Can’t wait to run all my games in 240p and have a blurrier lanczos “AI” blow it up unevenly to 1366x768 so it’s all extra blurry and dogshit on my 8k TV
.
/s
On a more serious note, why do people use this again?
The problem isn’t the tech itself. Getting a pretty darn clean 4k output from 1080p or 1440p, at a small static frametime cost is amazing.
The problem is that the tech has been abused as permission to slack on optimization, or used in contexts where there just isn’t enough data for a clean picture, like in upscaling to 1080p or less. Used properly, on a well optimized title, this stuff is an incredible proposition for the end user, and I’m excited to see it keep improving.
Upscaling and Framegen is a crutch unfortunately that enables studios to focus on “photorealism” that pushes sales instead of proper optimization and gameplay that many games care about. And I blame suits of course that make decisions not my fellow engineers who follow the company line.
I don’t like that this is the state of game graphics either but I would much rather it be an open technology than a proprietary, vendor-lock, black box tech like nvidia tends to do.
Well, if it was used to prolong the life of your gpu it would be nice, but unfortunately it seems to be mainly used by devs so that they don’t have to optimize their games, so the end result is people with new gpus get blurry crap, and the older gpus get left behind all the same.
Thanks but no thanks, I already play MSFS in VR on my Quest 2 via VD and the JPEG compression on that thing is enough, adding any kind of upscaling would turn my comfy Fenix A320 into the goddamn vomit comet
With the quest? Yeah so am i. You very much still have compression. Even at 800mbps bitrate, thats still not the same as a proper display port headset like the Index.
In fact you very likely don’t see the smear probably because of the blocky JPEG of the image as is, and maybe you’re used to it because you watch a lot of streaming videos rather than high bitrate vids. Not meant as an insult, that’s just the way things are nowadays.
Yeah it’s been great for a few years but it’s slowly falling apart. I’ve been putting off getting a replacement as it looks like the only options are downgrading or triple the cost (or more)
If my deck can’t handle a game('s lack of optimization) I move to desktop. Or if it’s a visual artsy game I play on desktop. The most egregious fsr glorpination in recent memory was water in horizon forbidden west. When I first saw the beaches I thought it looked like utter shit but when I switched to actual rendering it was suddenly one of the best video game beaches ever.
But yeah basically everything that moves looks like shit, and anything bigger than a steam deck display makes it easier to tell something is fucked. I’ll check out fsr 4 but I don’t have high expectations.
Samesies, the deck just isn’t all that powerful but it’s okay, it’s not meant to be a top of the line thing, it’s a portable after all. That said sometimes I’d rather play on the deck just for that sweet sweet OLED.
You… you understand that the entire point of releasing an SDK, a ‘software developement kit’, is to simplify and streamline the methods that game devs use to implement FSR 4, expose and document more lower level, fine-tunable ways to access the render pipeline, right?
The entire point of this is so that game devs can better optimize their games so that what you describe is less of a problem.
People use this because when it is implemented well, it can actually be used in ways that make it so the whole game does not have to be built around depending on only having a high end GPU.
This release could potentially accelerate the success folks have already been having at getting FSR4 to run on older, cheaper 7000 series AMD cards.
FrameGen and Upscaling can be very good things when games are made by people who know how to use them correctly, instead of relying on them as a crutch to cover the lack of ootimizations from their ‘high speed low drag’ development practices.
But it actually has very little to do with optimization.
An SDK is essentially just a package of libraries (and often complimentary tools) with docs on the proper usage of APIs, header files you can link to when compiling and depending on whether it’s closed or open source you may also get the actual implementation code or in the case of closed source - you get .DLLs or .so files to Dynamically link against and the .DLL provides APIs via header files for you to talk to.
While I’m very happy that FSR is open source and now devs can fork it and create modifications to the actual implementation to suit their needs (presuming the license allows, haven’t checked), I do wonder whether this kind of tech is the electron of desktop apps, if you catch my out of memory exception.
That pushing them, making them more available, will only lead to neglecting real optimization techniques, and an over-reliance on upscalers. I don’t see how making upscalers more available will as you put it:
less of a problem.
In fact, you confirmed my fears right here:
People use this because when it is implemented well, it can actually be used in ways that make it so the whole game does not have to be built around depending on only having a high end GPU.
This is the entire problem. It looks like ass. You don’t need much beyond an RTX 3090 to run mostly every game on ultra. GPUs are very cheap nowadays compared to the crypto boom era circa 2020.
