Topic: Poly Count versus Texture size...  (Read 4898 times)

0 Members and 1 Guest are viewing this topic.

Offline Magnum357

  • Lt.
  • *
  • Posts: 641
Poly Count versus Texture size...
« on: June 13, 2006, 06:31:47 pm »
Recently, I decided to try too get back into some modeling again and was wondering something... when making a SFC Model (or I guess any 3D model for that matter) does the amount of Polys or the size of the Textures make a difference when your computer trys too load the model in a 3D editor or a game?

I have been redoing a few of my older models and have bumped up there Poly counts a bit (I usually prefer to make my Models at low Poly count too help out low End systems) and so far I'm very suprised to find out that the higher Poly count does not seem too effect the performance of my SFC Games even though my computer is a bit dated.  So far, it seems the more Texture intensive I have a model made, the more it can eat up the resources of the Computer or how well a game or 3D app can handle the model.


I was just wondering what other modelers think is the big determination of what eats up system resources... Poly count or Texture Sizes/Number of Textures a model has?  Opinions?  Facts?
"I sure am glad I like SFB!" - Magnum357 (me)

Offline SkyFlyer

  • D.Net Beta Tester
  • Commander
  • *
  • Posts: 4240
  • Gender: Male
Re: Poly Count versus Texture size...
« Reply #1 on: June 13, 2006, 06:40:52 pm »
Well you must remember that SFC2 was made with a 500mhz machine was "great."

Upping the poly counts would have a definate effect on a  500mhz machine... But a 1.5 or a 2ghz+ machine... not much.

But yes, it does affect performance.
Life is short... running makes it seem longer.

"A god who let us prove his existence would be an idol" - Dietrich Bonhoeffer

Offline AlchemistiD

  • Lt. Junior Grade
  • *
  • Posts: 440
  • Gender: Male
  • No Replacement For Displacement
Re: Poly Count versus Texture size...
« Reply #2 on: June 13, 2006, 08:24:33 pm »
Well you must remember that SFC2 was made with a 500mhz machine was "great."

Ahh the old days when we used to kick it. *listens to too much hotspot*

Offline Starforce2

  • Bridge Commander Ambassador
  • Commander
  • *
  • Posts: 2827
    • Nightsoft SFC File Dump
Re: Poly Count versus Texture size...
« Reply #3 on: June 13, 2006, 09:24:51 pm »
using a few large textures is also supposed to be better than a ton of small ones, or so I am told.

Offline GotAFarmYet?

  • Lt. Commander
  • *
  • Posts: 1189
Re: Poly Count versus Texture size...
« Reply #4 on: June 13, 2006, 11:17:20 pm »
Your answer is yes.

They both effect the game. Larger textures allows more details to be placed on them but the newer cards handle that better.
In SFC 1 and 2 the whole ship is drawn every poly so the less the better the game maxs out at about 15,000 or so polys.

So both effect but but most newer computers handle things just fine, unless you are loading 15-20 10,000+ poly ships each with 10+ 2048x2048 high resolution textures.

anything with 5000 or less polies and 8 or less 512x512 textures runs with out a hitch.
People always said they wanted the government to listen to them and now the government is listening, taking notes and names...and coming to see you soon!

America-Not the land of the free anymore...
 Its the land of the freeloaders

Remember the axiom of big government bureaucrats: If it moves, tax it. If it keeps moving, regulate it. When, finally, under the crushing weight of taxes and regulation, it stops moving, subsidize it.

Offline SkyFlyer

  • D.Net Beta Tester
  • Commander
  • *
  • Posts: 4240
  • Gender: Male
Re: Poly Count versus Texture size...
« Reply #5 on: June 14, 2006, 05:23:14 am »
Your answer is yes.

They both effect the game. Larger textures allows more details to be placed on them but the newer cards handle that better.
In SFC 1 and 2 the whole ship is drawn every poly so the less the better the game maxs out at about 15,000 or so polys.

