The issue is just how good will the 3070 be... if the rumours are true (2080 Ti performance hopefully around £500) do we NEED more performance than that? I think 90% of us would agree and say that's enough... however I'm just looking back at previous X80 cards vs previous gen Ti cards it looks like the 3080 could be the best jump in recent history... let's start with Kepler... I won't be including clock speeds but rather pure specs of each GPU. Also my performance figures are at launch and not after a year where drivers have matured and there's new games etc..
GTX 780 Ti - 2880 cuda cores 7.1 billion transistors, 48 Rops 336gb/s bandwidth
GTX 980 - 2048 cuda cores 5.2 billion transistors 64 Rops 224gb/s bandwidth. 10% faster on average on release with 40% less cuda cores 36% less transistors 33% more Rops 50% less bandwidth.
GTX 980 Ti - 2816 cuda cores 8 billion transistors 96 Rops 336gb/s bandwidth.
GTX 1080 - 2560 cuda cores, 7.2 billion transistors 64 Rops 320gb/s bandwidth - 30% faster with 11% less cuda cores 11% less transistors 50% less Rops 5% less bandwidth.
GTX 1080 Ti - 3584 cuda cores, 12 billion transistors, 88 Rops, 484gb/s bandwidth.
RTX 2080 - 2944 cuda cores 13.6 billion transistors, 64 Rops 448gb/s bandwidth. Roughly 5% faster with 21% less cuda cores 13% more transistors 37% less Rops 8% less bandwidth.
RTX 2080 Ti - 4352 cuda cores 18.6 Billion transistors, 88 Rops 616gb/s bandwidth.
RTX 3080 - 4352 cuda cores *34.5 billion transistors* #number of rops? *760gb/s bandwidth* same amount of cuda cores, 85% more transistors, 23% more bandwidth.
What to take away from this? Previous X80 cards exceeded the performance of the previous gen Ti cards with much less resources whereas this time the 3080 exceeds the 2080 Ti everywhere other than drawing equal on the cuda cores and we don't know how many Rops it has... therefor I'm confident and I will be shocked if this card doesn't have a 35% performance uplift over the 2080 Ti...
I was bored and wanted to kill some time