Actually you might want to do a little bit of research before making claims like that. Crysis had real time ambient occlusion way back in 2007. Crysis 2 has real time global illumination, though obviously not as high quality as longer solutions.
@TheAussieStew real time ambient occlusion is a form of global illumination isn't it? that SSAO right? console crysis 2 doesnt have it but crysis 1 and 2 on pc do.
SSAO is only in the screen space and is only in one colour space: black. It works in conjunction with systems like this and Crysis 2 to produce a more realistic effect. SSAO is used for console versions of Crysis 2 as well as many other console games, it's not that costly to run. But what console Crysis doesn't have, is global illumination. Crysis 2 DX11 also uses an upgraded version of SSAO, called SSDO that takes into account lights and colour and is more accurate.
Radiosity is the affecting of nearby surfaces based on the current surface's brightness and color. Much like the real world, if a white light shines on a red object, with radiosity, red will be projected in to the room.
Depending on the engine, it may be calculated in reverse - with a surface "checking" the nearby surfaces for their color and brightness and changing accordingly.
IF is was real-time then why haven't they done a live demo and move the camera around instead of this preset camera path? looks cool but it's offline rendered.. nothing new
Well as the light moves, the entire lighting calculations change. Search for Geomerics Enlighten Workflow, it underscores the purely realtime ambitions, even with animated meshes. Assuming this isn't all extraordinarily misleading advertising, it looks like realtime assets will finally compare favorably in nearly every aspect to render-farmed results. Search for LightSprint for another similar light engine on the horizon. Man I can't wait to play with these!
how much is precomputed of this? is it like wann jensen's demo? a lot of points in the scene are precomputed and if they get lit there is light data for the whole scene for each one of them
you might be confusing my post with someone else's. I was talking about doom3 not crysis. someone else criticized crysis. I will agree though that while crysis looked good it still suffers from that indeterminate ambient illumination, maybe mostly just because they soften the sun to even out the light. I think it's possible to make something look really photoreal crisp even with today's hardware. maybe by using actual locations, a super version of google street view. I see it coming
This is what i won't ever get, how is it that they render photoreal things in 60fps at 1024* or much higher, and when it comes to a rendering software it takes hours for a simple frame?!
And if some very clever guy comes here to easily answer this, i just want you to feel the problem for a second... i'm not interested if games use the video card for their graphics, and programs use the CPU, or whatever, just feel the fuckin problem with me. It is annoying as hell.
lol I was wondering the same thing.... like how can my pc run crysis at medium to high at like 20~30 fps and can't render some low poly scene (w/ AA/Ambient Occlusion/Radiosity) in less than like 10seconds?
I KNOW, IT'S SO ANNOYING! THANKFULLY, there was a simple answer to this. See, it's all in the programming. A game can be programmed to take full advantage of that huge amount of processing power in the v. card, it's actually faster than the cpu. Programs that use the CPU to render simply cannot use the card due to some paradox or something that's similar to dual/quad cores and how you can't really spread physics onto 2-4 cores very easily. You can look around on google perhaps, it has everything
It's a problem of quality and scale. Games are optimized to render a few order of magnitude less data than offline renderers, and with much less quality (even if it's not obvious but the kind of defects we accept in games wouldn't be accepted in other realms - just grab a real screenshot from a game).
Of course, for what they do they are fast, but they won't scale to do the offline work faster than offline renderers. Also for offline, speed is secondary, first thing is stability and quality.
Okay but i wasn't necessarily talking about GI and dozens of shaders, the very simple scanline render for example in MAX takes seconds for some boxes, and more seconds for a little more advanced geometry. Which is renderable at hundreds of fps in games. Thats just unacceptable for me.
If it is that obvious, than make a freakin button for 'ingame' rendering into these programs.
there is already, max can use DirectX shaders in its viewports (the directx material), but it sux (and it's still slower than a game because the game is also optimized to do only what it has to do, while max is optimized to edit and create stuff).
Anyway, always presume that people can do their job, and thus there is a reason why things are like they are. Sometimes is false, but usually its not. You won't like for example, to have a 3d app that is fast, but has a lot of constraints, as game do
Hmm. The difference is that in games most effect you see are fake. Shadows for example is not calculated but part of the texture (fake shadows). Specularity and reflection is almost always absent or faked in games. Textures are very low. Light is calculated inaccurately (but fast) and for some are even faked (only textures). In some games shadows of moving objects are absent. That is the explanation why. Game renderers avoid those computationally expensive calculations or make shortcuts.
Tsk, tsk... your mistaken. Games now may have developed but this things have not changed yet. Try playing your favorite 3d games and see those cheats. Try looking for accurate and soft shadows. See how objects are lighted. Try blocking light sources. Try to look for color bleed or indirect lighting. You will find that games never render this things accurately.
