Modern game devs: “Umm if you want mirrors, you need to have a 5090 Pro Gt Rtx card to enable ray tracing and get a blurry mess, but it is LITERALLY impossible to have anything better right now”
Game devs 30 years ago:
Modern game devs: “Umm if you want mirrors...
Such an original thread. Did your twitter grandma write it for you?
Why can we get reflections off of windows and puddles but not mirrors? Can someone actually explain that?
who took this photo?
Lakitu, it's a mirrored reflection.
What is stopping developers from just putting an inaccessible copy of a room with a clone of your character in it? Does it increase the processing power that much?
Because windows and puddles usually just use your screen and flip it onto the reflection to get that effect or render a small portion of objects nearby at a low res but that doesn't work on mirrors because it's gotta be hi res and not just cast the screen.
The best option is to just duplicate the room in rooms with mirrors like how older games did it.
Unironically probably laziness of having to code a second player character just to be reflected
we can go way back
Devs are retards these days. As John Carmack, you should create your own engine so your game will be exactly the way you want it to be. With so much unity and now ue5 slop about it's little wonder game devs are doing barely any dev at all.
You don't need RTX for this
Does it increase the processing power that much?
Yeah, they put a ridiculous amount of triangles and shader instructions into the main characters and barely manage to hit 60 fps. But it's not just the player, depending on where the mirror is located and the viewpoint you'd have to potentially rendering intensive objects with transparency included twice. So you'd have to keep them mostly confined to simple scenes. You could use workaround like making the copy use lower level of details and less intensive shader but it's all about the spectacle nowdays so it's unthinkable to a modern dev.
Why can we get reflections off of windows and puddles but not mirrors?
We've been able to do both since before World Wide Web was easily accessible by most people on this planet. In the 2020s, there's literally no excuse.
Also, more and more devs succumbing to using mere Screen Space Reflections for EVERYTHING is just embarrassing attempt to make the RTX seem "better" in comparison.
What is stopping developers from just putting an inaccessible copy of a room with a clone of your character in it?
Nothing. But that would've been a pointless waste of time already come year 2000, because Planar reflections, as in render-2-texture, would've allowed capturing another view angle using secondary camera, which then would've been rendered on top of a game object, such as a flat mirror panel.
This is the sort of tech N64 was already capable of, and later games such as Half-Life 2 use everywhere.
Yes, even Unreal Engine has a toggle for this.
notice how extremely simple the scenes are in your pic. these are 100% duplicated objects and not actual planar reflection, which are unthinkable given how modern games already struggle to render a single view point due to their poor optimization and obsession with graphic fidelity.
Mario 64 did mirrors by duplicating the room and Mario though.
Literal brainlett take, mixing and matching countless old memes into one non-cohesive mess.
For starters, the topic was about the truly oldschool mirror-room replication, not the planar reflection tech.
Second, planar reflections and any sort of render-to-texture methods can be optimized in countless ways. For starters, you don't need to have the reflections enabled at all times; just cull them when out of sight.
Third, lowering the rendering resolution, LOD calculations and other effects is just as easy to do for the rendered mirror reflections as it is for the main camera. Some games outright disable things like dynamic shadow casting for the reflections.
Fourth, you can mix and match various reflection methods, such as having SSR reflecting parts of the image whenever possible, while rest of the reflection's pixels are rendered using secondary camera footage. You could also utilize other now oldass tech, such as perspective-corrected cubemaps, with only the NPCs and player characters cast on top, the cubemap thus working like a fancier pre-rendered background.
Not only that, but you could even go as far as have the reflection run with the faster calculation methods shit like UE use for the main scene by default, meaning dithered 1-bit alpha and shading that then gets smoothed via TAA or even material-specific smoothing algorithm. Don't need to have secondary reflections or alpha necessarily either.
Last: any part of the screen the "mirror" covers, essentially culls stuff from the main camera. So again, only if you are a Pajeet-tier codelett would a mirror reflection cause problems in a modern video game.
these are 100% duplicated objects not actual planar reflection
Resident Evil uses replicated objects,
Max Payne 2 uses planar reflections. This is EASILY recognizable in-game, because the mirrors use a very low rendering resolution, that in the modern HD and 4K time sticks out like a sore thumb.
GTA5 and Hitman do Planar great.
damn, almost like devs made use of creativity to improve presentation when faced with limitations
Because they're copyrighted you dumb nigger
given how modern games already struggle to render a single view point due to their poor optimization and obsession with graphic fidelity.
Modern UE5 and RE-engine games struggle because there's no artistic touches whatsoever, and everything's automated + outsourced to Pajeet render farms.
Meanwhile, 2003-tier PCs running 1/2000th the processing power of the modern day consoles could do all this and then some in real time.
Does it increase the processing power that much?
Never understood this excuse.
