Games Polygon count in the future?
polycounter lvl 9
I am following a project to see where will it go.
It's a mmo while they plan on creating the nerve gear from sword art online.
They plan on releasing the game in year 2022.
I can see some of the 3d models they are creating for their game as well as polygon count for them.
In my opinion the polygon count seeems too high,but sadly I cant find any info to confirm this.
They are using 50k for a sword.
So what do you think,what will be the polycount in that not so far future?
If you can find some information that would also be great to hear.
Btw,is there a way of working around the high poly count besides normal maps?
It's a mmo while they plan on creating the nerve gear from sword art online.
They plan on releasing the game in year 2022.
I can see some of the 3d models they are creating for their game as well as polygon count for them.
In my opinion the polygon count seeems too high,but sadly I cant find any info to confirm this.
They are using 50k for a sword.
So what do you think,what will be the polycount in that not so far future?
If you can find some information that would also be great to hear.
Btw,is there a way of working around the high poly count besides normal maps?
Replies
A decade or so ago we didn't even have normal maps.
As for other methods around the high poly counts aside from normal maps, there's also real displacement maps, as displacement and realtime tesselation are becoming more and more supported. There's also things like relief/parrallax + clip mapping, but that never seemed to catch on, and straight up higher polycounts+displacement seems like it's the likeliest route to me.
I suppose in ten years time we could also have returned to modeling *some* stuff with splines/curves again, and then we'd let the engine tessellate them at runtime.. that might be useful for very precise modeling requirements (accurate cars, perhaps)...but I can't imagine it would replace sculpting tools like zbrush.
managing and modifying geometry just becomes such a hassle when the meshes are very dense. i'd hope for new approaches to work on complex models beyond pulling vertices, really.
what we have right now requires effort enough to deal with and we'd better look at ways to structure production in more manageable ways first and foremost.
I am more like an observer of that project.
My proposition was that they work on nerve gear first cuz games can be done later easily.
But they are probably some kids and dont know much about any of this.
Can you explain more about how real dispacement maps work and relief/parrallax + clip mapping.
As you say the modeling could evolve in some way,but in case it doesnt we can go by how much polygons can our pcs handle.
With that in mind there might be a pattern like every 2-5 years 1000 - 2000 more polygons.(the increase would be applied to each part of the game such as characters,maps.objects...)
Even if we manage to calculate this it doesnt mean it's 100% accurate,still it would help me understand better so that I can somehow argument what I have to say when I respond to them.
Cuz they have some people following their project,so I am trying to "open their eyes"...both the developers and people that are following the project.
Edit:Thomeas@
We have zbrush.
Zbrush kind of changes the whole thing pulling vertices.
However we might figure out even better ways of modeling in the future.
Assuming an ideal coverage, an 1920*1080 image should not have more than about 520,000 polygons; but of course there's overdraw, you can usually find and drop back facing polygons, there's a lot of stuff drawn with transparency like foliage and such.
Then there's LOD, where you progressively swap out your highres meshes with less detailed ones as they get further away from the camera. Nowadays it's not uncommon for characters to have 5 to 10 discrete LOD versions in order to have an optimal triangle size for every case.
So, it doesn't make sense to have more than a few million triangles in a realtime scene and sacrifice pixel rendering efficiency. Still, the highest LOD version of a giant fantasy sword could remain to be relatively OK for close ups even at 50k polys, but these cases should be kinda rare in an MMO, especially if it's from a third person view.
Displacement mapping is even less "new". It's been around for decades in the pre-rendered world, but in real time it's more a recent thing due to the advent of GPU-based tessellation. Think of it like a combination of a tessellation and a push modifier on an entire model, masked with an image. All the geometry is added on the GPU at runtime. It's pretty limited since the pushing and pulling is surface-normal based (i.e., no doublebacking geometry), but in combination with higher res ingame meshes in the coming years you could have something pretty powerful without much change in the way you author the base content, since this would just be another baked map from a sculpt.
Personally, I don't see either of these methods being standardized over raw polygon usage within the next console generation--I'd advise spending more time learning about new shader and lighting technology rather than fussing over geometry tricks.