Well implemented, poorly implemented - it does not matter in the slighest.
FXAA, FSR, DLSS, and godforbid - TAA, are all a smeary mess that ruins visual clarity and makes games broadly unappealing no matter how good the underlying assets are.
This release could potentially accelerate the success folks have already been having at getting FSR4 to run on older, cheaper 7000 series AMD cards.
The only things those cards should be running is the same DVD of The Office on repeat in a retirement home in Sheboygan, Wisconsin.
Or idk, running some l33t riced out Linux distro with 10 different levels of transparency.
What they should not be doing is be used as an excuse for stagnating real time graphics, which if it becomes an industry standard leading to an overall decline of the art form as we have already seen with how much better games that implemented normal anti-aliasing methods like MSAA (or even SMAA as a fallback performant option) look.
FrameGen and Upscaling can be very good things when games are made by people who know how to use them correctly, instead of relying on them as a crutch to cover the lack of ootimizations from their ‘high speed low drag’ development practices.
They really can’t be implemented in any way that will make them ever preferable to native. And don’t be so naive, you know exactly what all this shit is for - it’s to make games sell more.
Shiny game = more sales = $$$
Back in the day this was done by just having the game run like shit. All the screenshots look good - and when you’re not playing, you don’t feel the impact of the framerate so much.
This goes way back to the N64, try playing the XBLA remastered GoldenEye on Xenia with mouse support - it is so much better - you’ll never want to go back.
Makes you wonder why didn’t they just make the game look worse, and actually run at an enjoyable framerate?
Well, they did this because shiny game sells better, whether you’re looking at trailers or screenshots in magazines, PS1 games looked like shit (including because the console didn’t have a hardware z-buffer and so jittered all over the place and relied on devs to implement sorting tables instead for ordering depth), so having top-end graphics was an easy selling point for N64 games.
Eventually they started making trailers that were just unrelated CGI, sometimes outsourced to a random CGI sweatshop in India, but people caught on and the industry switched to “in-engine”, which was marginally more honest.
FSR, DLSS, etc is the latest version of this same trick. You can’t really see the smear in gameplay trailers thanks to the YouTube’s 2009 jpeg-ass compression algo that people think is 1080p when it looks worse than a late 90s DVD, so graphics can be turned up to 11, then when you get the game it turns out your options are either to run it at 480p upscaled to 1080 and have it look like some crusty ass letsplay, or at a native res, but at 15FPS, which will be mostly generated of course.
Hyperbole - yes. But also truth.
Video games are art. They’re also a crazy technical achievement a lot of the time, I played TLOU Part 2 on PC recently, with DLAA at actually native 1440p and just… Wow! It was like that one tree from Uncharted 3 all over again, shit looked like real life. Once you turn the stupid sharpening off at least. Now play it on PS4 and it looks like whenever they’d show a woman close up on Star Trek: The Original Series.
It also made me sad because there wasn’t an option not to have this smeary DLAA shit, even if MSAA was available, the shimmering of e.g. hair or guitar strings or other thin objects that was never accounted for by devs because it would be covered up by the smear.
its just a shame, I love that game to death (I like weird games) and so I’d want to come back to it in 20 years and run it with 8x MSAA at 4K and be blown away all over again, just like I’d do today with Return to Castle Wolfenstein or something, but instead, it’ll be forever aged and tarnished by this crap, because it made it look slightly shinier on release and made it sell better. Ka-ching.
Sidenote: I’m not a professional dev or game dev, I only have a passing knowledge of graphics because I’m too smoothbrain for math. I may be wrong. Feel free to correct me.
MSAA has issues with a lot of transparency and high frequency textures. Both of these are important in pushing the upper limit of quality.
TAA based methods are more smeary but also have a higher ceiling for the same content.
Lot of teams understood where MSAA is a bad fit and designed content unlike that. Sadly TAA techniques are worked around that easily for a myriad of reasons (including lack of incentive to QA)
Capitalism’s misaligned incentive model pushing short-term revenue gains made by increasing sales of games through the overuse of upscalers to push graphical fidelity past actual graphical capability of the hardware that isn’t noticeable in promotional material due to video compression leading to dishonest advertising and bad looking games or worse yet - good looking games made to look bad with this AI shortcut instead of R&D bucks spent on actual solutions like adapting the MSAA approach to deferred rendering pipelines of modern game engines and finding solutions for aliasing on specular surfaces and overly small geometry?
