Search Unity

  1. Megacity Metro Demo now available. Download now.
    Dismiss Notice
  2. Unity support for visionOS is now available. Learn more in our blog post.
    Dismiss Notice

3D model polygon count... How much is optimal?

Discussion in 'General Discussion' started by MrDude, Jul 19, 2010.

Thread Status:
Not open for further replies.
  1. MrDude

    MrDude

    Joined:
    Sep 21, 2006
    Posts:
    2,569
    Hi all...

    Okay, where to begin...?

    Every now and then I come across a 3D model extracted from one or other commercial game and when I check out the source I notice polygon counts in the 13k to 30k for a single character model. Heck, one Volleyball game by a very reputable company I am not going to name spent over 10k just on the one girl's braided hair! Granted, the scene consists of a single static environment and only 4 characters, a net and a ball so I am sure they must have had extra processing power available but still... but still, one model I saw had 90k as each hair style and head type belonging to that particular model was contained within the same file... That one was just plain weird!

    With game studios using between 10k and 30k models for realtime games I get a tad confused when people say that 10k is pushing the limits for current gen real time applications. Certain people say 5k is decent enough, others say anything over 3k is too much while others argue that anything up to 7k is what is used in current commercial games like [your names here]... even though the 30k ones I mentioned above were on the Original XBox, so that was nearly 10 years ago last-gen games that used 30k models...

    So when someone speaks of "optimal" or "non excessive" polygon counts, what is that based on? A common denominator between average PC hardware specs or lowest-still-considered-modernish-PC hardware specs, target console specs, hearsay, guess, needs, artistic skill? What?

    I have decided, based on everything I have read thus far that 3-5k is a decent polygon count, but I have seen 10-30k used by people who know... Almost mentioned a name there... Huge franchise, available on the XBox, well known company, lots of titles behind their name, big in the gaming world... so I think they must have a clue what they are doing... So between my personal goal of 3-5k and the witnessed 10-30k, I have decided to be generous in my model count and allow for 5-7k as an "average polycount"

    ...but I am curious to know where the rest of the people get their figures from when they speak of next-gen and current-gen polygon limits. I sincerely doubt that anyone will speak about "next-gen graphics for [your favorite phone here]" so I am assuming they are speaking about consoles and far more powerful computers... so what does these generally accepted default values find base their value on?
     
  2. ChaosWWW

    ChaosWWW

    Joined:
    Nov 25, 2009
    Posts:
    470
    I think a lot of different people have different ideas about what is an optimal poly count. Really depends about how far you are going back in technology. However, most even semi-modern computers can render tons of polygons with hardly performance cost at all, or so I heard. Personally for my project, I'm going for around 10k for main characters and significantly less, like 2-4k for enemies, depending on how important they are. By the industry's standard, that is not even that much, but it works for me.
     
  3. equil

    equil

    Joined:
    Jan 18, 2010
    Posts:
    15
    they're based on old information. A lot of people are still sort of obsessed with the amount of triangles because it's easy to gauge the amount of them, but ultimately they're not a very good measuring stick for performance. It's just that triangle count and fps are the numbers that you can usually see, so they're probably the only tangible things people link to performance.

    For example the main character from Oni, a game released back in 2001, had more than 5000 triangles.

    what you should really look at is vertex count and batches, and this should shred some light on the subject.
    http://www.ericchadwick.com/examples/provost/byf1.html
     
    EvilGremlin likes this.
  4. MrDude

    MrDude

    Joined:
    Sep 21, 2006
    Posts:
    2,569
    Wow, that's a lot of reading but it sounds like should be an interesting read!

    Will have to get onto it... Cheers :)
     
  5. Vert

    Vert

    Joined:
    Mar 23, 2010
    Posts:
    1,099
    I personally have not done the complex math or tried to figure out how to perfectly optimize my detail by figuring out where everything will go when levels are finished as the article linked in this thread tells you to do. I am attempting the idea to use as few as possible. I figure the less polygons/vertex's there are the better. Also if I sacrifice them in one model I can add more models for added detail. I am not so concerned with super smooth edges or 10,000 polygon guns for first person view. I think that uniform detail is good and whatever I save I can use elsewhere. Basically my plan for polygon count is use as little as possible. Then when the level is complete, add as many detail meshes as possible without killing frame rate. I think a world with slightly less smooth objects but more of them is better than an almost empty level with a few high polygon models lying around. I am not sure how well it will work, but it allows me to not worry about if I can add extra polygons here and there for extra detail as I will just use as many as it takes and no more.
     