And oh! by the way. Games takes advantage of video cards. Video cards are separate cpu's that is dedicated for graphics. Its instruction set and logic is optimized for this. Still it is unable to render realistic images very fast. But because in games, images are always moving and some inaccuracy is accepted, such things are not noticed. Also, are you not talking about videos in games when you mean realistic? They are videos, recorded, not in real time.
@AMYuntold if they have some sort of script that samples the color of the primary hit object and then create a point light ofsseted along the normal of the hit surface in such a way that you get 1 or 2 lights per object that do not cast shadows; it's possible. that's how i'd do it to fake radiosity
@AMYuntold They aren't rendering photoreal things. Its just good enough to make you think its photoreal. In a production renderer they do accurate, physically correct lighting simulations. This is not nearly as accurate, nor as physically correct, but is enough so that you don't see the difference.
@AMYuntold - Because there's a dick-waving contest between real-time and pre-rendered graphics guys, and pre-renderers like to take advantage of their ridiculous amount of time.
Those very long rendering are made with actual simulations, and this real-time thing is made of a lot of heuristics (eureka!) stuff. So (I'm guessing) this could be a REALLY controlled environment, where you really have to take care about geometry, movement (things have to be still), etc. to maintain the looks very coherent, whereas those long scenes have a LOT of detail and more freedom of creation.
You could spend a month making something that runs in realtime using every trick in the book. Or you could spend a few days writing an inefficient raytracer (that still looks good), and let the computer suffer.
Because photo realistic rendering engines simulate real life physical phenomena and do not use "shortcuts" to achieve the somewhat similar results. Also CPU is unfortunately slower than GPU in these kinds of calculations, so i do not understand why we still use purely CPU dependent rendering solutions. But that is about to change soon because GPU rendering solutions are appearing to the market allowing up to 100 times faster rendering in 3d applications such as 3dsMax.
@AMYuntold For those wondering why GPU is not used everywhere instead of CPU .. is that because, of course, for real productions scenes the overall memory would not fit on GPU memory. For simples scene like CAD stuff and such theres no dubts that the GPU approach will revolution the entire field.
Games and applications that do these real-time things on GPU uses shaders and other techniques, that gives imperfect results, this way that they are faster, because in real-time you simply don't see these imperfections.
In the CPU models, we don't use shaders, normally is done ray-tracing, that is, every pixel have an ray that is calculated to give photo-realistic results, most of them almost (but not) perfect.
@AMYuntold i think you've got the answers you wanted. as a mildly experienced 3d creator, i would say that in real time engines, like this, it uses optimized builds, baked textures/shadows, and approximate effects.
in a high quality graphics scenario, like anything pixar has made in the last 5 years, uses raytraced lighting, reflections, transparency, diffusion, motion blur, depth of field, volumetrics, and other amazing effects you cant get in real time. (crysis 3 is hopefully an exception)
@AMYuntold both are using completely different techniques to achieve this.. games are absolutely not accurate and use a huge amount of fake techniques where rendering software is doing the real thing, simulating lightrays as they behave in the real world to achieve photo realism. That way is easier than developing all those fake techniques for realtime rendering. And rendering software need to be able to render any kind of scene or object which is not possible with such a 3d engine.
@AMYuntold The rendering software do a technique called Raytracing, that does that with an very small error margin, the GPU does things that look good but not perfect, you just don't notice some small errors because of high framerate, as you can't notice a small 2x2 pixel error somewhere on screen for just one frame, for example...
Simple, games are pre-rendered and almost everything you see on games are textures with lighting and displacement data to "fake" shadows and bumps, so no high geometry is used and some buffer shadows are used too, not ray traced shadows that are very very much slower to render but are high quiality shadows. reflections are reflection maps so no reflectios and almost everything on games is "faked" for fast rendering
@AMYuntold it's called a rendering network, companies that make these probably have 10 computers running at once with combined clock speed and graphical power to easily render this shit out, and those are probably computers just being used for rendering alone lol.
rendering speed is no problem if you've got tons and tons of money to spend
@AMYuntold it's called a rendering network, companies that make these probably have 10 computers running at once with combined clock speed and graphical power to easily render this shit out, and those are probably computers just being used for rendering alone lol.
rendering speed is no problem if you've got tons and tons of money to spend
and next to that, games are prerendered, that's why engine's are so important, because they place the rendered stuff exactly where it belongs.
@AMYuntold chillax man programms are being developed for use with GPU's so things are getting faster also this is a solver witch has BUGs thats normal in solvers because they dont have many substeps and lots of errors happen
switch renderer. There is no photoreal renderer/engine that can render photoreal frames at 60fps unless the scene is empty or something but you rarely need photorealism for believable results, if a single frame takes several hours you're probably doing something wrong
@AMYuntold I think you posted this 2 years ago so you know by now, right? If not: Games use rasterization and shaders for all the effects. They are basically faking reality.