Here is the link of my conversation about it with some of the guys from the team
http://www.moddb.com/members/lucarioman555/images/anneal-blade-new-hilt-wip
(dont comment there xD)
I am trying to explain,but how to explain something that cant be argumented cuz
it's something that may or may not happen in the future
Vargatom@
Nerve gear is going to be like real world vision - it's supposed to be.
Will the picture really be 1920 * 1080 or much much more?
I heard of lods before,but it still feels like a strech cuz mmos are supposed to be team play with nerve gear even more.(10 - 50 players on one place and a huge dungeon boss)
The code that would need to be executed to preform all the changes of lods and the size of the game would be enormous.
BlankSlateJoe@
That seems interesting.
Still cant figure out how much is that different from normal maps.
Now that I am thinking the coliders would play a big role.
For example to create a feel that there is a hole there even if it's just a normal map would be to clear that area off the coliders.
The hand of a player would pass trough the 3d model making the impression that there is indeed a hole.
Cuz the whole point of being in a virtual world is feeling like you are there with your mind a body...
Ugh, seems like you're quite unfamiliar with a lot of standard knowledge in this industry. But first, what the hell is 'nerve gear'?
As for the resolution, MMOs in particular should be capable of running on average hardware to maximize subscription income. There aren't millions of gamers buying super high-end systems and 1080p or so already assumes a 22-24 inch screen.
Sure there are these new retina display things but they require utterly fast graphics, so once again no.
LODs are absolutely necessary to get good efficiency and thus good rendering performance. And they're not even the most coding intensive feature in a renderer to get good framerates, in fact I'd say this is one of the most simple things to do.
Nerve gear is a virtual reality device - a fictional device from anime called Sword Art Online.
Virtual reality Nerve Gear creates should be close to real world.
Virtual reality is the next step in gaming evolution.
The closest thing to that so far is occulous rift and morpheos from sony.
But nerve gear cuts the electric signals from your brain to your body to stop your movement while the 2 I mentioned only change your image while you are awake.
Yeah,I dont know much about industry cuz I am not a part of one.
But I have better understanding than an average person.
At least I think
I dont belive they will manage to make it,but in 20 - 50 years we will have something like that.
Still I wanna give them a chance so I wanted to explain to them some things.
Even if they fail in the end they would still have a solid game that could be played on a regular pcs.
Btw,I am not associated with them or helping them with anything...I am more concerned if they sort of scam people while not even know that they are doing it.
Like promising something and then not being able to fullfill the promise.(7 years saying they will create a nerve gear and then not)
Why I also want to help is because I love anime in general and Sword Art Online is quite good.
But I can be objective.
Edit:
blankslatejoe@
What do you mean?
This makes the assumption that MMOs are even still popular at all in 7 years or if mobile MMOs become the norm.
I figured that,but mmos that I played arent going that far back imo.
I have a weak graphic card intel hd 3000,but still can play most of mmos on medium settings.
I tried to explain to that guy that he is using too much polygons - practicly wasting polygons,but he said it's fine.
Before I even come accros that project I modeled a sword from that anime :
As the gaming evolves mmos and online gaming will be more and more popular.
Your Game Idea is too Big
I think it low balls the "existing IP" check mark, I actually looked into purchasing an obscure IP to make a remake and it was valued at $100,000. I doubt the MMO you mentioned will ever make it far enough to be a game but if it does, it's going to get shut down. Trust me on that last part, I've been giving that advice for 17 years.
As far as the "where will polycounts be in the future" I don't know. I think lighting & rendering is the new frontier. I don't think we'll be exporting straight from Zbrush into game engines any time soon, that's not even the case in Hollywood, is it?
Well,what I see which could be worth is nerve gear.
I dont see why would they bother start creating a game now.
If I remember correctly they said that they would only charge money for nerve gears,not for the game.
And lol,I tried to tick some of options on that website and the number of years they would need is 25
But I still doubt that xD
Anways...I tried going in with a different approach that you,but it seems they probably wont listen.
The only thing now is just a dream/wish for nerve gear to become reality.
Maybe sony does something in the next couple of years
I myself started a project 4 years ago to build my skills and it's sadly still incomplete.
But definetly in a better shape than their project
Personally I wish I got in during the 90's when people got huge bonuses.
This sentence belongs in a resume!
I know it may seem impossible now,but I am pretty sure in 50 years we will have some sort of a virtual reality device similar to nerve gear.