    WiVonZulul likes this.
  6. Alric

    Alric

    Joined:
    Feb 17, 2009
    Posts:
    331
    I tend to agree with you, but unfortunately hardware doesn't generally perform that way.. fewer objects will tend to outperform many objects dramatically when polycounts are equal. Of course there are ways to mitigate this but it's always there.
     
  7. hmacyt

    hmacyt

    Joined:
    Jul 1, 2010
    Posts:
    203
    Drake in drakes fortune 2, drake was 30k, main characters were 15-30k, basic enemies were 12-15k

    in gears of war Marcus was 15k

    Alyx in halflife 2 was 8k, back in 2004

    shepherd in mass effect is 20-25k depending on the weapons and armor.

    Current engines have no problems pushing tris, but are slowed down by advanced shaders and lighting effects. The UDK deals really well with instancing and modular pieces, unlike unity which deals well with everything with shared materials merged into one object.
     
    taxvi likes this.
  8. loken

    loken

    Joined:
    Mar 25, 2009
    Posts:
    109
    This largely depends on target platform and target audience.

    A more casual game will most likely attract people with laptops. And What's in laptops? Crappy Intel chipsets that very much do have performance limitations when it comes to number of triangles.

    That's what it comes down to for me anyways.
     
    Aligdev likes this.
  9. MattCarr

    MattCarr

    Joined:
    Jul 5, 2009
    Posts:
    339
    Check out these 2 articles:

    http://www.rsart.co.uk/2006/11/20/how-many-polygons-in-a-piece-of-string/

    http://www.rsart.co.uk/2007/08/27/yes-but-how-many-polygons/

    They should help with the understanding of "it depends on the project".

    There are of course other many things to consider aswell besides polygon counts such as number of bones in a skinned mesh. It really comes down to what the platform and audience is and will require some initial benchmarking to help determine budgets. Also, LODing can be massively important not just with mesh complexity, but LODing character rigs aswell can be invaluable if you have lots of animating characters in your scenes.

    There is so much smoke and mirrors optimisation going on in all games from the low end to the big AAA releases that people don't realise. Generally speaking at the end of the day the core elements of different engines aren't that different, just some will have more optimisation stuff going on than others. The best thing to do is figure out what your project needs in terms of art content and quality and then build a framework to handle these elements in a fast way. Once you have the optimisation framework (which may include code/tools for mesh/rig/material/animation/AI/whatever LODing for example) complete, you'll be able to start testing the #s and find the limits.
     
  10. Vert

    Vert

    Joined:
    Mar 23, 2010
    Posts:
    1,099
    I can see why, but I think if I use enough duplicates of objects a properly optimized game engine such as Unity can deal with the duplicates well via the instantiation and use of prefabs. Or am I wrong that prefabs do not give that much performance gain? I would rather have 10 of the same lower polygon barrels all around than only 4 high polygon ones. Or would that still have performance suffering? I am still learning how to optimize art for game engines. My past 7 years of hobby design have all be theory research and learning just how games are made. :) So clearing this up would be very helpful. Perhaps I haven't done enough research in the art department yet.

    Edit: Well if I am wrong at least my game will be ready to be ported to the netbooks or smartphones of the future as I should have just about the right art for the job :wink:
     
  11. dogzerx2

    dogzerx2

    Joined:
    Dec 27, 2009
    Posts:
    3,964
    Good question. I must say, I have no idea what's the best number of polys. With today's PCs and consoles... 6k.. 10k maybe? Maybe the important thing is to make sure your models look at their best. Mesh is needed for two things as far as I'm concerned, silhouette and shading. Although with normal map shading looks awesome even in low poly meshes.