But 3D programs use raytracing, no shaders. Real reflections and lighting. It takes a lot more processing power but you get more detail and realism.
@AMYuntold rendering software usually uses methods that are a lot more photorealistic, but a lot longer to accomplish. Games usually make some concession, they use faster algorithms with slighty less accurate result and then they try to make it look better
@AMYuntold To answer your question: Video games use something called rasterization to render. It and the associated techniques are extremely mathematically complicated, BUT render VERY fast. However, to call it photoreal is not exactly true. Realtime rendering is almost entirely composed of shortcuts and tricks, which reduces the photorealism. You just may not notice because they are very good about hiding this fact.
@AMYuntold Movies, on the other hand, are rendered using ray tracing. It's far simpler, and more elegant, because it models how light actually works. It's slow because of how light works: When light hits an object, part of that light bounces off, part is absorbed, and part is sent through the material (think glass). Which each pixel on the screen casting a light ray that can do one of three things for many many bounces, the complexity of the scene can make rendering time exponential.
@AMYuntold This video is showing radiosity, which allows for "one bounce" (or maybe 2) rendering effects. Ray tracing potentially allows for "infinite bounces" of light. Something ray tracing can do that realtime rendering cannot: two mirrors aimed at each other.
Now, ray tracing CAN be done in real time, BUT you either need custom hardware (thousands of dollars) or you can only have very simple scenes (a couple of spheres and a box).
Note that while this technique looks gorgeous, it is still limited: only the light is allowed to move; the rest of the geometry must be fixed. Compare this to, say, the ambient lighting system used in the Source Engine. Here, the geometry can move, but the lights must be fixed.
The best of both worlds will probably have to wait for real time ray tracing.
Lightsprint can have objects moving but I'm not sure to what degree it affects the quality of it. For example, if you watch the demo, on the penumbra shadows section the robot moves but the penumbra is insanely noisy.
ray tracing is only good for reflections/refractions. It doesent solve the realistic illumination like path tracing (or another less time consuming algorithms) which is like a lot of times slower than ray tracing.
I heard about fake radiosity,called Screen Space Radiosity/Light Bleeding/Global Illumination or something else.The only thing I remember is that its Screen Space.
Consider a red floor and a yellow wall. The red floor reflects it's red ligth to the yellow wall, the now more red wall reflects the the new yellow (the more red one) to the floor. The more often you repeat this, the more realistic the final shading is. All in all it's a mess, and you have to set limits, on how realistic you want it. And i don't think, that the above is realtime radiosity, or real radiosity. Real radiosity images hardly differ from reality, even simple scenes.
to be honest,I don't think this is real-time.I know real time radiosity is truly possible,but this video looks like it's pre-rendered.If anyone can point me to a download of this demo,I'll be very happy XD!
Actually they stated that Geomerics has joined the Integrated Partners Program, which I believe will give developers the _option_ of using Geomerics technology in Unreal Engine 3.
That's my point..Skrew the Unreal Engine! Geomerics should have started their own Engine from the start. Look what NaturalMotion is doing with Euphoria and that football game BackBreaker : p ))
I have a more depressing opinion. Since Real-Time Readiosity take up one of the Xbox 360's (3) major cpu threads, Developers will probable end up using the Pre-baked Readosity illuminatelabs provides. They place their readiositiy inside the Textures. Their Games include Haze, Killzone 2, and Mass Effect.
This is incredible! I am so looking foward to see games implement this even though lighting looks good in Crysis and Far Cry 2 does implement some advanced indirect lighting looking much alike this radiosity.
This video is close to a year old... And we are yet to see Geometric techno anywhere else then in their movie. They could at least make a playable demo, don't you think?
Uh they demonstrated it at GDC and they released another vid since. And why would they make a playable demo. They are not a game developer. Plus last year I doubt the product was fully finished. Furthermore, take a look at Euphoria. I heard about it along time and it took FOREVER for the first in engine vid to be released. HEck even with the number of companies on board, not one game using it has been released yet. These things typically tend to take time.
It looks just like any other game to me but you guys seem to be experts. The whole point about Geomerics is that they have new equations which can calculate lighting effects many times quicker than now, so having this in real time is a reality.
4) in the full size screen shots you can see blocky edges of primary shadows (shadow buffer technique?), its only the indirect lighting that is soft
5) i think 'infinite bounce' just refers to it being an attempt at radiosity, there may be some fixed limit (2-3 is probably fine), just remember this includes diffuse reflection also now -- as for fake hdr, hdr just implies they use unclamped values for intensity, the bloom is the fake part in any app
Don't get me wrong... If they actually can do what they claim they can, it will be a revolution in games graphics. However, untill I see it as a scene I can walk in and interact with it, it will remain a huge pile of stuff anybody with a computer and any 3D rendering software can do in his weekend.