It's like those people that were alive when people started using electricity.
Most of the people thought it was impossible.
So I decided to keep an open mind.
I doubt this guys/kids will invent one,but still it will come sooner or later by someone else.
Virtual reality has many more uses than just video gaming.
Tidal Blast@
I agree that this lies on lots of ifs,but it's better than nothing.
Anways...my question was for 50k sword primarly.
I am a low poly modeler and usually model anime stuff.
So 50k for a sword is way too high for me and probably for a mmo too.
But when I said that to the guy with whom I had the disscussion he started talking something about compression and servers being able to handle it and the sword being one file...so I came here to ask about it...
Also yes,it's the greatest time to be alive.
I am born 1993.
Even though computers were invented before I was born I still feel like I and computer evolution are going together.
Before I die I would definetly like to see something like nerve gear.
And as for the project the reply was more or less "this is kid's play". What you make of that feedback is your responsibility (not the others). As people mentioned their choice of arguing for "future specs will solve it" is because it allows them to not really go deep in asset production, and use beginner's modeling techniques. Which fits the whole notion of this project, which is more or less let's do something without any restrictions, give us all cool titles... give it a name, a future release date. It's like kids playing and making things up.
Any person who as actually done serious work for a bit will tell you that working under "no restrictions" is never the case, there will never be the case where there is abundant power (working at NVIDIA I get a glimpse on a possible future a bit earlier). There will always be limits. Which is why getting dragged into project's like this does nothing good, if you want to be serious about the topic, you learn to work with limitations. You learn the proper craft... That will prepare you for the future.
There is no shortcut saying, hey let's work without limits, because in +x years the limits will be different, which means we are prepared.
It's not meant in a mean way by others to point this out.
Now you can either invest time into that and for whatever reason argue with them (someone's wrong on the internet), or do something else with your time
There were lots of good answers like you said,but there is always couple of those that are here just to put down people.
Anways...yeah,I was first one to question their way of doing things because I have some knowledge about it from my game development "voyage".
I guess I am a bit biased in a way that I still have hope in having something like nerve gear in my lifetime because that's almost like the transition before pc was invented and after pc was invented.(basily something big)
I was trying to get some answers on what could be the possible polycount future pcs could handle.
Cuz of Moore's law.
I work on a project with 1 coder and he said to me that these guys are kids and he turned them off.
So yes I understand most probably this project will fail.
So I wanted to reasearch a bit for the sake of "I want a game of Sword art online on a pc" or to at least open the eyes of other follovers to not donate any money or waste their time.
At the same time I kind of wanted to know more ways other than normal maps to have high detailed models with lower polygon count cuz more knowledge about game development will help my project.
in that case, learn what the current game technologies are, because this will be the basis for anything happening in the future. And if you care about nerve gear type of stuff, then you should look at medical and electronics research projects. E.g. the devices which allow paralyzed people to control computers with their brains and nerve impulses, or which allow blind people to see - I remember reading something recently how there's an implant stimulating the eye's nerves. Now that is pretty advanced stuff that sounds like sci fi! But that's a totally different track than games...
Point of my story is that just because you can, doesn't mean you should. You should always strive to be efficient.
I kind of know already some of the current game technologies and limitations.
But it's always interesting to hear more about it.
While I dont plan on creating the nerve gear,I would like it to exist while I am alive.
Even if that technology at first may be used for medical needs,I am pretty sure it would eventually lead it's way in to gaming.
"Allowing blind people to see" that's great!
Didnt hear about this one.
Edit:
skankerzero@
To be fair in some sense a bit more polygons can represent something better if that means you wont have to use normal maps to trick the player because most game developers actually strive towards reality.
As we go more and more in the future that is becoming more and more the case.
But I can agree with you that a box will always remain the same if we are talking about a box without any details on it.
But once you have a normal map, the underlying model's actual structure becomes less important, it only matters with silhouettes and self-occluding shapes. This is why there's research into using tessellation only on silhouette edges of an object, for example (I think it's used by Naughty Dog). So increasing the poly count will get you diminishing returns, or - if they get too small - a loss of GPU efficiency and low frame rates.
Offline CG of course doesn't really use normal maps, but the software renderers have very efficient methods to deal with small polygons so a denser mesh isn't going to ruin efficiency. But game engines running with GPUs are different.