    I think your standard should be something you find out for yourself depending on your needs. For next gen PCs you can push it quite a lot, graphic cards can handle their shyt, maybe the concern is the cpu, when you start making use of physics and other stuff, so it depends on what role will your models will take. Also the amount of characters on screen needs to be taken into account.
     
  12. MrDude

    MrDude

    Joined:
    Sep 21, 2006
    Posts:
    2,569
    This brings to mind another question... I thought I remembered reading somewhere a few years back... not sure where... but if I recall correctly, using prefabs in Unity somehow gives you a huge performance gain in terms of if you use a prefab of a crate model and place 20 of them in the scene, the game will run faster than if you just placed the same crate in the scene 20 times... apparently the other instances require no overhead or something... I am speaking under correction here but I could have sworn that was what I read....

    Am I remembering this correctly and why would this be the case when both methods use the same mesh. For instance, if this were true then I could make a 50k bad guy and place 10 of them in the scene with no worries of any kind...
     
  13. TwiiK

    TwiiK

    Joined:
    Oct 23, 2007
    Posts:
    1,729
    That seems like something you could easily test yourself, doesn't it?

    I for one, without having seen that article or read something that contradicts it, would say that is bullcrap. :) 20 instances of a prefab or 20 copies of an object all increase the drawcalls, tricount etc. the same so if there is a performance gain there could be a tiny one related to how Unity manages those objects in memory.

    But I would think a noticable performance gain would be written about in the documentation, perhaps under prefabs, and it's not.

    If you made a 50k bad guy and placed 10 of him in the scene you would have 10 times the draw calls and 500k worth of bad guys no matter if you instanced him from prefabs or copied him in the scene. :)
     
  14. bigkahuna

    bigkahuna

    Joined:
    Apr 30, 2006
    Posts:
    5,434
  15. MrDude

    MrDude

    Joined:
    Sep 21, 2006
    Posts:
    2,569
    now why the hell didn't I think of that?! :p

    I agree with everything you just said. I thought it was too good to be true so I never really paid it much mind. Suppose that was why I never bothered to test it...
     
  16. Vert

    Vert

    Joined:
    Mar 23, 2010
    Posts:
    1,099
    I think the performance gain with prefabs is that they all reference a base set of data such as its mesh and texture. Then instead of loading 20 instances of a crate, it loads one and then notes where each is supposed to go. I think its designed to stop texture and object duplication in memory.

    Reading this thread is making me realize I can get more generous with my polygons in my model. I think I will. Good thing I haven't made many models yet.
     
  17. MrDude

    MrDude

    Joined:
    Sep 21, 2006
    Posts:
    2,569
    What a contradictory ambiguous statement in my case... I don't have a lot of models, yet, and that is no good thing at all as the code is 85% done... but thank goodness as I haven't settled on a final theme, yet... :p

    This issue actually came to mind after reading yet another: "Your model looks good but you are using too many polygons here. Try and use more there and less there"-thread. It then got me wondering...

    Daz has some really decent models that you can make by taking the same figure and just adding different clothes to it. Now Daz models look good because they use millions of polygons so how can it NOT look good? Then they made the decimator and showed this model as 90 000 polygons. Now I am not daft enough to believe that ANY game would use 90 000 even for a lead character. But they claim they have had decent results with up to 4000 polys. So great... start with 2 million and break it down to 4000... What a workflow...

    The alternative is to go to a place that takes a base mesh, allows you to modify it by controlling certain parts of the shape and then exporting a custom, rigged model for dirt cheap and at a low polygon count of like 5k... but they all look basically the same. The face can be tweaked to a large extent but for the rest it is just a variety of bulky or not and like 3 clothing types. The rest is all done via textures. Boring as hell.

    Another alternative is MakeHuman. Has a low poly model that doesn't look very good from what I've seen and a quality model that is just to high in polycount to be of any use... Same deal as Daz in that you start from a base model and modify it to your tastes except they do not offer the clothes that makes Daz as useful as it is. Being free has it's advantages but the fact that you have to choose between low poly really low poly LOOKING model or print quality rather spoils it for me in terms of choice.

    So then there is 3 more alternatives:
    - Buy a ready-made model from somewhere like TSquid
    - Have a model made for you
    - Build your own.

