Forced upscaling and blurry TAA is compensating for the fact that they can’t push graphics much further on the hardware we have. The current hardware progression has stagnated, combined with the fact that we are seeing more diminishing returns in graphics as they improve, requiring more power to deliver less of a noticeable difference.
But it doesn’t mean these games won’t look great when you disable the fakeness and run it with brute force GPU power 10 years from now.
I honestly think the current graphics we can achive are fine and where the true improvements should come from are better animation and actually good art direction.
I think the primary reason for the GPU stagnation has been the AI / GPU compute bubble over the past 5 years.
So much on-die space has been diverted away from raw rasterisation power towards CUDA, that it has artificially held back GPU progress.
When we do see the current AI bubble burst (and it does feel like we’re fast approaching that point, due to all the recent incestuous business dealings), hopefully we can see some innovation return to the sector.
I’m no expert on the matter, but I know this yt channel argues that the technology is already available. The thing is, big players like unreal engine devs make sub-optimal decisions when implementing these new features, leaving a lot of games being blurry and/or mal-ajusted simply by not knowing any better. Of course, art direction will always be important for a games graphics, but when the vast majority of tools available make things look bad by default, it makes sense that people will assume a better result is just not available yet.
That’s the guy who’s asking for a million dollars to “fix” unreal engine 5 despite having 0 programming experience and sends out dcma strikes for any videos that call him out on it, lol
Idk his render pipeline breakdown videos seem fairly in-depth. Is it just mumbo-jumbo? I saw some discussion where some devs seemed to acknowledge the perspective but say basically past 10 years of graphics make non-deferred render pipelines utterly unfeasible and thus MSAA, not to mention the issues that TAA “solves” like particularly fine geometry (see guitar strings in TLOUpt.2) or shimmering on stuff that can’t be optimized e.g. hair.
Frankly though I think in practice the difference between graphics in 2015 and 2025 is negligible compared to the difference between TAA (or DLAA/FSR/XeSS/FXAA/SMAA) and x4 MSAA. The only that comes even close is Path Tracing in CP2077.
I agree he seems like a sketchy af grifter, but I’ve not seen a single good rebuttal of his actual points, and even if he was a grifter, that doesn’t invalidate what he’s saying.
That Half-Life Alyx render in flatscreen with MSAA looks better than practically any game I’ve seen.
Forced upscaling and blurry TAA is compensating for the fact that they can’t push graphics much further on the hardware we have. The current hardware progression has stagnated, combined with the fact that we are seeing more diminishing returns in graphics as they improve, requiring more power to deliver less of a noticeable difference.
But it doesn’t mean these games won’t look great when you disable the fakeness and run it with brute force GPU power 10 years from now.
I honestly think the current graphics we can achive are fine and where the true improvements should come from are better animation and actually good art direction.
I think the primary reason for the GPU stagnation has been the AI / GPU compute bubble over the past 5 years.
So much on-die space has been diverted away from raw rasterisation power towards CUDA, that it has artificially held back GPU progress.
When we do see the current AI bubble burst (and it does feel like we’re fast approaching that point, due to all the recent incestuous business dealings), hopefully we can see some innovation return to the sector.
Pc gaming can achieve 1080p@120 or more without much effort…
I’m no expert on the matter, but I know this yt channel argues that the technology is already available. The thing is, big players like unreal engine devs make sub-optimal decisions when implementing these new features, leaving a lot of games being blurry and/or mal-ajusted simply by not knowing any better. Of course, art direction will always be important for a games graphics, but when the vast majority of tools available make things look bad by default, it makes sense that people will assume a better result is just not available yet.
That’s the guy who’s asking for a million dollars to “fix” unreal engine 5 despite having 0 programming experience and sends out dcma strikes for any videos that call him out on it, lol
Idk his render pipeline breakdown videos seem fairly in-depth. Is it just mumbo-jumbo? I saw some discussion where some devs seemed to acknowledge the perspective but say basically past 10 years of graphics make non-deferred render pipelines utterly unfeasible and thus MSAA, not to mention the issues that TAA “solves” like particularly fine geometry (see guitar strings in TLOUpt.2) or shimmering on stuff that can’t be optimized e.g. hair.
Frankly though I think in practice the difference between graphics in 2015 and 2025 is negligible compared to the difference between TAA (or DLAA/FSR/XeSS/FXAA/SMAA) and x4 MSAA. The only that comes even close is Path Tracing in CP2077.
I agree he seems like a sketchy af grifter, but I’ve not seen a single good rebuttal of his actual points, and even if he was a grifter, that doesn’t invalidate what he’s saying.
That Half-Life Alyx render in flatscreen with MSAA looks better than practically any game I’ve seen.