Jensen Huang said a few years ago in an interview that Moore’s law is effectively dead, and that they are focusing now on optimizations rather than transistor count.
I suppose that means optimizing their pricing structure for maximum profit, and optimizing ways to improve performance metrics without significantly improving actual performance, like these frame generation routines.
Ironically this will lead to less optimization on the part of games being developed. Why spend time and money optimizing your game when you can rely on dlss and 4x framegen to get players good frames?
Is it just me, or is it depressing that modern gaming has succumb to requiring AI gimmicks to run performant? Why does every new AAA title require DLSS and why would I want AI generating 75% (as I perceive the DLSS 4x, correct me if wrong) of the frames that induces artifacts like halos and blur anyway? What is it about these games that some optimization wouldn’t be warranted for them to run without some AI layer on top to get 60 fps, let alone the 144+ fps that consumers now expect?
To be fair, the point of framegen is for high refresh rate monitors. You’re assumed to already be hitting good fps numbers at baseline. And dlss is made mainly for 4K native res. I’m sure nvidia doesn’t mind people thinking otherwise
The entire industry has hit a wall and NVIDIA is insisting on coming up with infinite problems and selling solutions. They are trying their damnest to effectively take control and manipulate the market in their favor with these BS frameworks that are hardware locked.
Games are very lazily optimized for anything other than NVIDIA these days and we can look forward to the trend continuing now that AMD and Intel are having a very tough time competing. A popular theory is that everyone only like competition cause it makes their grand green gpu a little cheaper.
TL:DR is that NVIDIA can do whatever they want while players and developers follow blindly because neither Intel nor AMD match their market share. Its kinda like linux vs windows in a way. We are witnissing NVIDIA attempting to takeover the gpu market similar to microsofts early strategies.
Games are very lazily optimized for anything other than NVIDIA these days
NVIDIA bad and all that, but I wish that was true. If the game runs like shit, it runs like shit everywhere. What are these mythical games, that are optimized for NVIDIA in 2025?
I guess they mean they’re optimized for Nvidia in the sense that they need upscaling technology in some form due to either demanding graphics or poor optimisation or both, and Nvidia has the best one at the moment. I wish it wasn’t the case, but DLSS does look better than the alternatives.
Depressing? There are a lot of other things right now that depress me way more than the type of tools used to project the image I’m consuming in my escapist entertainment medium of choice. Maybe I would be more outraged if I was a programmer and knew exactly what kinds of optimizations developers were (or weren’t) doing these days. As an end user I have no problem with DLSS to be honest, I just see it as another tool and frankly it strikes me as one of the better applications of AI we’ve developed. There have been multiple examples where I was happier with the visuals with DLSS on rather than turning it off but also having to turn off or down a lot of other graphical bells and whistles. Also, making games that require DLSS on current hardware is future-proofing in a way, is it not? A generation or two of hardware down the line and you can just turn off DLSS and enjoy it in its full glory.
As a hobbyist programmer/game developer, ya basically hit the “as a programmer” part on the nose as in it’s why I care. In my opinion the issue is because we have faster and faster chips, we don’t care as long as it runs which means it’s using as much of the device as possible which means higher energy consumption. Also, DLSS has to be trained so you can lump it into how AI is causing more coal to be burned and water to be wasted cooling data centers designed to train AI.
It’s because ML enhanced graphics are perceived as “good enough” by the majority of consumers.
I am no game dev, but I will speculate that optimization is difficult, expensive (time/cost) and quickly reached a point of diminishing returns.
That being said, I personally prefer SSAA anti-aliasing, but I play heavily CPU-bound games that rarely stress my graphics card. But the SSAA ship has sailed, very few games offer this mode.
Nvidia DSR is literally supersampling.
I am aware of Nvidia DSR. I have not had much luck with it with the various indie games that I play.
Why? It works on every single game. What’s the issue you’re facing?
All fun and games until you have to use a 30 fps source to convert to 120fps because the card can’t handle anything else and it looks like absolute ass
So if I understand it correctly, DLSS uses some sort of ML to interpolate the extra frames. Isn’t that like the “smooth motion” options on TVs that let you get 240Hz from a 24fps movie?
I think it’s 60 or 120hz on TVs. Idk the technicals but I think it’s the same concept. Dlss framegen just does it based on ML and is probably more complex.