    In the three cases above you again have a lot of freedom in terms of how many polys you are going to end up with. (Either don't buy it if it has too much, tell the artist this is how many you want, or just make sure you stick to your own limit, respectively) so then I got to thinking... With all the "No, it's too much", how many would be "too much" and how many would be decent?

    I figured that the PS3 can handle loads but PCs are more powerful than any console. So clearly huge polygon counts are no issue to worry about. From my readings of the Unity docs it seems something to keep in mind is the number of textures/ materials associated per model (to my mind each texture requires a complete redraw of the model in the scene complete with re-lighting the entire model) and the same goes for how many pieces make up your model. Anything more than 1 piece causing the same effect as above...

    So to my mind, take the maximum you are prepared to spend to make it look good (for sake of an example, say 50 000k), then divide the polygons by (the number of textures + the number of parts the model comes in) so if the head is separate from the body and I use 3 textures then use 10k for the model. Add in a bumpmap and a specular and I am down to like 3.5k.

    This is how my mind works... but then I didn't account for the number of bones... I have seen some rigs "made for games" include a bone for clavicle, shoulder, biceps, elbow, forearm, wrist... huh!?! Others include only clavicle, shoulder, elbow, wrist but then they have upper lip, lower lip, lips left corner, lips right corner, jaw, left cheek, right cheek... for a 3rd person game like Lara... Others have chest, shoulders, neck, head.

    Now, obviously, the less the faster (not necessarily better, just faster) but how many bones is enough and how many is too many and how many bones would dictate that you start cutting down on your polygons? Do bones really make THAT much of an impact?

    And then there is the other issue regarding textures. Staying with power of two images, each time you double the dimensions, you quadruple the required ram. So if a 512x512 image is not good enough, using a 1024x1024 image will require 4 times the RAM but using 2 textures of 512x512 will use half of that but require twice the processing power to draw. So waste the precious limited RAM or add more pressure to the overworked CPU? Which is best in the long run?

    I would consider the key to making a good model is:
    - Decide what your model will be doing and from there determine the range of movement required.
    - Based on the intended actions of the character, decide upon what bones are required.
    - Use as few textures as you can get away with but try to stick to 1 or max 2 512x512 images for lead characters and try to use smaller ones for less important chars.
    - Now that you have your functionality out of the way, use my previous formulae to calculate the polygon count.

    The only problem with this approach is "what do you base the original polygon count on? why 50k? Why not 5k or 500k?". I reckon, with the exception of hand held devices, current gen machines can handle pretty much anything you throw at them unless you intend to break it, of course...

    ...but at the end of the day you simply CANNOT say: "This is the best polygon count and this is what you must use for everything". Build a Smarties game with the lead Smartie being 20k in size or building the next gen army shooter with each character being no more than 3k... it just won't work.

    Knowing that, I suppose the question I am asking is not so much "How much is enough?" or "What is a decent number?" it is more along the lines of "When does one start to border on going overboard?" and how much do the other factors like textures and bones etc really influence the game's performance?
     
  18. Detocroix

    Detocroix

    Joined:
    Jul 2, 2010
    Posts:
    75
    Interesting discussion :)

    In my humble opinion, most important thing is the concentration of polygons. Do you need 5000 tris on that sword hilt while blade is done with 10? Are those eyes round enough with 250 tri, or do you really need those veins modelled in.

    If you just aim to use as little as possible to achieve the lovely look, and then look at the wirecage and there's no 'blackened areas' (areas where there's so many tiny triangles the wireframe itself fills more area than the tris behind it), you probably have good amount. Places where your eyes concentrate on (face and gear) are excused for higher poly counts, but shoes, legs, arms and chest generally do better with good normal mapping.

    Also when using textures, as long as you don't waste space you're good to go.

    One very good tip and voice of reason is to display your char on screen as close as you're most likely going to see it, then stare from LITTLE bit closer than normally, and slowly scale the texture smaller and when you can see blur and general trashiness, you've gone one step too close :)

    Just use power of two textures ALWAYS, generally game engines will mess up even very high detail textures.
     