Lightstriker: Ever heard of Gears of War? Geomerics wrote the Enlighten engine for it. They don't bullshit. This is real-time, it's not faked, and it is revolutionary. But congrats on the four full-length posts that make you out to be an uninformed ass-hat.
You would like that right? Well, guess what, I have GoW at home... And there is 0 radiosity in it. Next time, try to bring argument to your post insteed of just stupid pointless flame.
5) Lot of the infos in this video are some impressivly uber crap. Infinite ray bounce? You are aware that even in high end CG movie they often limited their ray bounce from radiosity to 2-3 as anything more is pointless? HDRI? Yes, it's being introduced into games... But it's faked. It's prerendered HDRI solution, combine with expensive calculation for dynamic object. Check Lost Coast from HL2. It requires quite a computer to run it... And there is only a fake HDRI.
3) I see some (a lot) of Anti-aliasing. Even modern game has a hard time to use it completly as it often require double or triple the processing power.
4) Shadows are not per-pixel nor stancil shadow, the currently 2 used technique in games. In this video, I see ray traced shadow, what is used in pre-rendering system. Soft shadow is currently impossible in games, as the 2 technique used, per pixel and stencil, doesn't allow easy and smooth integration of soft shadow.
It was 5 years ago, and still no radiosity in videogames :(
bestplugins 5 months ago
@bestplugins
There was radiosity in Crysis 2 and Stalker, not to mention the upcoming BF3.
TheAussieStew 5 months ago
@TheAussieStew prerendered textures, no real radiosity, BF3 has ambient occlussion, close to radiosity, but less realistic
bestplugins 5 months ago
@bestplugins
Actually you might want to do a little bit of research before making claims like that. Crysis had real time ambient occlusion way back in 2007. Crysis 2 has real time global illumination, though obviously not as high quality as longer solutions.
TheAussieStew 5 months ago
@TheAussieStew real time ambient occlusion is a form of global illumination isn't it? that SSAO right? console crysis 2 doesnt have it but crysis 1 and 2 on pc do.
dageezerboi 4 months ago
@dageezerboi
SSAO is only in the screen space and is only in one colour space: black. It works in conjunction with systems like this and Crysis 2 to produce a more realistic effect. SSAO is used for console versions of Crysis 2 as well as many other console games, it's not that costly to run. But what console Crysis doesn't have, is global illumination. Crysis 2 DX11 also uses an upgraded version of SSAO, called SSDO that takes into account lights and colour and is more accurate.
TheAussieStew 4 months ago
@bestplugins
/watch?v=O7MnzjTLxJA
TheAussieStew 5 months ago
so the Frostbite 2.0 engine does this easily with 1 computer?
megafio 7 months ago
This isnt radiosity, its global illumination... seriously, people...
Inityx 7 months ago
@Inityx
of which radiosity is an algorithm of... seriously...
gm3dgames 7 months ago
lukeeales, Please, what is this song?
yanasitta 8 months ago
lukeeales, Please, what is this song?
yanasitta 8 months ago
Radiosity is the affecting of nearby surfaces based on the current surface's brightness and color. Much like the real world, if a white light shines on a red object, with radiosity, red will be projected in to the room.
Depending on the engine, it may be calculated in reverse - with a surface "checking" the nearby surfaces for their color and brightness and changing accordingly.
Dracorat 9 months ago
radiosity? lightmaps? textures? where can I learn all these terms? is there a site which has all terms for 3d grahpic rendering?
ScrapeNow 10 months ago
SONG???
johhhavboh 11 months ago
i don't understand how real-time radiosity is so unbelievable.
just a matter of setting the light source as sort of a relay, right?
qiman2 1 year ago
@qiman2 take a graphics design class then we'll talk
Zarkyun 1 year ago
@Zarkyun Yesss!! I don't understand any of this Radiosity stuff!! All I know is that battlefield 3 is using it!!!
Thet3 1 year ago
Comment removed
hq2x 1 year ago
IF is was real-time then why haven't they done a live demo and move the camera around instead of this preset camera path? looks cool but it's offline rendered.. nothing new
randomdude3824 1 year ago
Well as the light moves, the entire lighting calculations change. Search for Geomerics Enlighten Workflow, it underscores the purely realtime ambitions, even with animated meshes. Assuming this isn't all extraordinarily misleading advertising, it looks like realtime assets will finally compare favorably in nearly every aspect to render-farmed results. Search for LightSprint for another similar light engine on the horizon. Man I can't wait to play with these!
TandyDandy 1 year ago
meaning it doesn't work if anything moves
DanFrederiksen 2 years ago
how much is precomputed of this? is it like wann jensen's demo? a lot of points in the scene are precomputed and if they get lit there is light data for the whole scene for each one of them
DanFrederiksen 2 years ago
There's really no proof that this is real-time...
rawrEvan 2 years ago
No proof that it's not.
