Poly Count Guide? - GameCreators Forum

Author

Message

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 1st Aug 2009 07:23

Link

Is there some form of guide/can someone make me a guide that says what the poly count should be for models such as: a human character, a gun, a knife, a tree, a house, etc. And if it helps I mean like for programs like GDK or DBP.

Back to top

Profile PM Email

Azunaki

15

Years of Service

User Offline

Joined: 11th Feb 2009

Location:

Posted: 1st Aug 2009 12:26 Edited at: 1st Aug 2009 12:37

Link

i think med poly is what most systems can handle in those programs.

a character would depend on what your aiming for in detail but about 1000-1500 polys is what i aim for. (hands,head, and feet are the killer in polys)

a gun, again depends on what gun that is rather vague (also i aim away from them....)

no idea for a knife.

a tree i would think about 8 diameter...4-5 high... 6-8diameter for each branch... 4-5 length.. 40 for the trunk plus 24 for each branch(if you use 6 width 40 if you use 8) so with about 5 branches you would be looking at about 160.

a house is rather vague... and varies greatly but i think you could manage a fairly good one with 250-300 polys. give or take 50...

these are rather vague areas. and vary from object to object but that should help you a bit.

Back to top

Profile PM Website

Sid Sinister

19

Years of Service

User Offline

Joined: 10th Jul 2005

Location:

Posted: 1st Aug 2009 18:51

Link

Don't worry about it. Just make your character. Next-Gen characters can get up to 10-15,000.

Two brilliant 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/

Quote: "
GOW - XBOX 360
Wretch – 10,000 polygons with diffuse, specular and normal maps
Boomer – 11,000 polygons with diffuse, specular and normal maps
Marcus – 15,000 polygons with diffuse, specular and normal maps"

"If I have seen a little further it is by standing on the shoulders of Giants" - Isaac Newton
Current Project: http://strewnfield.wordpress.com/ (Last updated 06/11/09)

Back to top

Profile PM Email

That1Smart Guy

15

Years of Service

User Offline

Joined: 26th Feb 2009

Location: Somewhere...... yep

Posted: 1st Aug 2009 19:04

Link

Quote: "GOW - XBOX 360
Wretch – 10,000 polygons with diffuse, specular and normal maps
Boomer – 11,000 polygons with diffuse, specular and normal maps
Marcus – 15,000 polygons with diffuse, specular and normal maps"

wow I play that game and I didnt know they were that high poly

Back to top

Profile PM Email

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 1st Aug 2009 19:20 Edited at: 1st Aug 2009 19:20

Link

Quote: "Gears of War, Xbox 360, 2006 (according to D’Artiste book)
Wretch – 10,000 polygons with diffuse, specular and normal maps
Boomer – 11,000 polygons with diffuse, specular and normal maps
Marcus – 15,000 polygons with diffuse, specular and normal maps"

DANG! How is that possible? Is it the game engine, or the processing power of a 360? Or is it the programming language (if used)?

Quote: "Resident Evil 4, Gamecube, 2005
Leon – 10,000 polygons"

I didn't know that a gamecube had strong processing :o.

Quote: "Metal Gear Solid 3: Snake Eater, PS2, 2005
Snake – 4,000 polygons"

Didn't doubt that. Honestly some of this news is shocking to me... How can one acheive these high poly heights? If they programmed in c++ directly and had it sent off to play for a 360 or ps3? Or does optimazation play a big role in this?

Back to top

Profile PM Email

Jaeg

18

Years of Service

User Offline

Joined: 16th Mar 2006

Location: Indiana

Posted: 1st Aug 2009 19:27

Link

A Gamecube has pretty good processing power. It is more powerful than the PS2 I think. If only it wasn't for Nintendo's family friendly game restrictions it would have had quite a few more impressive games for it.

If you get mad and want to type something nasty about another person do this-Type what you want to say in the box then press ctrl-a and hit delete then type what you should say.

Back to top

Profile PM Email Website

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 1st Aug 2009 19:44 Edited at: 1st Aug 2009 20:09

Link

Quote: "If only it wasn't for Nintendo's family friendly game restrictions it would have had quite a few more impressive games for it."