No one, repeat, no one starts working on a game before the console is created. Also, never starts on a proj before you got an engine.
Those guys got nothing.
I don't even know if you see the game with Nerve Gear or not.
Maybe in 50 yrs, we could just plug our brain into the game. No need for display.
8 years from now? Why even work on art at this point? So they are going to go into development hibernation and chip away at a project for 8 years?
Damn.
Yeah, my point is that if future engines can handle 100 million triangles on screen, you shouldn't put 100 million triangles on screen.
Wasted triangles for the sake of upping the polycount is a terrible practice.
This is actually not true for tangent-space normalmaps. As the name suggests tangent space is "tangential" to the geometry, but that orientation must come from somewhere. The vectors required for tangent space with which the normalmap vectors are then "rotated" into proper orientation, are coming from vertex interpolation, and the normal vector is typically one of them.
Rasterization hardware will always favor bigger triangles, which are not "thin". Especially as you approach triangles that are smaller than a pixel it becomes quite wasteful. Because you pay a constant price for all vertices and primitive setup and so on.
If an object takes few pixels you normally "win" because there is less pixel shading work to be done. But once you have too many triangles/vertices, that win is neglected by those constant costs.
If I remember correctly they mentioned a custom engine called The Seed Engine.(they created it)
I doubt we would go from normal pcs to no display...
What you are talking about may come after vr devices.
I pretty much thought and think the same thing.
As well as many things people already mentioned.
I wonder if there is any forum thread or article somewhere talking about creating a vr that sounds promising
What I was talking about is rendering triangle meshes without normal maps. As far as I'm aware current GPUs can't really do per-pixel shading without them, so it'd look kinda ugly, even with very high polygon counts.
Also we had per-pixel shading before we had normalmaps.
There can indeed be a quality benefit to normalmaps which helps reduce aliasing artifacts via mipmapping and such. But fundamentally your hi-poly models look good in the viewport as well.
simply put: 50k for a sword? just cos you could... would you?
http://www.fotos-hochladen.net/uploads/workschutzpmqorbc51a.jpg
gotta be kidding me... and says theres no way to better do that? o.o
So you better have a damn good reason to be using 50,000.
Hehe,I agree with you.
If you check on the comments you will see that I suggested them to make models with less polygons.(my sword was 1500 - which is probably a bit much as well but I did it for showing off,not games)
I am just checking the project from time to time mostly to see 3D models cuz before everything,I am a 3D modeler.
But so far I didnt notice the huge polycount they are using
Mostly because I use 3ds max and they use blender.(they didnt talk about it)
Nice,I hope to be as good as 3D modeler who did this one
Every time I see some good 3D model I feel good lol
I also believe a game like SAO probably won't ever happen even in a decade. To make something so close to real life to a point where you actually believe it is real, including all senses (taste, touch, pain etc) .. No chance. Children can believe though. I know when I was younger I thought graphics in games would one day look like they do in movies and they almost do now!
Besides, most of the human brain hasn't been mapped yet, we're more likely to get a reality TV show of people living on Mars in 10 years time than we are of Nerve Gear.
Er, most of the time you bake your normals from a surface with proper curvature and usually with added detail. I'm not talking about cubes; think about cars in racing games. They require normal maps as well because the underlying mesh surface is still far too faceted to get a nice smooth shaded result. If you discard the normal map you'll have to work with the vertex normals and what the GPU interpolates along the triangle surface.
Granted, I'm not working in games - but I don't think that'd look nice. Last time I've read about it, around the first Geforce, all that was supported was Gouraud shading; I guess it should be possible to write your own in vertex shaders, but I doubt it could look as nice as 16-bit normal maps.
Offline renderers can of course get you nice results with a proper amount of subdivision but that means plenty of sub-pixel triangles. But supersampling AA can easily deal with that.
Yeah but it was also fed by textures and the final shading was done in the pixel shaders. Like EMBM and such.
Standard Maya viewport gives me ugly artifacts unless I turn on Smooth Mesh. Our head models are in the 25-50k quads range and they still need that to look good.
But again, I'm not as familiar with raw shading capabilities of GPUs nowadays. So feel free to correct me here. I mean how does a sphere look like with and without normal maps? I believe that there should be a huge difference.