I remember when the first Doom 3 tech demo was shown and everyone was like "FAEK!!!"
fail.
masterpiraka 2 years ago 18
doom3 was never impressive though. it was always ugly. the released game was even impressively ugly
DanFrederiksen 2 years ago
Sure the lighting was bad, but no one really saw anything better at the time.
It was the shit. It was Crysis. It is foolish to say otherwise.
masterpiraka 2 years ago
you might be confusing my post with someone else's. I was talking about doom3 not crysis. someone else criticized crysis. I will agree though that while crysis looked good it still suffers from that indeterminate ambient illumination, maybe mostly just because they soften the sun to even out the light. I think it's possible to make something look really photoreal crisp even with today's hardware. maybe by using actual locations, a super version of google street view. I see it coming
DanFrederiksen 2 years ago
I was comparing Doom 3 to Crysis.
Doom 3 was the crysis of the time. No one had a rig that culd run it with full everything, despite the fact that it wasn't visually perfect.
masterpiraka 2 years ago
i love enlighten but it sucks that dynamic object dont affect the light..so breakable house wont be dark inside (if im right)
i wish there will be an update for future console generation to support it.
mat967 2 years ago
This is what i won't ever get, how is it that they render photoreal things in 60fps at 1024* or much higher, and when it comes to a rendering software it takes hours for a simple frame?!
And if some very clever guy comes here to easily answer this, i just want you to feel the problem for a second... i'm not interested if games use the video card for their graphics, and programs use the CPU, or whatever, just feel the fuckin problem with me. It is annoying as hell.
AMYuntold 2 years ago 13
lol I was wondering the same thing.... like how can my pc run crysis at medium to high at like 20~30 fps and can't render some low poly scene (w/ AA/Ambient Occlusion/Radiosity) in less than like 10seconds?
djarnexus 2 years ago
I KNOW, IT'S SO ANNOYING! THANKFULLY, there was a simple answer to this. See, it's all in the programming. A game can be programmed to take full advantage of that huge amount of processing power in the v. card, it's actually faster than the cpu. Programs that use the CPU to render simply cannot use the card due to some paradox or something that's similar to dual/quad cores and how you can't really spread physics onto 2-4 cores very easily. You can look around on google perhaps, it has everything
super6plx 2 years ago
It's a problem of quality and scale. Games are optimized to render a few order of magnitude less data than offline renderers, and with much less quality (even if it's not obvious but the kind of defects we accept in games wouldn't be accepted in other realms - just grab a real screenshot from a game).
Of course, for what they do they are fast, but they won't scale to do the offline work faster than offline renderers. Also for offline, speed is secondary, first thing is stability and quality.
PesceA 2 years ago
Okay but i wasn't necessarily talking about GI and dozens of shaders, the very simple scanline render for example in MAX takes seconds for some boxes, and more seconds for a little more advanced geometry. Which is renderable at hundreds of fps in games. Thats just unacceptable for me.
If it is that obvious, than make a freakin button for 'ingame' rendering into these programs.
AMYuntold 2 years ago
there is already, max can use DirectX shaders in its viewports (the directx material), but it sux (and it's still slower than a game because the game is also optimized to do only what it has to do, while max is optimized to edit and create stuff).
Anyway, always presume that people can do their job, and thus there is a reason why things are like they are. Sometimes is false, but usually its not. You won't like for example, to have a 3d app that is fast, but has a lot of constraints, as game do
PesceA 2 years ago
Hmm. The difference is that in games most effect you see are fake. Shadows for example is not calculated but part of the texture (fake shadows). Specularity and reflection is almost always absent or faked in games. Textures are very low. Light is calculated inaccurately (but fast) and for some are even faked (only textures). In some games shadows of moving objects are absent. That is the explanation why. Game renderers avoid those computationally expensive calculations or make shortcuts.
mapflu 2 years ago
You described here a game from about 5 years ago...
AMYuntold 2 years ago
Tsk, tsk... your mistaken. Games now may have developed but this things have not changed yet. Try playing your favorite 3d games and see those cheats. Try looking for accurate and soft shadows. See how objects are lighted. Try blocking light sources. Try to look for color bleed or indirect lighting. You will find that games never render this things accurately.
mapflu 2 years ago
And oh! by the way. Games takes advantage of video cards. Video cards are separate cpu's that is dedicated for graphics. Its instruction set and logic is optimized for this. Still it is unable to render realistic images very fast. But because in games, images are always moving and some inaccuracy is accepted, such things are not noticed. Also, are you not talking about videos in games when you mean realistic? They are videos, recorded, not in real time.
mapflu 2 years ago
@AMYuntold if they have some sort of script that samples the color of the primary hit object and then create a point light ofsseted along the normal of the hit surface in such a way that you get 1 or 2 lights per object that do not cast shadows; it's possible. that's how i'd do it to fake radiosity
TheNemesiah 1 year ago
@AMYuntold They aren't rendering photoreal things. Its just good enough to make you think its photoreal. In a production renderer they do accurate, physically correct lighting simulations. This is not nearly as accurate, nor as physically correct, but is enough so that you don't see the difference.