    Aligdev likes this.
  19. SanderAKALego

    SanderAKALego

    Joined:
    Jul 17, 2010
    Posts:
    32
    3-7k would probably be okay, but iv heard that the character (that one u play) in Crysis 2 is over 50k ;o
     
  20. WillBellJr

    WillBellJr

    Joined:
    Apr 10, 2009
    Posts:
    394
    I believe you really have to look at your target platform(s) before you can make an educated decision.

    If I were targeting a phone device, I'd definitely limit my polygon count at the 3-7k range max.

    Depending still on how many objects I plan on having on screen at one time.

    Since my platform is strictly PC, I'll risk going into the 10K+ region for my main character objects.

    Unity netting me the web player for free, if I were to decide to release a web version, I may shave my limit down to 8K or so polys...

    Until you can actually test your final levels on your target platforms, you're getting ahead of yourself anyway.

    Other factors such as AI, physics, audio, multi-play infrastructure, and other CPU cycle stealers may force you to lower your poly counts even more, especially once it's all done and running (or crawling) on screen.

    -Will
     
  21. unitrix

    unitrix

    Joined:
    Mar 19, 2010
    Posts:
    279
    Lower poly count is always better and you will be amazed at how low you can go with polys before loosing quality. Always aim to the lowest amount of polys you can get away with, simple as that. Your time is better spent on texturing because this is what makes or breaks your game. A well textured character with bump mapping and specularity on a low poly model will look awesome as aposed to a high poly model with 1 texture.
     
  22. WinningGuy

    WinningGuy

    Joined:
    Aug 10, 2009
    Posts:
    884
    Lower polygon count isn't ALWAYS better. Depending on your graphics card, once you get lower than the amount of polys rendered in a single pass, then going lower than that is pretty much useless.
     
  23. unitrix

    unitrix

    Joined:
    Mar 19, 2010
    Posts:
    279
    Going lower will let you target a wider range of graphics cards no?
     
  24. MrDude

    MrDude

    Joined:
    Sep 21, 2006
    Posts:
    2,569
    This has been a most interesting read. Something that got my attention, though:
    I read that on Unity one should aim to have 1 texture for 1 mesh as "there really is no reason to have more" since it only slows down the rendering of that model...

    Wait, I think I have that wrong.... 1 MATERIAL per mesh. I used to only work with normal diffuse texture so to me 1 texture and 1 material was exactly the same thing so I wanted to ask about "if you are supposed to have 1 texture for one mesh as more than one just unnecessarily slows down rendering, then what about normal maps and specularity maps" as that surely takes 3 textures then!

    But now I can still ask that same question... say you do do a low polygon character with 3 decent texture maps... each with a bump and a specular... that's 9 textures for a single character... surely that can't be a good thing???

    BUT!!! Last week I found this link on this very forum to a site that offers free 3D character models. When I went there to have a look I found it was all illegal models. All commercial models like Resident Evil, Tomb Raider, Harry potter... I have noidea how one site got the source of so many different commercial games but yeah, there it is... So the models cannot be used but I still downloaded them to see how they do their models and I was real surprised by what I found...

    This entire time I was under the impression that I should limit my models to using as few textures as possible... I was also under the impression that when you create a normal map, each face should have it's own space in the texture file, no overlapping allowed... So for the Lara Croft model they actuall only texture half the face and then duplicate it for the other side... I never knew you could do that with normal maps!!! But tat wasn't the interesting part... the interesting part is that (for some character in some obscured game I don't know of) they used 8 textures and 8 normal maps...

    Surely they are amateurs and down know what they are doing... Then I get to this other game and they use 25... Another uses 10.... Lara uses 8 textures and 5 bumps, this angelic character uses 11 textures with no bumps and the two Resident Evil's characters combined with their respective weapons use up 83 textures...

    That in the world is up with that?!?!?
     
  25. unitrix

    unitrix

    Joined:
    Mar 19, 2010
    Posts:
    279
    Sorry I may have made a mistake when I said texture, I may have meant material or shader. My point was that it isnt the polys that make it good looking but the texturing and shaders and all that good stuff that make it look good.

    Edit: They may use a program for capturing it like dxripper.
    The program itself may have code that divides the meshes up and gives them more textures, it may not be a exactly what the studio uses for the game.
     