So both effect but but most newer computers handle things just fine, unless you are loading 15-20 10,000+ poly ships each with 10+ 2048x2048 high resolution textures.

anything with 5000 or less polies and 8 or less 512x512 textures runs with out a hitch.

+karma for the exact info that I did not have. :)
Life is short... running makes it seem longer.

"A god who let us prove his existence would be an idol" - Dietrich Bonhoeffer

Offline Magnum357

  • Lt.
  • *
  • Posts: 641
Re: Poly Count versus Texture size...
« Reply #6 on: June 14, 2006, 06:15:44 pm »
Hmmm... very interesting.  Seems like everyone has a little bit different opinion on which one (Poly Count versus Texture Size/Number of Textures) effect an SFC Model the most.  I think we all agree that you put enough Polys and enough Texture on any model, its going to need more Horse Power from your computer too load it, I was just wondering which one of these effect Performance the most in the game?

I make my models Very Low Poly (usually 500 too 2500 Poly Max) and because of the limitations of Milkshape (it will only allow 256 X 256 Textures for exported SFC Models) all of my Textures are small so maybe that is part of the reason why even a model at 2500 Polys doesn't seem too effect performance on my old P3 600 Mhz Computer.
"I sure am glad I like SFB!" - Magnum357 (me)

Offline OlBuzzard

  • renegade
  • Lt. Commander
  • *
  • Posts: 1759
  • Gender: Male
Re: Poly Count versus Texture size...
« Reply #7 on: June 14, 2006, 06:37:40 pm »
IMHO Magnum that is essentially the key to the matter:  what sort of system you are working with.  If ya remember a lot of the old Taldren models were some what bland compared to todays standards.  It wasn't that the model designers were bad ..  just very limited to the "space" provided so to speak.  The general conscious is that if you are designing for a more "basic" system then the less polys the better.  Also ...  just as you indicated ... the smaller bmp files are needed as well.

Today there are some really awesome PC's ... many with well over 1gig ram, 180g hds and 3.2 gh or more p-4 chips  (like mine ... and it's almost ready for the next big overhaul ...almost 2 years old now).

The newer PC's can flat get with the program (so to speak). 

At any rate that is almost getting OT ..   but I think you have the idea !
If you aim at nothing:  you WILL hit it every time !

Offline GotAFarmYet?

  • Lt. Commander
  • *
  • Posts: 1189
Re: Poly Count versus Texture size...
« Reply #8 on: June 14, 2006, 06:48:27 pm »
Magnum357

if you can only export at 256 x 256 then resize one set for the export then use the orginals in the game. The names have to be the same for it to work but any texture once exported will use the same mapping and name. So make a low one for the export and a high one for the game or release.
People always said they wanted the government to listen to them and now the government is listening, taking notes and names...and coming to see you soon!

America-Not the land of the free anymore...
 Its the land of the freeloaders

Remember the axiom of big government bureaucrats: If it moves, tax it. If it keeps moving, regulate it. When, finally, under the crushing weight of taxes and regulation, it stops moving, subsidize it.

intermech

  • Guest
Re: Poly Count versus Texture size...
« Reply #9 on: June 14, 2006, 10:15:01 pm »
I have found that the more mapping on a texture, the slower my computer gets. Polycount has not really been an issue. If I put a very high poly ship into the game without a texture it works like a dream. If I put that same ship in with a 4X4 texture (yes small, and yes I am a time waster) I still have no problem. But when each poly is mapped to a different part of a large texture, or the texture is tiled in some way, then I have problems. Keep in mind I am using a 900 mhz computer with 512 mb ram.

Offline I, Mudd.

  • Lt. Junior Grade
  • *
  • Posts: 251
  • Gender: Male
  • Still Building Models Nobody Wants ...
Re: Poly Count versus Texture size...
« Reply #10 on: June 14, 2006, 10:27:16 pm »
That's an interesting theory, Interstellar Machine. As much a PITA as multiple texture files might be, that might be worth experimenting with ...