ChewyGumbal 1 year ago 2
@AMYuntold - Because there's a dick-waving contest between real-time and pre-rendered graphics guys, and pre-renderers like to take advantage of their ridiculous amount of time.
cyborgtroy 1 year ago
@AMYuntold I feel the same as u.
Those very long rendering are made with actual simulations, and this real-time thing is made of a lot of heuristics (eureka!) stuff. So (I'm guessing) this could be a REALLY controlled environment, where you really have to take care about geometry, movement (things have to be still), etc. to maintain the looks very coherent, whereas those long scenes have a LOT of detail and more freedom of creation.
riverjoe128 1 year ago
@AMYuntold I reckon it's plain lazyness.
You could spend a month making something that runs in realtime using every trick in the book. Or you could spend a few days writing an inefficient raytracer (that still looks good), and let the computer suffer.
matieman77 1 year ago
@AMYuntold
Because photo realistic rendering engines simulate real life physical phenomena and do not use "shortcuts" to achieve the somewhat similar results. Also CPU is unfortunately slower than GPU in these kinds of calculations, so i do not understand why we still use purely CPU dependent rendering solutions. But that is about to change soon because GPU rendering solutions are appearing to the market allowing up to 100 times faster rendering in 3d applications such as 3dsMax.
nVidia6000 1 year ago
@AMYuntold For those wondering why GPU is not used everywhere instead of CPU .. is that because, of course, for real productions scenes the overall memory would not fit on GPU memory. For simples scene like CAD stuff and such theres no dubts that the GPU approach will revolution the entire field.
maxtarpini 1 year ago
@AMYuntold I would simply say that:
Games and applications that do these real-time things on GPU uses shaders and other techniques, that gives imperfect results, this way that they are faster, because in real-time you simply don't see these imperfections.
In the CPU models, we don't use shaders, normally is done ray-tracing, that is, every pixel have an ray that is calculated to give photo-realistic results, most of them almost (but not) perfect.
MatheusMK3 1 year ago
@AMYuntold i think you've got the answers you wanted. as a mildly experienced 3d creator, i would say that in real time engines, like this, it uses optimized builds, baked textures/shadows, and approximate effects.
in a high quality graphics scenario, like anything pixar has made in the last 5 years, uses raytraced lighting, reflections, transparency, diffusion, motion blur, depth of field, volumetrics, and other amazing effects you cant get in real time. (crysis 3 is hopefully an exception)
kerog6 1 year ago
@AMYuntold both are using completely different techniques to achieve this.. games are absolutely not accurate and use a huge amount of fake techniques where rendering software is doing the real thing, simulating lightrays as they behave in the real world to achieve photo realism. That way is easier than developing all those fake techniques for realtime rendering. And rendering software need to be able to render any kind of scene or object which is not possible with such a 3d engine.
ecreif 1 year ago 2
@AMYuntold The rendering software do a technique called Raytracing, that does that with an very small error margin, the GPU does things that look good but not perfect, you just don't notice some small errors because of high framerate, as you can't notice a small 2x2 pixel error somewhere on screen for just one frame, for example...
MatheusMK3 11 months ago
@AMYuntold
Simple, games are pre-rendered and almost everything you see on games are textures with lighting and displacement data to "fake" shadows and bumps, so no high geometry is used and some buffer shadows are used too, not ray traced shadows that are very very much slower to render but are high quiality shadows. reflections are reflection maps so no reflectios and almost everything on games is "faked" for fast rendering
jlrazr 11 months ago
@AMYuntold
look for realtime path tracing
Dostape 9 months ago
@AMYuntold
Games are using preRENDERED graphic's
Damatelon 8 months ago
@AMYuntold it's called a rendering network, companies that make these probably have 10 computers running at once with combined clock speed and graphical power to easily render this shit out, and those are probably computers just being used for rendering alone lol.
rendering speed is no problem if you've got tons and tons of money to spend
GuitarGasmProduction 7 months ago
@AMYuntold it's called a rendering network, companies that make these probably have 10 computers running at once with combined clock speed and graphical power to easily render this shit out, and those are probably computers just being used for rendering alone lol.
rendering speed is no problem if you've got tons and tons of money to spend
and next to that, games are prerendered, that's why engine's are so important, because they place the rendered stuff exactly where it belongs.
GuitarGasmProduction 7 months ago
@AMYuntold chillax man programms are being developed for use with GPU's so things are getting faster also this is a solver witch has BUGs thats normal in solvers because they dont have many substeps and lots of errors happen
dahahaka 5 months ago
switch renderer. There is no photoreal renderer/engine that can render photoreal frames at 60fps unless the scene is empty or something but you rarely need photorealism for believable results, if a single frame takes several hours you're probably doing something wrong
haloandre 4 months ago
@AMYuntold I think you posted this 2 years ago so you know by now, right? If not: Games use rasterization and shaders for all the effects. They are basically faking reality.