Can’t wait to run all my games in 240p and have a blurrier lanczos “AI” blow it up unevenly to 1366x768 so it’s all extra blurry and dogshit on my 8k TV
.
/s
On a more serious note, why do people use this again?
The problem isn’t the tech itself. Getting a pretty darn clean 4k output from 1080p or 1440p, at a small static frametime cost is amazing.
The problem is that the tech has been abused as permission to slack on optimization, or used in contexts where there just isn’t enough data for a clean picture, like in upscaling to 1080p or less. Used properly, on a well optimized title, this stuff is an incredible proposition for the end user, and I’m excited to see it keep improving.
Upscaling and Framegen is a crutch unfortunately that enables studios to focus on “photorealism” that pushes sales instead of proper optimization and gameplay that many games care about. And I blame suits of course that make decisions not my fellow engineers who follow the company line.
I don’t like that this is the state of game graphics either but I would much rather it be an open technology than a proprietary, vendor-lock, black box tech like nvidia tends to do.
Agreed.
Well, if it was used to prolong the life of your gpu it would be nice, but unfortunately it seems to be mainly used by devs so that they don’t have to optimize their games, so the end result is people with new gpus get blurry crap, and the older gpus get left behind all the same.
Yeah… It sucks.
Try flight simulators in VR. Upscaling saves you at least €400 in GPU cost
Note that Frame Gen does not help here, as not getting nauseous in VR requires low latency. Frame Gen actually increases latency.
Thanks but no thanks, I already play MSFS in VR on my Quest 2 via VD and the JPEG compression on that thing is enough, adding any kind of upscaling would turn my comfy Fenix A320 into the goddamn vomit comet
Understood, I’m wired so no compression. It works well for me
With the quest? Yeah so am i. You very much still have compression. Even at 800mbps bitrate, thats still not the same as a proper display port headset like the Index.
In fact you very likely don’t see the smear probably because of the blocky JPEG of the image as is, and maybe you’re used to it because you watch a lot of streaming videos rather than high bitrate vids. Not meant as an insult, that’s just the way things are nowadays.
I use a Reverb G2 which is directly driven by the host pc through OpenXR, no compression or onboard processing at all.
I only use it for flight sims.
Oh sure yeah that’s fine then, you specified wired as opposed to the headset so I was confused, now I’m jelly haha
Yeah it’s been great for a few years but it’s slowly falling apart. I’ve been putting off getting a replacement as it looks like the only options are downgrading or triple the cost (or more)
If my deck can’t handle a game('s lack of optimization) I move to desktop. Or if it’s a visual artsy game I play on desktop. The most egregious fsr glorpination in recent memory was water in horizon forbidden west. When I first saw the beaches I thought it looked like utter shit but when I switched to actual rendering it was suddenly one of the best video game beaches ever.
But yeah basically everything that moves looks like shit, and anything bigger than a steam deck display makes it easier to tell something is fucked. I’ll check out fsr 4 but I don’t have high expectations.
Samesies, the deck just isn’t all that powerful but it’s okay, it’s not meant to be a top of the line thing, it’s a portable after all. That said sometimes I’d rather play on the deck just for that sweet sweet OLED.
You… you understand that the entire point of releasing an SDK, a ‘software developement kit’, is to simplify and streamline the methods that game devs use to implement FSR 4, expose and document more lower level, fine-tunable ways to access the render pipeline, right?
The entire point of this is so that game devs can better optimize their games so that what you describe is less of a problem.
People use this because when it is implemented well, it can actually be used in ways that make it so the whole game does not have to be built around depending on only having a high end GPU.
This release could potentially accelerate the success folks have already been having at getting FSR4 to run on older, cheaper 7000 series AMD cards.
FrameGen and Upscaling can be very good things when games are made by people who know how to use them correctly, instead of relying on them as a crutch to cover the lack of ootimizations from their ‘high speed low drag’ development practices.
Yes I know what an SDK is.
But it actually has very little to do with optimization.
An SDK is essentially just a package of libraries (and often complimentary tools) with docs on the proper usage of APIs, header files you can link to when compiling and depending on whether it’s closed or open source you may also get the actual implementation code or in the case of closed source - you get .DLLs or .so files to Dynamically link against and the .DLL provides APIs via header files for you to talk to.
While I’m very happy that FSR is open source and now devs can fork it and create modifications to the actual implementation to suit their needs (presuming the license allows, haven’t checked), I do wonder whether this kind of tech is the electron of desktop apps, if you catch my out of memory exception.
