Originally Posted by kithylin
Actually that's not true. Nvidia has released other video cards in the past where the top spec card of the new family is only +30% faster than the top spec card of the previous family (not including Titans) and they sold very well and they were released as a retail product. It is entirely possible based on the history of Nvidia products in the past that the new Ampere video cards could possibly only be +25% to +30% faster than Turing and still be sold as a successful product and be very profitable for the company. It does not have to be "50% faster" for Nvidia. They have no competition at the high end right now (As of typing this) and their new Ampere video cards will have zero competition on the high end when they release a few months from now. A smart thing to do from the business perspective would be to release new products only as fast over the previous one that people are willing to pay for. I'm sure if Nvidia wanted to do it they could release a card that is just +15% faster or +20% faster for first generation 7nm cards (Ampere) then the other +20% faster for the next family of 7nm (revised) products to milk more money out of their customers. There's really no reason to give us a single card that's all +40% faster in one gulp if they could split it up across two families for more profit. They may not do that this time around but as we get closer and closer to the physical limitations of silicon when they can't shrink chips any further (it's currently physically impossible to produce anything CPU, GPU, or Memory beyond 3nm for example) they'll have to stretch out their profits thinner and thinner at some point.
Except for Turing and Kepler refresh (GTX 780, same architecture) when did this happen? Going back until Fermi the generational leap has been 50%.
GTX 680, 50% faster than 580, as fast as 580 SLI
GTX 980, 50% faster than 780, as fast as 780 SLI
GTX 1080, 50% faster than 980, as fast as 980 SLI
GTX 980 Ti, 50% faster than 780 Ti, as fast as 780 Ti SLI
GTX 1080 Ti, 60% faster than 980 Ti, as fast as 980 Ti SLI
GTX 970, as fast as 780 Ti @ 200W
GTX 1070, as fast as 980 Ti @ 200W
With exception of of GTX 970, 980, 1070 and 1080 I've owned all of the cards in question (GTX 580M SLI and GTX 680M SLI).
Turing is anomalous because NGreedia renamed the entire product stack one GPU higher in an attempt to justify an actual doubling in asking price. FACT.
RTX "2070", TU-106 SKU (60 card SKU), no SLI (every 70 card up until here has had SLI, no 60 card has had SLI), only as fast as outgoing 80 card, not the 80 Ti card like every new 70 card before it. MSRP: $600, historical price of 60 card: $350
RTX "2080", only as fast as the outgoing 80 Ti card, not 25-30% faster like every new 80 card before it: $900, historical price of 70 card: $450
RTX "2080 Ti" see above. $1200 before taxes, historical price of 80 card: $600