But 3D programs use raytracing, no shaders. Real reflections and lighting. It takes a lot more processing power but you get more detail and realism.
TrickyEmu 1 month ago
@AMYuntold rendering software usually uses methods that are a lot more photorealistic, but a lot longer to accomplish. Games usually make some concession, they use faster algorithms with slighty less accurate result and then they try to make it look better
stephbeaudet 1 month ago
@AMYuntold To answer your question: Video games use something called rasterization to render. It and the associated techniques are extremely mathematically complicated, BUT render VERY fast. However, to call it photoreal is not exactly true. Realtime rendering is almost entirely composed of shortcuts and tricks, which reduces the photorealism. You just may not notice because they are very good about hiding this fact.
dwightdesign 2 days ago
@AMYuntold Movies, on the other hand, are rendered using ray tracing. It's far simpler, and more elegant, because it models how light actually works. It's slow because of how light works: When light hits an object, part of that light bounces off, part is absorbed, and part is sent through the material (think glass). Which each pixel on the screen casting a light ray that can do one of three things for many many bounces, the complexity of the scene can make rendering time exponential.
dwightdesign 2 days ago
@AMYuntold This video is showing radiosity, which allows for "one bounce" (or maybe 2) rendering effects. Ray tracing potentially allows for "infinite bounces" of light. Something ray tracing can do that realtime rendering cannot: two mirrors aimed at each other.
Now, ray tracing CAN be done in real time, BUT you either need custom hardware (thousands of dollars) or you can only have very simple scenes (a couple of spheres and a box).
Hope you understand now.
dwightdesign 2 days ago
Note that while this technique looks gorgeous, it is still limited: only the light is allowed to move; the rest of the geometry must be fixed. Compare this to, say, the ambient lighting system used in the Source Engine. Here, the geometry can move, but the lights must be fixed.
The best of both worlds will probably have to wait for real time ray tracing.
BrassVaz 2 years ago
Lightsprint can have objects moving but I'm not sure to what degree it affects the quality of it. For example, if you watch the demo, on the penumbra shadows section the robot moves but the penumbra is insanely noisy.
gbushimprov 2 years ago
ray tracing is only good for reflections/refractions. It doesent solve the realistic illumination like path tracing (or another less time consuming algorithms) which is like a lot of times slower than ray tracing.
JonnyRobbie 2 years ago
man this is 3 years old and it looks better then crysis
mrprofile101 2 years ago 2
beautiful
boilerheadz 2 years ago
Realtime radiosity is possible. There are a number of papers on the net that describe how to do it in different ways.
How is this possible? a) programmable pixel shaders and b) very fast video cards and c) good programmers :^)
M1nd5scape 2 years ago
I heard about fake radiosity,called Screen Space Radiosity/Light Bleeding/Global Illumination or something else.The only thing I remember is that its Screen Space.
speedaddict2008 2 years ago
what is radiosity?
jackabascal 3 years ago
Consider a red floor and a yellow wall. The red floor reflects it's red ligth to the yellow wall, the now more red wall reflects the the new yellow (the more red one) to the floor. The more often you repeat this, the more realistic the final shading is. All in all it's a mess, and you have to set limits, on how realistic you want it. And i don't think, that the above is realtime radiosity, or real radiosity. Real radiosity images hardly differ from reality, even simple scenes.
agrimm61 3 years ago 3
wow your smart!
jackabascal 3 years ago
cool vid
kalsikum 3 years ago
to be honest,I don't think this is real-time.I know real time radiosity is truly possible,but this video looks like it's pre-rendered.If anyone can point me to a download of this demo,I'll be very happy XD!
speedaddict2008 3 years ago
Also, what is the name of this song, I love it!
tosmos1 3 years ago
Wow. Great radiosity, I'm sure that with focused development, this technique (or engine) could be used in games
tosmos1 3 years ago
I wonder how they got it to run in realtime. Like, some kinda radio buffer?
snezegard 3 years ago
Real radiosity is very slow. It's not possible to make real-time radiosity without sacrificing accuracy.
I've seen many (true) radiosity scenes, and this doesn't look as good IMO. None the less, it doesn't look bad either, so it's a nice effort.
johnking87 3 years ago
Youre a pro!
starncat 3 years ago
whats the name of this song nice video btw the way!
corbono35 3 years ago
I hope that, people who uses this in their games engine, will look into gameplay too. There is too many games out there that's just about graphics.
God, I love to bash everything that's about realism in games these days. They just don't make anything as they were in the good ol' days.
Hasuak 3 years ago 3
we hear ya
Nkatsikanis 3 years ago
Actually they stated that Geomerics has joined the Integrated Partners Program, which I believe will give developers the _option_ of using Geomerics technology in Unreal Engine 3.
