If these new graphic cards are so powerful why is there so many instances where they use smoke and mirrors to achieve great graphics? I am talking about things such as frame generation, Dssl,FSSR etc. If the power is there what's the need for all the software based stuff? I understand drivers and such but it just appears that there are alot of gimmicks being thrown around when it comes to graphics cards. It's like they are feeding more instructions to the gpu for these gimmicks and there has to a performance hit.
-
Hey Guest. Check out your NeoGAF Wrapped 2025 results here!
GPU Question
- Thread starter FMX
- Start date
SirKicksalot
Member
Every copy of Doom 3 shipped with a tiny John Carmack waving a flashlight around your PC to make shadows.
If you ever used MSAA, your frames were not freshly harvested organic frames either.
If you ever used MSAA, your frames were not freshly harvested organic frames either.
Last edited:
SlimySnake
Flashless at the Golden Globes
They are literally using ray tracing and path tracing in games nowadays. The complete opposite of smoke and mirrors. And ray tracing is expensive. Path tracing is literally how they make cg movies.
Besides, I play a lot of games at native res. If you are getting a 5090 you will be playing almost every game at native 4k 60 fps. You will use dlss only as AA, not for upscaling.
Besides, I play a lot of games at native res. If you are getting a 5090 you will be playing almost every game at native 4k 60 fps. You will use dlss only as AA, not for upscaling.
Windom Earle
Member
Totally not bait thread
LordOfChaos
Member
If you have a high end card you're boosting it to very high resolutions and framerates, if you have a midrange card you're doing a fair trick to get higher apparent performance?
There's always room for more. Plus, ray tracing which threw a giant wrench in many raster games being pretty easy for any relatively modern high end card but can use these tricks to do more again.
phant0m
Member
Ready for the one that's really gonna bake your noodle?
It's ALL fake. Doesn't matter if your alleyway in CP2077 is lit by baked lighting and rendered at 860p or path traced at native 4k, it's all pixels in the end of something that doesn't actually exist.
IMO, it's pretty weird that gamers have taken to this ideology that rasterized graphics are 'real' but upscaling/super sampling/framegen are 'fake'
Literally every graphics card on the planet exists to generate frames.
Last edited:
Durin
Member
GPUs have always had "gimmicks", and many ways graphics improve is through new techniques than just cranking up the same features with more horsepower. Or techniques that feel like cheats:
Normal maps are a cheat to make 360-era games have surface level detail that doesn't exist. This example has a wall where the artist made a high-res 3d sculpt of all the surface details, baked that surface depth/detail into texture map, and the lighting just fakes those details. No individual polygons to model out the cracks because the hardware couldn't do it. If you tilt the wall even more and look at it from a side angle, you'd see right through the limitations that both are flat-ass planes, but light and texture fake it well enough.
The harsh reality is GPUs in general are hitting a wall on how much rasterization improvements they can make from gen to gen now. So they need to make use of other techniques around AI using dedicated cores that can do stuff like upscaling from lower resolutions, or generating frames in-between rasterized frames to improve motion clarity.
If pushing rasterized graphics were still viable without the "gimmicks", AMD or Intel would've just done it better than Nvidia, but they can't because none of them know how.
Normal maps are a cheat to make 360-era games have surface level detail that doesn't exist. This example has a wall where the artist made a high-res 3d sculpt of all the surface details, baked that surface depth/detail into texture map, and the lighting just fakes those details. No individual polygons to model out the cracks because the hardware couldn't do it. If you tilt the wall even more and look at it from a side angle, you'd see right through the limitations that both are flat-ass planes, but light and texture fake it well enough.
The harsh reality is GPUs in general are hitting a wall on how much rasterization improvements they can make from gen to gen now. So they need to make use of other techniques around AI using dedicated cores that can do stuff like upscaling from lower resolutions, or generating frames in-between rasterized frames to improve motion clarity.
If pushing rasterized graphics were still viable without the "gimmicks", AMD or Intel would've just done it better than Nvidia, but they can't because none of them know how.
Last edited:
Crayon
Member
ML upscaling works very well. Most people will at least flip it to quality mode because the artifacts are not always that easy to notice at all. It's an especially good trade when you don't have a powerful gpu. Most pc gamers are working with less power than a ps5.
It's not mandatory anyhow. Whether you have a powerful card or not, it's up to you to use upscaling. Brute 4k at 300 watts or fake it with indistinguishable results for 200 watts. Whatever you want.
It's not mandatory anyhow. Whether you have a powerful card or not, it's up to you to use upscaling. Brute 4k at 300 watts or fake it with indistinguishable results for 200 watts. Whatever you want.
Danny Dudekisser
I paid good money for this Dynex!
cinnamonandgravy
Member
nemiroff
Gold Member
Because games in general don't look like this anymore
Anyway, here's the thing: games are all about smoke and mirrors. They aren't real life, they're crafted and processed through computing.
The balanced end result is where DLSS shines. And then there's the challenge of dealing with the broken Moore's law..............
Anyway, here's the thing: games are all about smoke and mirrors. They aren't real life, they're crafted and processed through computing.
The balanced end result is where DLSS shines. And then there's the challenge of dealing with the broken Moore's law..............
Last edited:
analog_future
Resident Crybaby
Improvements made in software aren't "gimmicks". They can be just as impressive as improvements made in hardware.
Truth is that graphics are getting better but visual fidelity is getting WORSE.