True... honestly the gamecube was pretty good, but then the wii became not so well... Anyways, is it possible to acheive this power? I think with DBP X10 it is possible, or at least good enough, because if FPS Creator X10 was made in DBP X10, and FPSC X10 can support these modern day graphics, then shouldn't DBP X10 be good enough, granted it does even come out any time in this lifetime? EDIT: Found an even bigger one

Quote: "Project Gotham Racing 3, Xbox 360, 2006
Vehicles – 80,000-100,000 polygons"

That's just not right...

Back to top

Profile PM Email

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 1st Aug 2009 20:25 Edited at: 1st Aug 2009 20:31

Link

I searched up some more about polycounts and got this link for a mgs thing: http://www.softimage.com/products/xsi/customer_stories/metal_gear_4/default.aspx an image:

a quote:

Quote: "The 512 size was not exceptional for normal maps either. The team imported into XSI a face model to which details of about 500,000 polygons had been added with ZBrush. From this, they generated a normal map with Ultimapper. Hideki Sasaki said that to obtain the maximum effect at a low resolution, it was essential to raise the quality of high-res models. He said that if work is performed carefully using XSI's modeling function and subdivision surface, it should be possible to achieve sufficient quality even at low resolution."

and a movie of the mouth rigging: http://www.softimage.jp/user_case/mgs4/Flash/snake_face/snake_face.html I'm still a little shocked.
back to the question: for DBP or GDK, or even C++ or DBP X10 when and if it comes out, what would be the poly count for a high quality human model? And for that matter, considering the poly count, would DBP, GDK, DBP X10, or C++, be considered the graphics of what current day gaming system? Like I read that Snake from MGS3 was 4000 polygons, would that mean if DBP can reach 4000 polies for 1 character, then DBP could make a game with the graphics of a ps2 game? Its hard to explain, but this way I can get a better grasp at what type of graphics I'm aiming for, as in making 3D models, and then programming them. EDIT: hey, MGS4 was made in windows ^^

Back to top

Profile PM Email

Sid Sinister

19

Years of Service

User Offline

Joined: 10th Jul 2005

Location:

Posted: 2nd Aug 2009 01:11 Edited at: 2nd Aug 2009 01:32

Link

SJH: I don't think there's a clear answer to what your asking, but to be honest, I'm not exactly sure on the relation that game engine's have on the maximum polycount. As far as I know, the bottleneck is the hardware because the engine just needs to bring everything together. I don't see why you can't plop a 100k scene into an engine that was designed for a 50k scene at the time because of the technology at the time. The engine is suppose to control the overhead and manage assets, so I suppose if it had difficulty doing that fast and effectively you'd start seeing some issues. This is all just speculation though. You bring up an interesting question. Let me see if I can get this answered on Polycount and get back to you.

If anything, I think language would play the biggest part in the engine's speed. That's not to say that there is a definitive polygon limit though. I think it really depends on if you have a complex engine or a simple engine. You could probably make a Pong clone with 100k poly's in DBP, but if you tried doing an RPG where the engine is managing a lot of data on top of the graphics, then I think you'd start to see the polycount come down.

Again, I could be way off. I'm gonna go ask this on Polycount.

Also, I think you guys have been scared into a corner when it comes to making something "low poly." Seriously, it's okay, you can lighten up! You should really learn how to do High Poly art first anyway, especially to break the fear you have. After you make it high poly, work on making it low and bake the detail in.

REPLY FROM POLYCOUNT

Quote: "
I think the issue of polycount is far less important that texture resolution, shaders, draw calls, ect

Polycount is one of the simplest things a game engine has to do. It's all just basic math so if your computer is fast enough to handle all the points then its going to render fine.

There are limits in game engines that define how big meshes can be, but these limits are usually just put in place because some kind of limit has to exist and are often times arbitrary.

I don't know any specific limits off the top of my head, but really I think it's more a question of your target system(s) power than what engine you are using when it comes to a models polycount."

"If I have seen a little further it is by standing on the shoulders of Giants" - Isaac Newton
Current Project: http://strewnfield.wordpress.com/ (Last updated 06/11/09)

Back to top

Profile PM Email

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 2nd Aug 2009 01:39

Link

So pretty much the only thing influencing the poly count is the console itself? At one point I was convinced that DBP And GDK were slower compared to things like C++, or was I wrong? And Sinister, thank you for your interest in this topic, it helps knowing that someone's helping.