  26. unitrix

    unitrix

    Joined:
    Mar 19, 2010
    Posts:
    279
    I thought I would give you some poly counts of models in top mmorpg games just for your to see and compare if you are interested:
    I have all of these models on my computer or did at one point:
    World of Warcraft Human Model no gear just the model and hair has just over 1000 tris.
    World of Warcraft Human Model full gear with weapons and armor, the hole shebang, is just over 6000 tris at most.
    Everquest Original Models no weapons or gear just the model had about 700 tris
    Everquest Upgraded Models had about 2000 no gear no weapons
    Everquest 2 Models no gear just starting character has about 7000 tris

    A hidden gem for finding models is google warehouse
    A quick search there like this:
    http://sketchup.google.com/3dwarehouse/search?q=world+of+warcraft&styp=m
    you can find all the starting world of warcraft characters and some with models. Or you can search for Halo models and you will find them from all 3 halo games.
    Its really good for seeing how many polys are in models and what not. Also really good place to find models to use depending on their licence.

    Type in google : google models
    Warehouse is the first link
     
  27. WinningGuy

    WinningGuy

    Joined:
    Aug 10, 2009
    Posts:
    884
    Not by itself. But if you're targeting anything newer than 2005-06, going under 200 polygons is enough for anything.
     
  28. WinningGuy

    WinningGuy

    Joined:
    Aug 10, 2009
    Posts:
    884
  29. unitrix

    unitrix

    Joined:
    Mar 19, 2010
    Posts:
    279
    I was refering to quality nex gen models while still targetting more/lower graphics cards. Go as low as possible while optimizing your model untill quality starts to suffer. In my opinion a naked human character around 2000 tri is sufficient enough for nex gen.

    Then again is also depends on the game, for example if you were making a nex gen diablo type game where the camera is further out the polys may even be lower, if you are making a fps where the polys are right up in your face then they may be a little higher, however I still believe that optimizing models to as low as you can go without loosing quality is best. Think of how much optimizing big studios do, I heard they do insane amounts of it.

    You also goto remember your monitor is only so big, yes while there are bigger monitors coming into the picture, right now they arnt that big and the human eye wont detect anything more then around more then 3000 polys for naked human in game. With the textures and all the other stuff going on you arnt really going to notice strait edges, especially when you take into account the animations, the character is always moving. He is not standing still where he gives you a chance to point out that strait line or this strait line.

    I would say for a fully modelled with armor, weapons and the hole shebang I would try to keep it under 10k tris for a character.

    EDIT:
    Also depends how clever and skilled you are when comes to modelling. A good modeller can make amazing looking models with low polys.

    EDIT: Another thought, it may not be that what makes a nex gen game a nex gen game is how high one models poly is but how many models that look high poly can be on screen at one time? Thus going lower poly is still good and dont go to extremes of having 30 k poly models, if you want to have good performance and have lots of characters on screen then go lower poly.
     
  30. Vert

    Vert

    Joined:
    Mar 23, 2010
    Posts:
    1,099
    It's not about high polygon count, its about making a game look good. I am modeling with that idea in mind. Use just enough to make it look good. Anything else is extra and can be used somewhere else in the scene where it will be needed on some other model. All else it seems that this idea will keep polygons low that I could add more graphically demanding stuff such as more objects etc. I know some have said that polygon count and performance aren't tied much anymore as polygons are cheap but object count is not.

    However, with the new stream processor design in graphics cards can't I relocate my graphics resources elsewhere without an impact? From my understanding that was the big thing with the stream processor design is that if I need more power/memory for a task I can have it if its available. I remember an example saying that a scene might have a few thousand particles but then you move to a room with no particles and a bunch of high polygon models and the GPU handles it all fine as it can dynamically allocate its stream processors and memory to accommodate what is needed. This is of course with the idea that I am tailoring my game to run on mainstream cards. Do Intel cards dynamically handle this yet? I know this kind of tech was new about 3-4 years ago with the Geforce 8 series.
     