That pushing them, making them more available, will only lead to neglecting real optimization techniques, and an over-reliance on upscalers. I don’t see how making upscalers more available will as you put it:
In fact, you confirmed my fears right here:
This is the entire problem. It looks like ass. You don’t need much beyond an RTX 3090 to run mostly every game on ultra. GPUs are very cheap nowadays compared to the crypto boom era circa 2020.
Well implemented, poorly implemented - it does not matter in the slighest.
FXAA, FSR, DLSS, and godforbid - TAA, are all a smeary mess that ruins visual clarity and makes games broadly unappealing no matter how good the underlying assets are.
The only things those cards should be running is the same DVD of The Office on repeat in a retirement home in Sheboygan, Wisconsin.
Or idk, running some l33t riced out Linux distro with 10 different levels of transparency.
What they should not be doing is be used as an excuse for stagnating real time graphics, which if it becomes an industry standard leading to an overall decline of the art form as we have already seen with how much better games that implemented normal anti-aliasing methods like MSAA (or even SMAA as a fallback performant option) look.
They really can’t be implemented in any way that will make them ever preferable to native. And don’t be so naive, you know exactly what all this shit is for - it’s to make games sell more.
Shiny game = more sales = $$$
Back in the day this was done by just having the game run like shit. All the screenshots look good - and when you’re not playing, you don’t feel the impact of the framerate so much.
This goes way back to the N64, try playing the XBLA remastered GoldenEye on Xenia with mouse support - it is so much better - you’ll never want to go back.
Makes you wonder why didn’t they just make the game look worse, and actually run at an enjoyable framerate?
Well, they did this because shiny game sells better, whether you’re looking at trailers or screenshots in magazines, PS1 games looked like shit (including because the console didn’t have a hardware z-buffer and so jittered all over the place and relied on devs to implement sorting tables instead for ordering depth), so having top-end graphics was an easy selling point for N64 games.
Eventually they started making trailers that were just unrelated CGI, sometimes outsourced to a random CGI sweatshop in India, but people caught on and the industry switched to “in-engine”, which was marginally more honest.
FSR, DLSS, etc is the latest version of this same trick. You can’t really see the smear in gameplay trailers thanks to the YouTube’s 2009 jpeg-ass compression algo that people think is 1080p when it looks worse than a late 90s DVD, so graphics can be turned up to 11, then when you get the game it turns out your options are either to run it at 480p upscaled to 1080 and have it look like some crusty ass letsplay, or at a native res, but at 15FPS, which will be mostly generated of course.
Hyperbole - yes. But also truth.
Video games are art. They’re also a crazy technical achievement a lot of the time, I played TLOU Part 2 on PC recently, with DLAA at actually native 1440p and just… Wow! It was like that one tree from Uncharted 3 all over again, shit looked like real life. Once you turn the stupid sharpening off at least. Now play it on PS4 and it looks like whenever they’d show a woman close up on Star Trek: The Original Series.
It also made me sad because there wasn’t an option not to have this smeary DLAA shit, even if MSAA was available, the shimmering of e.g. hair or guitar strings or other thin objects that was never accounted for by devs because it would be covered up by the smear.
its just a shame, I love that game to death (I like weird games) and so I’d want to come back to it in 20 years and run it with 8x MSAA at 4K and be blown away all over again, just like I’d do today with Return to Castle Wolfenstein or something, but instead, it’ll be forever aged and tarnished by this crap, because it made it look slightly shinier on release and made it sell better. Ka-ching.
Sidenote: I’m not a professional dev or game dev, I only have a passing knowledge of graphics because I’m too smoothbrain for math. I may be wrong. Feel free to correct me.
MSAA has issues with a lot of transparency and high frequency textures. Both of these are important in pushing the upper limit of quality.
TAA based methods are more smeary but also have a higher ceiling for the same content.
Lot of teams understood where MSAA is a bad fit and designed content unlike that. Sadly TAA techniques are worked around that easily for a myriad of reasons (including lack of incentive to QA)
What’s your problem?
Capitalism’s misaligned incentive model pushing short-term revenue gains made by increasing sales of games through the overuse of upscalers to push graphical fidelity past actual graphical capability of the hardware that isn’t noticeable in promotional material due to video compression leading to dishonest advertising and bad looking games or worse yet - good looking games made to look bad with this AI shortcut instead of R&D bucks spent on actual solutions like adapting the MSAA approach to deferred rendering pipelines of modern game engines and finding solutions for aliasing on specular surfaces and overly small geometry?