Nvidia has used the gains of TSMC 5nm to heavily cut down everything under the 4090 and while the that card sees impressive gains both 4080s have less of an uplift over a 3080 compared to what the 3080 had against the 2080 while costing 50% and 100% more.Donno it's that's true, but it's also clocked very high, is built with the 4nm EUV TSMC process, with apparently more cache.
The only thing that concerns me is the continued use of GDDR6X, which no doubt runs very hot. Never had a problem with my RTX 3070 FE.
EDIT - It's possible NVIDIA has made some improvements to either their cooling or VRAM, as the GDDR6X VRAM apparently peaks around 84 Celsius:
https://www.kitguru.net/components/...s/nvidia-rtx-4090-founders-edition-review/25/
Thermal imaging results looking good too:
https://www.guru3d.com/articles_pages/msi_geforce_rtx_4090_suprim_liquid_x_review,8.html
Suppose we'll see.
I expect a 4080 12gb to average 10% faster than a 3080 10gb while the 4080 16gb will land around 35% faster in raster.
The 4070 is a cut down 4080 12gb silicon so I imagine it'll be around 3080 speed while also costing 3080 prices and again is a farcry from the 3070 which matched a 2080ti while costing £469. So yeah while TSMC 5nm is a great process the customers under a 4090 are not really getting any benefit from it in terms of performance but rather nvidia is just dishing out smaller cheaper to produce dies while doubling card prices so they are the only ones benefitting.