Games today:
main character has 15,000 triangles, we can't duplicate them it's too much!
Games then:
mario only has 752 triangles easy to duplicate!
But
Processors today:
4.0 GHz (4 billion operations per second)
N64 processor:
62.5 MHz (62 million operations per second)
Modern CPUs are 64x stronger than N64 so why can't they duplicate 20x the triangles?
For starters, the topic was about the truly oldschool mirror-room replication, not the planar reflection tech.
nowhere in that post did I talk about planar reflection you fucking nigger
Literal brainlett take
your own irony is lost on you.
characters aren't just a bunch of triangles
nowhere in that post did I talk about planar reflection you fucking nigger
Yes, you did.
But apparently you didn't even realize this yourself, because as I predicted, you're a tech-illiterate zoomie who just parrots random bits and bobs he's read on the interwebs, and does not understand jack shit about the topic matter.
The whole ancient meme about "OMG you have to render the scene TWICE!!" is related to the traditional "planar reflections" alone. The origin of this narrative comes from the Valve's own voice documentary found in the Lost Coast chapter of HL2, where they describe their struggles of creating those kickass, smooth water reflections, that also got realistic refractions, depth- and distance-specific fogging effects and culling... etc.
The sentence essentially says that they need to do two passes for the view underneath the water, when viewed from above the surface level, just so that the intended effects would stack correctly. This does NOT equal "wow we need to run the ENTIRE SCENE twice!!", and does NOT mean "this will take TWICE THE COMPUTING POWAH!!!!" like plebs seem to claim.
Once again, we got plenty of recent AAA games (and even more indie titles) that prove that you can do sharp, non-RTX reflections just fine on very underwhelming specs. The new Hitman games use planar reflections for not only actual mirrors, but for literal window reflections as well. They can also be broken, with each part reflecting a slightly different angle.
GTA5 and many HL2 mods also do this.
Why did you reply to the right answer with embarrassingly cavalier wrong answer?
i dont even think its Dev incompetence but Nvidia giving Publishers money to fuck up on purpose so they can sell new GPUs
Games today:
main character has 15,000 triangles
That's literally Nintendo Gamecube numbers, mate.
Original RE4 Leon model has 10k triangles.
The MODERN AAA slop has MILLIONS of tris for the characters and their pubic hair. And they justify this with the typical mantra of "it's faster to draw triangles than it is to have HQ textures!".
Which then is demolished by the very same devs using a dozen 4K quality PBR texture sets per character.
But yes, you are correct on the performance to quality ratio not matching.
But even then, you are downplaying a lot of things:
modern games need 20 Gigabytes of VRAM to run smooth at 1080p, an obsolete poorfag resolution
retro games ran on literal single-digit Megabytes, to do the industry standard 480p SD quality
Hell, MGSV, a 2015 game, ran silky smooth 60fps on my 1 Gigabyte GTX 560ti back in the day, at High settings and 1080p. That if anything should tell how badly the optimization of vidya has declined.
The whole ancient meme about "OMG you have to render the scene TWICE!!" is related to the traditional "planar reflections" alone.
No you illiterate nigger, it means you have to render the same object twice depending on the angle you're viewing the duplicated scene. I accept your concession nonetheless, dumb fuck.
Why do 47s shades lose their lenses on the mirror world on my machine? 8s that a consistent thing in H3? I could never find a setting to fix it
just to finish you mongrel off, I am literally using the word "copy" in my original post
You could use workaround like making the copy use lower level of details and less intensive shader
BUT THOSE AREN'T ACTUALLY REAL MIRROR REFLECTIONS
THOSE ARE FAKE MIRROR REFLECTIONS USED TO GET AROUND THE HARDWARE CONSTRAINTS OF THE TIME
WHY AREN'T WE USING THEM FOR MODERN GAMES?
DUNNO
I did not. See: Toddlers like you just don't know what you're talking about.
A replicated mirror room method would require a tiny fraction of the performance cost of planar reflections, which themselves are not some space science voodoo like the Nvidia shills would like you to believe.
Otherwise, we would not have any vast open world games with dozens of NPCs roaming about.
No you illiterate nigger, it means you have to render the same object twice depending on the angle you're viewing the duplicated scene
And why would this be such a massive issue? Do you understand how many HUNDREDS of times we render the same objects in any typical video game?
Once again, you are mixing up countless unrelated subject matters, and don't understand the language you're speaking.
Object instancing has been a thing since DirectX10 days, eliminating draw calls to ONE (1).
It's the matter of calculating the millions of color combinations per frame, per each transparent pixel in the scene, dozens if not hundreds of times per second, that are the biggest performance drains in modern vidya. But once again, we had this shit sorted 20 years ago, and now run fuckton faster hardware to boot.
I am literally using the word "copy" in my original post
That does not free you from your idiotic statements, buster.
