No, as I've told you numerous times now i was expecting something akin to 7900 Xt performance for the price of the 6800 XT, for a 10-20% performance improvement over previous generation cards for similar money
as has been the case for the last 10-20 years.
I was not expecting there to be stagnation, i was not expecting there to be almost no point in buying a new graphics card because it offers little in the way of price/performance improvements.
But then i suspect you know that already and you're just being argumentative for the sake of it.
Moore's law may not be totally dead, but it certainly has slowed down a lot. That plus only two contestants and one barely bothering with dGPUs means Nvidia largely have a license to print money. Especially with their crazy mindshare (3050 selling for more than 6600 XT, 3060 non-Ti selling for more than 6700, etc. that is crazy!)
But yes, there is a real danger that they will kill the golden gaming goose - even if they are raking it in with the golden turkey of AI - and like in the years of the Intel CPU stagnation of quad cores there is a real danger that the market collapses because there is nothing worth buying for most people's budgets. If people only buy new dGPUs when the old one breaks, then volumes will have to go way down.
Well, the best thing Nvidia could do at a small loss in margins is to start being "
generous" with VRAM. Planned obsolescences might take major hit though. And planned obselescenece has worked wonders for Nvidia despite all the denials and
is X VRAM enough? threads being locked: just witness the amount of people upgrading from 3060 Ti, 3070 and even 3080 because 8GB/10GB is now borderline.
They won’t need to cut the cards die sizes down any further next gen aside from maybe the bus getting cut as GDDR7 will make up for that. It will be the case of normalising the cuts that have already been made and moving those same size dies onto 3nm would give all the cards a 30% uplift and make it look a fairly decent upgrade when in reality you’d just be getting the performance you should have got with this gen.
Unless their next gen is coming in 2026, I doubt anything except for the top die gets 3nm as it seems to command a huge increase on 5nm/4nm per wafer. I was actually surprised that 4090 offered such a large improvement over 3090 but to do so they - again surprisingly - went to what was near the limit for the node.
10-20 years ago, ever 18 months to 2 years there would a new node and while wafer prices, design costs, masks etc. went up a bit the price per transistors went down massively. Now things are far slower. What Nvidia (or AMD) can no longer do is take a 80nm design, port it to 65nm like 8800 GTX to 8800 GT. Those days are gone.
So those are the excuses, but having said all that: Nvidia's margins and overall profits even
just from Gaming is crazily high for a fabless company. So they could afford to offer far better generational improvements. But since at least Titan, they have noticed that some whales are willing to overspend, which means Nvidia have little reason to offer better value.
Barriers to entry into the dGPU market are huge not only because of patentwalls, but also drivers, and the complexity of design, which suits Nvidia and AMD fine.