It's a shame that consoles are so far behind on a technology that they need so much. In a time when hardware costs are not going down it is extremely important to get things like this right.
Upscaling Technology Has Become A Crutch
- Thread starter BitByte
- Start date
Charlietus
Regular
This console gen was created at the perfect time to miss out on the major improvements that would have made them "next gen" (graphics wise).
Instead we have this ankward middle ground where developers want to use technologies that aren't really fit for what those consoles can do.
It's the ps3-360 all over again (but nowhere near as terrible).
Looking at PS4 performance, it could have been delayed another couple of years. Get PS4 looking really ropey and then present something far more compelling for a next gen. Although a calculation of lost hours due to people still using slow loading systems and no quick resume would probably show the cost to civilisation too high to bear! 
Charlietus
Regular
I still have a PS4, and I tried some recent games on it. On the 1080p 32" inch screen that I got downstairs, games still look completely fine. That memory bandwidth and gpu is still great.Looking at PS4 performance, it could have been delayed another couple of years. Get PS4 looking really ropey and then present something far more compelling for a next gen. Although a calculation of lost hours due to people still using slow loading systems and no quick resume would probably show the cost to civilisation too high to bear!
Yep. And discussion here spawned DF! Grandmaster came in with his lossless full HD capture tech and started about a technique to derive frame-scales from video by counting pixels.
I've been comparing DLSS and FSR in Enshrouded. It makes me think the Switch 2 might come closer to the Xbox Series S than the specs would suggest, simply because it has access to DLSS. If you control for image quality, DLSS kind of gives you a generational performance advantage over FSR.
It's four years later and AMD haven't released anything that is a significant improvement on RDNA2. I feel it'd have been silly to keep holding out.
They were about as good as we could have hoped in my opinion. I was certainly suitably impressed by the specs when they were announced. We should remember that nobody even expected them to have HW accelerated ray tracing capabilities at all originally.
Charlietus
Regular
I have a feeling that RDNA 2 came out good thanks to the investment and the help of Sony and Microsoft. And if that's true, then RDNA 3 isn't representative of what AMD could have really delivered.
raytracingfan
Newcomer
The best case scenario for RDNA3 if it had the benefit of Sony and MSFT's investment instead of RDNA2 is the RDNA 3.75 the PS5 Pro has. There would be a greater graphical uplift and less cross-gen games compared to what we got but that's it. All the games like CP2077 that released in the 2020-2022 but clearly needed hardware more powerful than the last-gen consoles would have been left in an awkward position. UE5 would be delayed, and thus the optimization and improvement the engine has gone through would also be delayed. And if it established the precedent that console generations would last 9 years that would be a major setback for technological advancement. Having the current-gen launch as it did and getting next-gen with RDNA5/UDNA1/UDNA2 in the 2026-2028 period is better for the industry than current-gen launching in 2022 with a superior RDNA3 and next-gen launching in 2031.
the point of DLSS and so on is that they help to create a high-end experience with low-range or mid-range hardware. I don't think they are using it as a crutch on most games. It's like upscaling on consoles, native 4K is not attainable on many games, so they use upscaling to compensate for that, they have no choice.
It also provides anti aliasing in a method that is superior to most temporal algorithms.
So I would say that it’s a necessity at this time. TAA with 4K will very often look worse than DLSS upscaled from a lower resolution.
I’ve seen enough videos and screenshot comparisons where I think many times people attributed the blurriness to being a difference in anisotropic filtering or texture quality between Xbox consoles and PS5, when now thinking back, it’s a likely a result of TAA being slightly better on PS4Pro/5 due to devs incorporating id buffer.
And you can see how large a jump is with NN based anti aliasing. The difference is night and day when it works.
Upscaling is a crutch. But NN based antialiasing is the future, which is often bound to upscaling. I guess another way to look at it is, if you're going to eat the performance cost for superior AA, you may as well claw some of that performance back by reducing the resolution and taking advantage of the upscaling.
Last edited:
DavidGraham
Veteran
Or, for a pure image quality result you can run NN Anti Aliasing on top of native res as in the case of DLAA, with minimal performance cost (often less than 10%).
Some people even used NN Super Sampling through DLDSR where they run the game at 1.5x the native res of 4K, downsample it to native res using DLSS, and then downscale it using DLSS Quality and achieve very good results.
I did notice that DLAA isn't a very common option, maybe I'm wrong here on that. I do usually see DLSS for sure, just about nearly all the PC titles have it, but DLAA as an option is not quite as supported yet, but I agree, with the right hardware it is probably the best image quality that one can get. I certainly can't run native with DLAA with a ultra wide monitor.
DavidGraham
Veteran
Yeah, DLAA is not common vs DLSS, but in the last year or so, it has seen a huge uptake in support, I think about a 100 titles now support DLAA natively. There is also a very powerful and stable mod than can enable DLAA in any DLSS supported game.
IMO, using a lower resolution to get better performance has been a thing since the dawn of 3D gaming. upscaling is just a better version of it. it's like driving slower to get better fuel mileage. but if you want the best mileage at higher speeds, don't buy a low efficiency vehicle.
Yep. Back in the crt days you just changed your resolution. Then lcd came out and had a native resolution, but it wasn’t so bad until 4K showed up. 4K is just kind of ridiculous as a native target, especially when it’s getting more and more difficult to produce the hardware to render on.
arandomguy
Veteran
I've been consistent on this but I feel because of the context of the release into consumer hands with DLSS this caused upscaling to be looked at from a hardware comparison stand point which I have always felt is the wrong approach. From the hardware standpoint the discussions ends up being based around the idea that is there is always faster hardware and that you look at it from the lens of how upscaled compares to native with all else being equal.
However in terms of the actual user experience angle it should have always been looked at from a fixed hardware standpoint.
Effectively you have the following actual scenarios -
1) "Native" - higher performance demand and therefore lower fidelity and/or FPS
2) "upscaled" - lower performance demand with higher fidelity and/or FPS
The comparison then should actually be between those two scenarios and not how native compares vs. upscaling at the same fidelity/settings and FPS. Due to the sheer performance efficiency advantage relative to quality of current upscaling techniques it's hard to argue that scenario 1 would be preferable to most users in a blind test.
Upscaling as discussed above also solves the single target native resolution issue. I know hardware enthusiasts/PCMR tends to be fixated on the idea of buying hardware for their most demanding games and target resolution but creates a huge in issue in that you essentially overbuy as not all the games someone plays (for most people) are that demanding. No longer having to worry about whether or not all your games can run at 4k or even 1440p is very useful from a real world stand point.
However in terms of the actual user experience angle it should have always been looked at from a fixed hardware standpoint.
Effectively you have the following actual scenarios -
1) "Native" - higher performance demand and therefore lower fidelity and/or FPS
2) "upscaled" - lower performance demand with higher fidelity and/or FPS
The comparison then should actually be between those two scenarios and not how native compares vs. upscaling at the same fidelity/settings and FPS. Due to the sheer performance efficiency advantage relative to quality of current upscaling techniques it's hard to argue that scenario 1 would be preferable to most users in a blind test.
Upscaling as discussed above also solves the single target native resolution issue. I know hardware enthusiasts/PCMR tends to be fixated on the idea of buying hardware for their most demanding games and target resolution but creates a huge in issue in that you essentially overbuy as not all the games someone plays (for most people) are that demanding. No longer having to worry about whether or not all your games can run at 4k or even 1440p is very useful from a real world stand point.