It pretty much depends on the game. It looks to me that some rely on raw geometry for the car body (probably with some smart vertex normals editing), like Forza or Project Cars ; while others might use actual normalmaps.
@JacqueChoi I can see a few edges standing out that could easily use more geometry.
[ame]
Smart guy. You should listen to him.
I did a bit of browsing on that project and it's going to fail. From what I could tell it sounds like a bunch of pre-20 (some of them have posted their age) fans of an anime that don't understand reality yet. That's not a bad thing as I myself have been there before. It's just that there are a lot of things that are wrong with the project.
They seem to focus a lot on the programming and software and no updates on the actual hardware. That would be like programming an ios app in 1999. Why the strong focus on the software? The project leader seems to shrug off the advanced hardware tech with a 'I know a guy' and 'It will work like in the anime'. :poly142:
One major thing they probably don't realize is that the Nerve Gear will be classified as a medical device. This means lots of money and time in development and approval costs. Not to mention it could very well be banned on a consumer level.
One common thing this project has with so many other game projects is there's no payment to the developers. It's a guy running a project of 'slave' labor. This is fine if they're really doing it because they're passionate about the project, but these type of projects fizzle out all the time. That's just from a software perspective. This is a major hardware and tech project, way more than it is a software project. It's practically impossible to achieve tech and hardware without money.
Another problem is they're doing it as a MMO. Most people who run down this path trip over their own feet, even some of the experienced game developers. MMO aren't just massive online games, they're massive projects that take a lot of resources to develop and maintain. They're also limiting it to 10,000 people and 1 server. That's so far from the idea of what a MMO is that would be like making Pipe Mania and calling it a Mario game.
Secret World, DC Online, and Rift all cost more than $50 million. Destiny cost well over $100 million. Star Wars Old Republic about $200 million. Everquest was about $3 million but that was back in 1999. Yeah those are all high end titles but you're talking a high end goal here. 10,000 people at $2000 a pop (assuming no special sales) is only 20 million. You really think a hardware/software project with 'hyper real' (a term I really hate) virtual reality is going to costs less to make? If you think QA is time/resource consuming now try adding touch and smell into the mix. What happens when your game literally stinks? Editing video and sound can be tedious what about editing smells?
This is a far fetched fantasy project based on a science fiction anime. The creators are too focused on creating the content. If they really want to see this become a reality they need to get a proof of concept and some funding and keep throwing money at research and development of its hardware, not the software. Until that focus is changed this car is heading for dead man's curve without any brakes.
:poly142::poly142::poly142::poly142::poly142:
If the IP is so good why not put your passion and effort into making your own game for it? Talk to the creator and get permission and start a project that isn't a unicorn farting rainbows.
An 'artist' that just throws more polygons to create more 'detail' is a terrible modeller. Doesn't matter how many polygons an engine or rig can handle. It's a very bad habit to have that mindset.
Thanks for taking the time and actually look a bit in to their site.
I totally agree with you on what you have to say.
But I see over 1000 followers of the project so I wanted to get answers that would prove to me and everyone that they are actually serious about this,or that they arent so that those 1000+ followers dont waste time on something like this.
I actually do have a project going on,but I dont have a team because it's a free project so you cant keep people if they dont want to.
While the coder is on a break because he has school and other stuff...
When I started the project the team consisted of 15+ members.
My project is also about an anime,but it doesnt involve anything that cant be done today.
The previous coder did a good job and coded a lot of stuff,but the last game version he sent me before he couldnt work anymore for some reason doesnt work correctly and the code itself is kind of a mess.
So new coder is supposed to write everything from scratch and optimize it for adding more content.
Anways....going too long about this...
I dont have that mindset.
I am doing low poly models all the time.(no triangles)
It's almost like I am stuck at that kind of modeling.(cant remember the last time I modeled a high poly model)
When I said that I meant something else.
For example.
Lets say you can represent a table legs with a 2D plane.
Would you do it?
That's fine.
But if you are going for strictly realistic (not game realistic) look you would use a cylinder for those legs.
(maybe name "autocad" might explain this a bit better)
That doesnt mean you would throw a 50 sided cylinder for the table leggs.
If vr age comes things like this will definetly be needed.