  31. Artimese

    Artimese

    Joined:
    Nov 22, 2009
    Posts:
    794
    its a big topic to discuss imo, but for me id never go over anything 14k (for a main character atleast)
     
  32. scarletsnake

    scarletsnake

    Joined:
    Nov 27, 2009
    Posts:
    106
    A pretty late entry but better late than never;

    Up to recently I had around 1200 polys for my main character's head, hair and eyes. While it certainly doesn't look that different from a distance, when the camera got close you could really count those polygons around the edges, and that certainly wasn't something I wanted. Add to that, when entering scenes with multiple light sources, no matter how good the bump map was it always looked crude. So now I'm using Z-Brush data on characters, and I'm a bit of a tidy person, I got a 5 year old rig that still runs today's games on high settings above 40-60fps; I got a really high FPS on my project, now I can sacrifice some of that unnecessary speed to have better looking characters. For example, my avatar is the old, low poly version. With the models I'm using now I'm amazed at what I was missing out in sake of performance.

    Please keep this topic alive, people, it's pretty informative and, in my opinion, needs to be updated with experiences.
     
  33. jjsuperspy

    jjsuperspy

    Joined:
    Oct 11, 2011
    Posts:
    71
    to me a high-poly model to me is were it has all the detail such as detail in skin or each thread in a piece of jeans and a low-poly is just the basic shape of a character or object but it all depends on what the poly count for high and low is and how much a game engine such as Unity or UDK can handle
     
  34. janpec

    janpec

    Joined:
    Jul 16, 2010
    Posts:
    3,520
    As long as you are using LODs there isnt really matter how many polys you use for FPS game. You can have very detailed 3k+ for face model if you have very close view distance. It depends what is your target platform and specifc game settings.
     
  35. DigitalCandy

    DigitalCandy

    Joined:
    Jul 16, 2014
    Posts:
    5
    Old thread, but some bad advice.

    Use as many poly's as you need to create the silhouette you need, no more, no less. If you're modelling a bullet shell, the player will only see standing up, they won't notice slight angle variations in the model, and those lines/faces can be excluded completely. Any extra faces you wouldn't notice in the game world, should be left out, period. Everything else can be handled with textures/normal.

    That said, yes.. poorly optimized models (and that's what models with unnecessary faces are), aren't THAT big a deal on modern hardware. If you go over your poly budget by 10%, don't stress about it.
     
  36. AGameTester

    AGameTester

    Joined:
    Mar 3, 2018
    Posts:
    1
    I have the same problem. I was trying to import a model into vrchat but, it said the avatar had 23,632 polygons. When I saw this, I was disappointed that it failed. I just need to know how to decrease the polygons.;)
     
  37. Ryiah

    Ryiah

    Joined:
    Oct 11, 2012
    Posts:
    20,950
  38. XenomorphGames

    XenomorphGames

    Joined:
    Feb 14, 2018
    Posts:
    4
     
  39. XenomorphGames

    XenomorphGames

    Joined:
    Feb 14, 2018
    Posts:
    4
    maybe you could use decimitate to reduce its polies i created a scene with 9 million polys and runs fine
     
  40. warthos3399

    warthos3399

    Joined:
    May 11, 2019
    Posts:
    1,726
    None of this really matters TBH. All big studio teams use a custom in-house engine, and its Nothing like Unity, and you dont know what the systems, and tech they have...
     
  41. Antypodish

    Antypodish

    Joined:
    Apr 29, 2014
    Posts:
    10,753
    It is not true. At least not anymore.
    While they got multiple own in house engines, they do use Unity, Unreal and other engines too.
     
    warthos3399 likes this.
  42. sxa

    sxa

    Joined:
    Aug 8, 2014
    Posts:
    741

    Well, true, except where they use Unity, of course.

    https://en.wikipedia.org/wiki/Hearthstone

    (And yay, for what looks like the 5th or 6th necro of a 2010 thread...)
     
    warthos3399 and MadeFromPolygons like this.
  43. MadeFromPolygons

    MadeFromPolygons

    Joined:
    Oct 5, 2013
    Posts:
    3,967
    maybe dont pointlessly necro threads?
     
    xVergilx and UnityMaru like this.
Thread Status:
Not open for further replies.