You are legit stating shit that only applies to Planar reflections. But you don't seem to understand this simple fact.
The room-copy shit would all run on the main scene, the main camera, meaning NO extra work for the GPU. The secondary, mirrored player character would literally be just another "NPC" in the game world. This is why even some NES games used such trick.
concession accepted
Not an argument, toddler.
As John Carmack,
Hey Carmack, hows it goin?
you can't just say "concession accepted" after someone explained why you're wrong
There is no argument to be had because there is no discussion since you couldn't even read my first post properly and immediately went schizophrenic.
pic
at least you're aware of your condition
his rant was about me not understand planar reflection even though I talked about duplicated scene. When I pointed that out he doubled down. I showed him in my original reply the proof and he triple down by ignoring it. That's a concession.
Ray tracing is a fraud, peak graphics it's still the first crysis from 2007
you couldn't even read my first post properly
But I did. And I replied to the idiotic points you made.
Let's go through it together, shall we?
OG question Anon presented:
What is stopping developers from just putting an [inaccessible copy of a room with a clone of your character in it]? Does it increase the processing power that much?
This is referring to the oldschool "fake" method of doing mirrors.
You have a closed "room", that has an "NPC" mimicking your actions.
It is literally done in the game-world, on the main camera. No extra calculations are needed compared to normal game scene.
It's fast and cheap, but requires a bit more planning because it takes some physical real-estate in the game world.
Now, your reply to this was, and I'll cut this into tiny parts :
Yeah, they put a ridiculous amount of triangles and shader instructions into the main characters and barely manage to hit 60 fps.
Unless your character model's polycount runs in the 100s of Millions of quads and packs numerous 8K textures, a single character will not tank your scene's performance.
An average street scene has dozens of NPCs, vehicles and whatnot running around, and requires a mere PS4 era hardware to operate, no matter the whole game being poorly designed and optimized in general.
This already raised my suspicion of you parroting the typical ancient arguments against properly working (planar) reflections in modern games. But then you continued with:
... But it's not just the player, depending on where the mirror is located and the viewpoint you'd have to potentially rendering intensive objects with [transparency included twice]
This is the smoking gun right here.
For starters, why would we have to render the SAME model twice, unless instancing is not coded into the engine? Also, most modern video games using the 1-bit dithered alpha to FAKE transperency, downsampling it to binary "ON/OFF" state, then blurring the output to poorly fake smoother see-thru parts.
nothing
that's how realtime RT works as well, you render the scene from a different point of view by casting rays
the only benefit to ray-traced reflections is that you can have reflections OF reflections with multiple bounces, but then developers go and limit it to a maximum of 1 or 2 bounces anyway so that idea goes out the window
you talked about perfomance increase, which matters way less if using a doubled mirror room vs. planar reflections.
so originally he assumed you must be talking about them, because otherwise your input would make no sense.
you just keep proving you don't know what you're talking about
But I did. And I replied to the idiotic points you made.
he quadruples down even though his rambling
about planar reflections as response to my post about duplicated scenes is still up
why are you like this?
I'm just gonna bite on this one because it's hilarious:
why would we have to render the SAME model twice, unless instancing is not coded into the engine?
what do you think instancing is? It not about magically not having to redraw an object twice but not having to send the same data more than once to the GPU. The triangles will still go through the shader pipeline. Funny how high and mighty you act when you not only can't read but don't understand what you are talking about.
you talked about perfomance increase
I talked about the impact of duplicating scenes and the reasons devs don't use them often anymore.
you just keep proving you don't know what you're talking about
you missed an occasion to shut the fuck up.
We don’t even need a physical space, as shown by non-eucledian techdemos
The lazy game devs’ excuses are even more pathetic
Either they are literally paid by nvidia to not do these shortcuts to sell rtx cards, or their middle managers are just as midwit as everyone else replying to you
Any sort of arguments against non-RT'd reflections go out of the window when you examine some PS4 era titles that managed them just fine:
They did this by literally just rendering everything in the room twice. It's an intentionally designed space such that doubling it would not be an excessive ask for the hardware. You can't necessarily do that with modern games with such high polycounts. But you could probably work around that by designing your environments properly. The real problem is that a lot of the visual effects and details that exist in games now are literally dependent on the position of the camera. They're essentially shortcuts to creating illusions of higher fidelity than the hardware is capable of rendering in real time on its own. So objects in the mirror might look like they're lit from the wrong direction, or simply render in a way that's obviously incorrect.
A solution like the one Hitman took in this post is both in a constrained space with limited geometry and is obviously using a rendering to texture method with predetermined splitting locations so each piece of the mirror can be tuned to look correct. This was a vanity set piece which they were naturally very proud of, because it does look good, but you'd never do this if it wasn't the centerpiece of a scene.