Quote: " Seriously, it's okay, you can lighten up! You should really learn how to do High Poly art first anyway, especially to break the fear you have. After you make it high poly, work on making it low and bake the detail in"

Lol, good plan

, but I have a theory: Take this image as an example:

If you look at his chest, you can see there are dark lines to show what I think are the original polygons (in the article it said that ZBrush added 500,000 extra polies, and since I saw 16 polys per original face, I got that there were about 31,000 polys originally). I remember once that I subdivided a smoothed out face in blender because I wanted the shine on the model to look much better. Then when I got Dark Shader (I am such a NOOB at shaders) I noticed the per pixel lighting shader. In theory, in the image aboves case, instead of subdividing a lot to make a rubber like shine, wouldn't a per pixel lighting shader work better than making several faces? And is there a way to assign a shader to a certain part of the object (I think I remember Green Gandalf saying that I should use the set limb effect, and then assign a shader per limb, but I'm forgetful, is what he said (I think) true?) And another thing: correct me if I'm wrong, but from all these aricles, aren't the textures and other "maps" actually small in size? Oh yeah, and when I said the different programming languages might help, I thought different languages have different speed to which they recall comands faster, and then that way they lagg at different points... I have no idea, and I think this paragraph is becoming too long, so I'll end it at that.

Back to top

Profile PM Email

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 2nd Aug 2009 01:56 Edited at: 2nd Aug 2009 01:59

Link

I just made another post because the last paragraph was getting too long: [/quote] I think the issue of polycount is far less important that texture resolution, shaders, draw calls, ect Polycount is one of the simplest things a game engine has to do. It's all just basic math so if your computer is fast enough to handle all the points then its going to render fine. There are limits in game engines that define how big meshes can be, but these limits are usually just put in place because some kind of limit has to exist and are often times arbitrary. I don't know any specific limits off the top of my head, but really I think it's more a question of your target system(s) power than what engine you are using when it comes to a models polycount. Really? I think I understand now, the more faces the your mesh has, the more you are tempted to make the texture per face a much more high resolution, then comes more shaders, normal maps, etc; so a game console like 360 or ps3 can use high res textures and shaders much more easily? Yeah, I thought of that a while ago, when I heard the repeated phrase "it all depends on the power of the engine". I thought,

Quote: "So all 360's and ps3's have the same processing power, therefore the game creators already know their target poly count for a game"

Eventually I will ask which programming language is the fastest, therefore can use more polys, shaders, and higher res. textures in a game, but since I'm a noob on the subject, I'll stick to DBP for a while. Then lastly, what is the poly count for normal things in DBP, including texture size and amount of shaders, etc.? And hey, I've learned a lot of stuff in this post, mainly that a lot of modern day games such as Metal Gear Solid 4, Devil May Cry 4, and Ninja Gaiden 2 use Autodesk Softimage. Seems important

Back to top

Profile PM Email

Sid Sinister

19

Years of Service

User Offline

Joined: 10th Jul 2005

Location:

Posted: 2nd Aug 2009 02:55

Link

Quote: "Really? I think I understand now, the more faces the your mesh has, the more you are tempted to make the texture per face a much more high resolution, then comes more shaders, normal maps, etc; so a game console like 360 or ps3 can use high res textures and shaders much more easily? Yeah, I thought of that a while ago, when I heard the repeated phrase "it all depends on the power of the engine". I thought,"

Er, no. Haha, guess you still got more to learn

A texture is like the paper around a pop can. You can deform the pop can all you want, but the texture still remains the same around it. The only thing that creates a higher resolution texture, is a higher resolution texture

Adding geometry does not create a higher res texture.

Advantages of creating a high poly model? Baking the ambient occlusion map out to texture over it to make your low poly look high poly. Those wrinkles you see in Snakes face are NOT actual geometry on that model. However, they were at one point though! Those wrinkles were modeled in Zbrush, and then superimposed into the texture to make the lower poly model look high poly.

Question: So if you can add all the detail you want by faking it, what's the point of having a 15k GOW character? Answer: There's only so much you can do by just faking it. Faking it will never replace the realism that the actual geometry would otherwise create. Also, there are just some parts of the mesh that need to be helped along. In other words, faking a smooth curve isn't enough, you would need to add more to it. For example, you can't fake a smooth edge on a cube. However, add a few faces to the edge and it starts to look rounder. Eventually you'll strike a balance between detail created with geometry and detail created with texture. That's where you want to be at!

As far as your shader questions go, I don't know anything about them. Ask Green Gandalf, as he's the shader king. However, I can tell you that the amount of faces an object has, has nothing to do with the type of material your faking. It's all about reflection and texture.

