we are getting early rumors and reports and claims than Vega 10 is closer to GTX1080 than Titan-X Pascal.  So on one hand, there wouldn't be a reason to bring a GTX1080 Ti since their Titan isn't being challenged.  But, if AMD played their cards right (no pun intended) and priced Vega 10 cheaper, then not only does nVidia react with a price drop on GTX1080, they have to debut something faster to preserve the premium expected for Titan-X Pascal to remain at $1200.  So maybe a GTX1080 Ti would come, but around $800-1000.

I think people misunderstand the TFLOPS compute values, that because Vega 10 is supposedly at 12.5 versus a full 3840-CUDA GP100 at just 12.0, then it would imply that Vega 10 is faster than anything nVidia has available for gaming-- i think that isn't proper logic.  

Vega 10 is said to be 40% faster than Fiji, which if it has the same 4096-sp, then the clock would have to be up at 1500MHz-- very possible for 14nm, sure, but it would need >60% to chase after GP102.

I don't believe AMD's flagship GPU is going after nVidia's flagship GPU this time, rather the 2nd tier, which means a x80 Ti product will be expensive and not in the same ranges we've seen 780 Ti and 980 Ti at.  That also means there isn't a hurry to debut it since AMD isn't going after nVidia's best.

I agree with this, which is somewhat regrettable since many, like myself, are unwilling to pay the money for a Titan X and Vega is the first real chance for that price to come down. It is also possible the 12.5 TFLOPS model uses binned cores, which it undoubtedly does, which means this is what we would see from the full chip going as fast as it can, using as little power to achieve said speed. If that is the case, the consumer cards are going to be more down to Earth for a current gen GPU, which is basically what this is since AMD never released a GPU that exceeded their previous generation of cards in every aspect. So, even though 12.5 TFLOPS is impressive and not to far fetched for what the current generation is achieving, it is likely not indicative of all cards while GTX 1080s 8 TFLOPS is likely a more indicative speed for the consumer models.
 
EDIT: I forgot to mention the size of the die on Vega. The amount of silicon in ONE GPU is insane. The die on one Vega chip is over 500 mm2 alone, that is just the GPU not including the memory. The original consumer Pascal die is just over half the size and is the GP 104 core used in the GTX 1080 and GTX 1070. If Vega competes with the 1080 with almost twice the silicon in the core alone, the price will be high. For comparison, GP 102 die is 471 mm2 and GP 104 die is 314 mm2. I would rather see the Vega 10 chip compete with Titan X since that uses the larger GP 102 core that is still smaller than the Vega core, while it competes with GP 104 which is far smaller. I don't expect to see Titan prices drop as a result of this until closer to the launch of Nvidia's Volta GPUs since that will exceed current gen GPUs more easily than a current gen GPU being called next gen because it took an extra year to make it equal to a GPU that came out before it.