Maexxus 3 years ago
That's my point..Skrew the Unreal Engine! Geomerics should have started their own Engine from the start. Look what NaturalMotion is doing with Euphoria and that football game BackBreaker : p ))
St1kyFinguz 3 years ago
I have a more depressing opinion. Since Real-Time Readiosity take up one of the Xbox 360's (3) major cpu threads, Developers will probable end up using the Pre-baked Readosity illuminatelabs provides. They place their readiositiy inside the Textures. Their Games include Haze, Killzone 2, and Mass Effect.
St1kyFinguz 4 years ago
This is incredible! I am so looking foward to see games implement this even though lighting looks good in Crysis and Far Cry 2 does implement some advanced indirect lighting looking much alike this radiosity.
unlicter 4 years ago
Too bad theres no color bleeding :/
Sakuukuli 3 years ago
Huh, I am pretty sure I am seeing some.
shletten 3 years ago
I mean in the CryEngine 2 theres no color bleeding.
Sakuukuli 3 years ago
!!! This is friggin awesome!
Grobulous 4 years ago
uhhh its proven that you can do it in games. just google it i found a download link for this thing. its real time interactive.
ragdollpwner 4 years ago
I don't think Epic would have incorporated this into thier Unreal Engine if it was fake Mr. Lightstriker. They announced it last month.
lumzi23 4 years ago
I put my hands in Unreal 3... And there is no real time radiosity in in etheir.
Lightstriker 4 years ago
There no games using it yet. it was a recent announcement, post-GDC. I am pretty sure mentioned 'last month' 'somewhere.'
lumzi23 4 years ago
This video is close to a year old... And we are yet to see Geometric techno anywhere else then in their movie. They could at least make a playable demo, don't you think?
Lightstriker 4 years ago
Uh they demonstrated it at GDC and they released another vid since. And why would they make a playable demo. They are not a game developer. Plus last year I doubt the product was fully finished. Furthermore, take a look at Euphoria. I heard about it along time and it took FOREVER for the first in engine vid to be released. HEck even with the number of companies on board, not one game using it has been released yet. These things typically tend to take time.
lumzi23 4 years ago
It looks just like any other game to me but you guys seem to be experts. The whole point about Geomerics is that they have new equations which can calculate lighting effects many times quicker than now, so having this in real time is a reality.
BreezeOnTheKeys 4 years ago
4) in the full size screen shots you can see blocky edges of primary shadows (shadow buffer technique?), its only the indirect lighting that is soft
5) i think 'infinite bounce' just refers to it being an attempt at radiosity, there may be some fixed limit (2-3 is probably fine), just remember this includes diffuse reflection also now -- as for fake hdr, hdr just implies they use unclamped values for intensity, the bloom is the fake part in any app
rndmcnlly 4 years ago
Don't get me wrong... If they actually can do what they claim they can, it will be a revolution in games graphics. However, untill I see it as a scene I can walk in and interact with it, it will remain a huge pile of stuff anybody with a computer and any 3D rendering software can do in his weekend.
Lightstriker 5 years ago
Lightstriker: Ever heard of Gears of War? Geomerics wrote the Enlighten engine for it. They don't bullshit. This is real-time, it's not faked, and it is revolutionary. But congrats on the four full-length posts that make you out to be an uninformed ass-hat.
VigEuno 4 years ago
You would like that right? Well, guess what, I have GoW at home... And there is 0 radiosity in it. Next time, try to bring argument to your post insteed of just stupid pointless flame.
Lightstriker 4 years ago
5) Lot of the infos in this video are some impressivly uber crap. Infinite ray bounce? You are aware that even in high end CG movie they often limited their ray bounce from radiosity to 2-3 as anything more is pointless? HDRI? Yes, it's being introduced into games... But it's faked. It's prerendered HDRI solution, combine with expensive calculation for dynamic object. Check Lost Coast from HL2. It requires quite a computer to run it... And there is only a fake HDRI.
Lightstriker 5 years ago
3) I see some (a lot) of Anti-aliasing. Even modern game has a hard time to use it completly as it often require double or triple the processing power.
4) Shadows are not per-pixel nor stancil shadow, the currently 2 used technique in games. In this video, I see ray traced shadow, what is used in pre-rendering system. Soft shadow is currently impossible in games, as the 2 technique used, per pixel and stencil, doesn't allow easy and smooth integration of soft shadow.
Lightstriker 5 years ago
Sorry to burst your bubble.. But I see no proof it's made in real time.
However, I see PLENTY of proof it's pre-rendered.
1) Radiosity, even fake, take a LOT of processing power. Mathematicly speaking, I hardly see how any radiosity could be calculated in real time.
2) I see no segmentation. Even with all the progress, polycount in a game scene stay extremly limited. Every curve in this video show no segmentation.
Lightstriker 5 years ago