I hope that helps?

"If I have seen a little further it is by standing on the shoulders of Giants" - Isaac Newton
Current Project: http://strewnfield.wordpress.com/ (Last updated 06/11/09)

Back to top

Profile PM Email

Sid Sinister

19

Years of Service

User Offline

Joined: 10th Jul 2005

Location:

Posted: 2nd Aug 2009 03:17

Link

Another reply from Polycount on the matter:

Quote: "polycount doesn't seem to be as crucial as texture budgets, fill rate (shader complexity), draw calls/batch calls, and etc.

polycount does still come into affect though--having over 32,000 verts can sometimes be a problem (I think it has something to do with storing the verts in a 16bit versus 8 bit). Older engines may assume a mesh is going to be under that and may import the mesh all wonky if at all, connecting edges to the wrong verts and screwing up vertex order and what have you.

Also, having an object take up too much 'space' in the world can sometimes be problematic, since towards the extremities of an engine's limits you can run into precision errors (collision will act funny, things may render weird, etc)

And just because you CAN import a 30k+ mesh doesn't mean you should ... Even current gen engines can crawl when it gets to a million or two verts. If that were ALL characters at 30k then thats only a few dozen on screen..with nothing to account for environments, FX, UI, or etc.

Anyway, I dunno if that answers any questions. There's no quick 'clear answer' for a lot of this stuff"

"If I have seen a little further it is by standing on the shoulders of Giants" - Isaac Newton
Current Project: http://strewnfield.wordpress.com/ (Last updated 06/11/09)

Back to top

Profile PM Email

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 2nd Aug 2009 04:15

Link

Quote: "Er, no. Haha, guess you still got more to learn"

You're right, and the way I know that is that I knew little aobut what you said... back there, or at least the technical terms. And sorry for the mistake about faces and texturing, I guess that only applies to me

. Anyways is that what ZBrush does? I kind of figured from the fact that the name has brush, and Z means the z axis, but anyways, Is it all it does? Add rinkles to a texture? Or sculpt in 3D... That seams actually VERY helpful, but programs like photoshop and the gimp can do that if you're skilled enough and get the parts of the UV map you want to be wrinkled in your imagination... And besides, blender, a free program, has a sculpting option, and by watching a video of making a normal map in blender, I learned that you can bake a model down... So, in other words, stuff like baking these wrinkles in the model can be done easily and much more cheaper? And thanks for the info, you should be a tutor or something, you're good with this stuff ^^. And about what you said, I think the fact that other engines use quad polies instead of tri polies, they save a poly count and overall work. But that's directX for you...

.

Back to top

Profile PM Email

Jeff032

17

Years of Service

User Offline

Joined: 13th Aug 2007

Location:

Posted: 2nd Aug 2009 04:37

Link

Quote: "And about what you said, I think the fact that other engines use quad polies instead of tri polies, they save a poly count and overall work. But that's directX for you..."

Quads need to be converted into triangles to be rendered. In the end, there is no difference in poly count between a model made out of quads and the same one made out of tris.

-Jeff

Back to top

Profile PM

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 2nd Aug 2009 04:56 Edited at: 2nd Aug 2009 05:14

Link

Quote: "Quads need to be converted into triangles to be rendered. In the end, there is no difference in poly count between a model made out of quads and the same one made out of tris"

Really, I always thought it was just directX that uses tri polies when rendering. So in reality game creation uses tri polies period, and 3D animated movies use quad polies and tri polies, right? That is... interesting o.O. And I'd really like to try the baking thing with blender, you know, just make something like a basic mouth shape, and then try the wrink thing with blenders subdivision, sculpting, and baking options. Oh, and is there a way like anti-subdividing? I mean like is there a way to subdivide , and then simplify the subdivised face without using the undo command?

Back to top

Profile PM Email

Sid Sinister

19

Years of Service

User Offline

Joined: 10th Jul 2005

Location:

Posted: 2nd Aug 2009 05:15

Link

Both tri's and quad's are always there =] You create using both, you can convert to both, both industries use both.

