I expect only 5090 to be a worthy step up from my 4090, so yeah, kinda hoping to skip it, and wait for RTX 6000 series on 2-3nm TSMC instead.
So far RTX 5000 series looks like a stop gap solution really. Same process node pretty much, they need to make chips physically bigger to deliver something good, which means higher power usage as well.
GDDR7 might bring some performance, however will also drive prices up more. It might matter in 4K/UHD+ gaming but I don't expect it to do much for 1440p and below. Nvidia looks to be skimping as well, going with 28Gbps instead of 32Gbps GDDR7 modules. Thats not really a huge upgrade over OC'ed GDDR6X.
5090 will be a beast, however, I also expect 1999+ MSRP price tag, with little availability on launch, hence hiking up price to 2500-3000 dollars probably. I think output will be minimal.
I also think 5090 will be very power hungry as well. Like 3090 Ti, maybe even worse, considering some 800 series motherboards are adding a new 8-pin GPU connector for RTX 5000 series, because 600W from 12VHPWR is not enough I guess?
5080 looks to be a massively step down according to rumours and think it will barely beat 4090 yet still sell for 1199 dollars like 4080 on release. I paid 1500 dollars for my 4090 like 2 years ago, so I might just keep it and get a 6080/6090 in a few more years on a true next gen node, 3nm TSMC or less.
GDDR7 might be the driving factor for 5000 series because the node is pretty much identical to 4000 series. TSMC 4N aka optimized 5nm again.
I don't think "DLSS 4" will matter, if we even see it, it will probably be RT/PT stuff which I care little about.
It is pretty clear that Nvidia have most of its focus on AI and Enterprise at this point.
I think we will see 5090/5080 in Q1 next year with 5070/5060 series Q2/Q3 because they are competiting with themselves for the most part and don't really need to release something new.