Have you ever noticed that your graphic driver have many "optimizations" turned ON by default? All these techniques lower visual fidelity to imrpove performance. Upscaling and frame generation is just the next step in this diversion.
All of these while GPU vendors keep lowering hardware specs (VRAM size, BUS width, ROPS, memory compression), making us pay more for less
Windom Earle
Member
Before we had hardware acceleration there was only software rendering. Those damn gimmicks man.
Fbh
Member
Because the complexity of what can be rendered on a computer is on a very wide spectrum. Regardless of how expensive and powerful a GPU is, there are always going to be limitations on what it's capable to render dozens (or hundreds) of times in a single second.
Movie quality CGI can take literally hours to render a single frame on work stations with much more computing power than your average gaming PC. So why wouldn't there be a limit to what an RTX5090 can render 60 times in a second at native 4K?
Movie quality CGI can take literally hours to render a single frame on work stations with much more computing power than your average gaming PC. So why wouldn't there be a limit to what an RTX5090 can render 60 times in a second at native 4K?
poppabk
Cheeks Spread for Digital Only Future
They are illusions, a trick is something a whore does for money.
But more seriously - everything in a video game is an illusion and fake - AI upscaling and frame generation is just much more efficient than increasing resolution by 2 or framerate by 2, so why not use it?
But more seriously - everything in a video game is an illusion and fake - AI upscaling and frame generation is just much more efficient than increasing resolution by 2 or framerate by 2, so why not use it?
Fbh
Member
Should I do that before or after I install that 32GBRAM.exe file I download to add more Ram?
Windom Earle
Member
You don't understand. It's about the ethics of native resolutions. /s
simpatico
Gold Member
They're not. The consumer GPU benchmark for rendering resolution has been 1080 for nearly two decades now. They work long nights to come up with filter effects like DLSS and sprinkly flowery tech words around it, but the bottom line, these fools cannot make a native 4k GPU. They do not have the technology.
rodrigolfp
Haptic Gamepads 4 Life
Eh, not really. Every upscalling and FG are optional plus things like DLSS improves stuff like some anti aliasing vs native or even super-sampled resolutions while some little blurriness and/or artifacts are trade offs.
V2Tommy
Member
The cards are not as powerful as people would have you believe. There is no "overkill" GPU on the market at any price, not even close. My 5090 can't even do 4K60fps in recent games. It's merely about having "the best performance today" and nothing more, all this fake frame trickery is practically required to let people use the refresh rate their displays are capable of.
Worse quality than MSAA and much worse performance than software antialiasing such as SMAA. I don't buy this "DLSS looks better than native" BS
Bojji
Gold Member
MSAA doesn't work in modern engines and FXAA/SMAA/MLAA looks like crap. TAA is used to make aliasing go away but in many games lower resolutions + DLSS4 produce better results. You don't have to believe in it, if you have GPU with that functionality you can check it yourself.
Already have, DLSS always looks softer than native
Bojji
Gold Member
Last edited:
rodrigolfp
Haptic Gamepads 4 Life
MSAA doesn't cover everything, only polygons edges.
turbo_blade
Gold Member
the power isn't there to run a game at 4K 60+fps with full pathtracing.
DLSS/Frame gen might use AI but it's how Nvidia has been able to introduce and enable the adoption of ray/pathtracing which is incredibly demanding.
DLSS allows games to run at a lower resolution (freeing up performance) while maintaing native or better resolution. Frame gen takes 1 real frame and generations "fake" frames to boost frame rate.
it's going to be a long while before pathtracing is standard in games and can be done without "smoke and mirrors".
This is actually one of the best uses of AI right now. it's just insanely good upscaling and while some people might not like the latency from frame gen i personally can't even tell a difference.
also the frame gen benefit is tied to your hardware. if you want the best performance you still need the best card.
"Native + TAA"...
Alex from DF did an interview with Intel's GPU lead discussing that just came out.
How things are rendered is changing, raster has hit a brick wall in terms of whole big improvements can be generation to generation.
It's time to start thinking in terms of raster frames and temporal frames.
Bojji
Gold Member
Native without TAA is full of aliasing.
Puscifer
Member
All that stuff is for features like path tracing which is used for things in movies that have entire rendering farms. Being able to do that at a consumer level is amazing but costly, so it helps us see future graphics now.
Also, this was tried on consoles years ago so it's not anything new.
LucasArts' 60FPS Force Unleashed II tech demo
A frame-rate upscaler that really works? Digital Foundry investigates
www.digitalfoundry.net
SlimySnake
Flashless at the Golden Globes
yep. its unplayable.
Itchy Tickles
Member
Ive always been on the side that things like DLSS and frame gen are meant for the end user only, but when devs can't be bothered to polish their games they use these tools as a crutch and then make it mandatory.
Hawk The Slayer
Member
As long as it reach the point where you can't tell the difference between native and DLSS.
I remember the Moguri mod for FFIX being amazing making them game look so much better while still keeping the feel of the original snd that uses A.I heavily.
So if they keep improving it to the point where you can't tell any difference than it would be great.
I think us 5000 series user will get more improvements with future new dlss.
I remember the Moguri mod for FFIX being amazing making them game look so much better while still keeping the feel of the original snd that uses A.I heavily.
So if they keep improving it to the point where you can't tell any difference than it would be great.
I think us 5000 series user will get more improvements with future new dlss.