SJH, even though you're still uber confused, I admire your thirst for correct knowledge. I'll see if I can't track some resources for you in the next couple days so you can understand better

"If I have seen a little further it is by standing on the shoulders of Giants" - Isaac Newton
Current Project: http://strewnfield.wordpress.com/ (Last updated 06/11/09)

Back to top

Profile PM Email

Jeff032

17

Years of Service

User Offline

Joined: 13th Aug 2007

Location:

Posted: 2nd Aug 2009 05:18

Link

I believe its more that working with quads can be better for doing modeling/animating, but graphics cards work with triangles. So quads & tris when making models --> tris when rendering them. But I could be wrong.

Back to top

Profile PM

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 2nd Aug 2009 05:29 Edited at: 2nd Aug 2009 06:05

Link

Quote: " you can convert to both"

Hmm? I didn't know you can convert tris to quads.

Quote: "you're still uber confused, I admire your thirst for correct knowledge. I'll see if I can't track some resources for you in the next couple days so you can understand better "

Wow, thanks

, that's very nice of you

. Oh, and is there a way like anti-subdividing? I mean like is there a way to subdivide , and then undo the subdivised face without using the undo "ctrl+z" command? And hey about the baking thing with wrinkles, how does that work exactly? Is there a low poly version of the model that has a UV Map already, and then you open a second version of the model and subdivide it, and using a sculpting tool to make the rinkles, and then finally using a bake tool to make the the 3D wrinkles part of the texture... EDIT: just tested baking, a wrinkle in, and what do you know, its easy:

I guess it would look better on some thing like Snake's face above, the wrinkles blend in better than a solid edge, when a smooth surface like a face supports wrinkles... Bottom line blender seems perfect for the whole wrinkle thing

. It can subdivide faces as much as you want, it can use its sculpt mode to make the wrinkles, and then it has a simple render baking system that can be saved as the UV Texture... Now I know how the whole sculpting and baking process can make low poly models look smoother. Normally I have to add MANY faces into a model to make it look like it has a lot of faces, but now I can save the low poly model, save a second version of it, subdivide the model several times, and then sculpt out some things like wrinkles or keep the smoothness of the current look of the model, and bake it and save it as a new UV TEX. That's really handy ^^, if only I could make the lighting high res on the model as well, as in have a high res lighting effect on a low poly model. I assume that this whole baking wrinkles thing is just for that: wrinkles, otherwise the lighting on those models would look weird...

Back to top

Profile PM Email

Sid Sinister

19

Years of Service

User Offline

Joined: 10th Jul 2005

Location:

Posted: 2nd Aug 2009 06:08

Link

Quote: "Hmm? I didn't know you can convert tris to quads. "

Two tri's = quad. In Max you can switch between Editable Poly and Editable Mesh, Mesh gives you Tri's, Poly gives you Quad's.

I'm headed to bed, I'll look at your post tomorrow. Let's focus on you learning correct info on 3D stuff before setting off to do UV baking. That's kind of advanced stuff.

Also, if you can help us help you, format your paragraphs a bit better. Like, use paragraphs so it's not a huge wall of text

"If I have seen a little further it is by standing on the shoulders of Giants" - Isaac Newton
Current Project: http://strewnfield.wordpress.com/ (Last updated 06/11/09)

Back to top

Profile PM Email

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 2nd Aug 2009 06:18

Link

Lol, K. Thx for the help, Too bad Blender doesn't use that polygon mesh thing, but that's an wasy thing to work around: Just don't convert any of the quads to tris, and save it, and if I want a tri thing just use convert the model to tris and save or just load into dbp, because I heard somewhere that it converts the polys into tris.

Back to top

Profile PM Email

SJH

16

Years of Service

User Offline

Joined: 19th Oct 2008

Location:

Posted: 2nd Aug 2009 06:41

Link

Oh and just a thing, in that article about the MGS making thing, they ussually made an action figure of the model, and the nused the action figure as reference images. Would something like a drawing be good enough? I certainly hope so

.

Back to top

Profile PM Email

Sid Sinister

19

Years of Service

User Offline

Joined: 10th Jul 2005

Location:

Posted: 2nd Aug 2009 22:04

Link

Yeah, something like a drawing would work just fine. It's best to have a front, side, top, bottom and three quarter view though. You can set your viewport backgrounds to this images, or make a plane and set the texture to it, so you can model it accurately.

"If I have seen a little further it is by standing on the shoulders of Giants" - Isaac Newton
Current Project: http://strewnfield.wordpress.com/ (Last updated 06/11/09)

Back to top

Profile PM Email

Sorry your browser is not supported!

3 Dimensional Chat / Poly Count Guide?