To be fair, we're reaching the limits of what can be done with silicon chips and how much the nodes can be shrunk. Nvidia will have no node advantage this gen while AMD will go, I think, from 5/6 to 4nm, which isn't a big improvement there.

There's also the physical limitatin of how big can the chips be. AMD tried the chiplet approach to solve that and failed, in its first attempt, while Nvidia has yet to try.

We can expect architectural improvements and ne designs, of course, but that alone can't make up for the other two problems.

And let's not forget that we're seeing new engines that don't give a f*ck about optimization and shoehorn in the different upscaling techs to make the games run by brute force.

So, all in all, this is the new reality and you'll be a fool if you think that a miracle will happen and things will be as they used to.

While I do agree that there is an impending wall of doom coming. I think Nvidia still could have configured it a lot better while still making a ton of money. There's nearly 10,000 cuda core difference between 5080 and 5090. If they configured 5080 as GB202 and gave it 12,000 or 14,000 cuda cores, it probably could have matched a 4090. And if they priced it $1000 while pricing the rest accordingly and upping 5070 vram to 16GB, I think this would have been a pretty good generation. But instead they choose the super greedy route where the only GPU in the line up that's remotely worth it is the 5070 Ti. The rest of the GPUs feel bad to get including the 5090 where as at least the 4090 felt godly if you had the big bucks. The 5090 feels meh at best.