I’ve seen a lot of people lately saying that upscaling (fsr, dlss, etc.) is a bad thing, including some calling it ‘fake frames’, which is probably due to them confusing it with frame generation.
What upscaling does is take an input (a frame rendered at 1080p, for example) and attempt to improve it by generating more information (bringing that 1080p frame to 1440p). this does make things a little fuzzy, but it also frees up resources to allow stuff like improved lighting to be rendered which makes games like cyberpunk able to be rendered at a decent framerate without a $5,000 gpu.
Frame generation is different. It takes an input as well (same 1080p frame, for example), but it doesn’t improve the frame. It makes a new one based on that frame, sometimes several. These actually are ‘fake frames’, and this is what the people who called upscaling fake frames were really talking about.
I won’t lie, upscaling is definitely a crutch and the goal should be to be able to render that cool stuff at native resolution. however, the tech that can render that stuff is too expensive to be worth buying unless you have money to throw away, which real people typically don’t. it’s up to you whether a little fuzziness in the graphics is worth it to you, but the fact is it’ll give you the leeway to choose between higher framerate and prettier lighting. without it most people are stuck just setting their graphics to ‘no’, because they can’t afford the kind of processing power making things look good at native resolution takes.
Part of why I am making this post is that I wanted to see what other people think of this take, and more importantly get feedback so I can improve the take later. I’m currently running a laptop with a 1650, and I’ve had it for years. I’m used to balancing frames and quality and making compromises, and upscaling tends to be one of them that’s worth making.
Upskaling is a fabolous technology and the split that quality needs to do between hardware upgrades and software support. Overall the existence of the technology is definitly a positive one.
However people are worried about a development that we are already seeing where games are just not efficient with their resources and require way to much computing power. People are afraid studios will decrease the amount of work they put into optimising because they feel like Upskaling will solve all perfomance problems for them. But optimisation needs to happen on both parts. That’s what people are afraid of.
I’m a game developer and I will 100% confirm that studios have already started and will continue assuming the user has DLSS/FSR/XeSS enabled because it turns out rendering half as many pixels can get you across the finish line.
It was already fairly standard practice to try as hard as you can for performance, and when that fails to bring you good performance at native resolution, just cut some resolution (for example, to 900p from 1080p).
However, I do want to add that DLSS/FSR/XeSS is great technology for the low end of the market who can’t afford insane rigs but do get to have a slightly sharper image than previous upscalers could accomplish.
ofc optimization is still needed, I meant that upscaling improves the maximum you could reasonably get out of your hardware. I’d also like to see actual low options come back bc it’s really annoying not being able to turn the graphics down anymore.
I think the greatest growth for me is realising how good “high” or “medium” presets are nowadays. There is a lot of FOMO on missing the best (looking) experience, but IMO modern medium settings are stunning. I was watching a graphics comparison of Kingdom Come 2 and the improvements from “ultra” were so miniscule. The jump from “low” to “medium” to as incredible tho.
Upscaling an old game on fixed hardware that can’t output at high resolutions is good.
Upscaling a new game as part of the graphics pipeline instead of optimizing it is terrible and shouldn’t be accepted by gamers that have to spend $1000+ on a GPU
Upscaling is great for older games that don’t support modern resolutions. I’m not really a fan of using it in modern games because of the blurryness you mentioned.
Regarding frame gen… Optimise your damn games! Game optimisation is becoming a lost art and now people think they should spend thousands on new hardware too keep up :(
I don’t think I’ve ever seen someone say that upscaling OPTIONS are bad but I’m worried about games like monster hunter where upscaling and frame gen is used to make the game playable in most cases.
I agree in principle, but it’s a crutch that shouldn’t substitute good code. It’s like having a powerful car that runs sluggish, and then someone suggests that removing a couple seats could improve things.
Totally agree though would say it’s more along the lines of needing premium gas and newer performance air filters and tires when you’re thinking it should be capable out of the box
I’ve personally gotten a lot out of all the AI enhanced graphics technologies, and pretty much consider these applications the absolute perfect use case for the AI we have today. Yes, they shouldn’t be a substitute for optimisation, but overcorrecting the other way and attempting to claim that DLSS is garbage that ruins everything and looks like shit is also bad (and untrue).
Even frame generation has its uses, as long as you don’t play something fast paced where there is a lot of camera movement and/or you’ll feel the added input lag too much.
A special shout-out to the redheaded stepchild of the family too: DLDSR is a fantastic technology and once you’ve tried it you’ll never want to go back.
I’m fine with the concept of upscaling tech. DLSS 4 with the transformer model looks excellent. And FSR 4 is looking pretty damn decent as well. The earlier attempts weren’t as good. Ideally it would be acting more like DLAA, but 8.3 million pixels is a lot to render (4K). And if 8K is going to be a thing one day, it makes even more sense there.
I think too many people focus on the now and can’t imagine what things will be in the future as they progress.
Now frame generation, that one I feel less optimistic about. Especially when I see people using it for 60fps or less. It should really only ever be used at 80fps or higher, where the lag is less of a problem. But one day inferred frames, where it only looks at the prior frames and does not wait for the next frame, might make it a better experience.
Lastly, it’s NVidia and AMD’s marketing departments fault for having them all conflated. DLFG & FFG is what the frame gen tools should have been called, rather than shoehorning them under their super sampling and super resolution branding.
Don’t pay attention to idiots who don’t know what they’re talking about. Lemmy and Reddit are full of tech misinformation. It’s not even worth replying to.
Okay, you convinced me to give it a shot with some future games
If I really need upscaling, I just let my monitor do it with bilinear scaling or whatever. No fancy hardware required.
That upsampling, which doesn’t fill in the details.
Yeah, taht is more or less where I come down. “AI” upscaling is spectacular. Frame gen is much more hit or miss
The main problem is that, as with most things, people are stupid. They don’t understand that an outlet like Digital Foundry or even Gamers Nexus are going to be harsh on upscaling/frame gen because it actively makes it hard for them to give you guidance on what performance you can expect. So “This is horrible for benchmarking” becomes “This is horrible”
I’m confused. Digital foundry clearly spells out the performance you will get with both FG and without. They’re not harsh on it at all?
I think upscaling is a good idea. Most of the time I’m running around while dodging bullets, arrows or fireballs, so I don’t really have time to examine the details of the foliage around me at the pixel level. I also will not buy an overpowered space heater so that the grass in my game looks more realistic. I don’t want a triple fan monster sounding like a turbojet near me.