Would the size of the multiple texture files  figure into that as well, I wonder, and at what point would they become a tax on the system?

Definately food for thought.

I, Mudd.

Offline Sochin

  • Lt.
  • *
  • Posts: 747
  • Gender: Male
Re: Poly Count versus Texture size...
« Reply #11 on: June 15, 2006, 04:31:15 am »
I remeber when the Courageous came out, it had three download type Hi - Med - Low poly versions which to me was a very good idea as you could pick and choose the quality of the model you wanted. Reason for this, or atleast for me at the time is that another and I on 9th Fleet did the signatures and we where allways on the lookout for high quality models that we could use. Not everyone has 3DS or equivalent so Taldren Model Viewer was ideal, and having high poly models was excelent.

Keep up the good work.

Offline OlBuzzard

  • renegade
  • Lt. Commander
  • *
  • Posts: 1759
  • Gender: Male
Re: Poly Count versus Texture size...
« Reply #12 on: June 15, 2006, 08:02:24 am »
That's an interesting theory, Interstellar Machine. As much a PITA as multiple texture files might be, that might be worth experimenting with ...

Would the size of the multiple texture files  figure into that as well, I wonder, and at what point would they become a tax on the system?

Definitely food for thought.

I, Mudd.

hmmm   yes ,...  me thinks this has a strong, valid point.  My son worked extensively on a Mod ...  and seems like this might be ..

frankly I'm not well enough versed in the area of loading an entire mod ....  let's face it reviewing one model at the time is one thing ...  but an entire game ...   that presents another issue.

I know when I transfer a file of the model by its self to one of my buds  ( without any bmp files ...  just the mesh ) ...  it only takes a couple of minutes.   Even a larger model  (poly count wise) ...  but as Mudd has pointed out ... WOW what a difference in time with the entire model (complete with all bmp files).  Even zipping them up still takes a lot more time.  That has to say someting about how the size of those bmp files must also affect the process.

If you aim at nothing:  you WILL hit it every time !

Offline Bartok

  • Lt. Junior Grade
  • *
  • Posts: 293
  • Gender: Male
Re: Poly Count versus Texture size...
« Reply #13 on: June 15, 2006, 08:41:43 am »
WOW - Awesome thread here....

I've been wondering about all the variables as a newbie modder and a dyna admin.  Some of the lwo and ads? files I converted in Milkshape had huge poly counts and huge bmps (when imported) and I would see some performance issues when locally testing my server.  My game PC and the one i use Milkshape on has a G of ram and (i think) a 2 gig processor.... 3 year old Optiplex GX270

Anyhow - I would see some lag with the high res ships, particularly when they closed with the enemy... Nothing major but I imagined it not on a Lan but on the dynaverse with folks variable connection speeds and so.... have stalled on that mod.  The suspect .mod files are in that 3-5meg range, not including some fat bmps...

Lotsa food for thought and experimentations here :)

Offline GotAFarmYet?

  • Lt. Commander
  • *
  • Posts: 1189
Re: Poly Count versus Texture size...
« Reply #14 on: June 15, 2006, 10:10:48 am »
Bartok,
What you are discribing is why LODs were used. LOD 2 was the one most often used with medium resolution textures and a lowered poly model. LOD 3 was the far away one and had low resolution textures and a very low poly model. LOD 1 was only used up close had high resolution textures and a high poly model. now what normally happens is your ship flicks between LOD 1 & 2 depeding on the distance you set the model from you in the view. Enemy ships start off at LOD 3 and move to LOD 2, upon occasion if you get really close and have your ship close in the view it qwill make LOD 1 as well. The reason the LODs where there was to reduce the impact on the systems playing the game. It was also the reason so many ships never make it into mods, they don't have LODs and too many impact the system. Over time allot of ships got high res textures over all the LODs and it does make a difference, but not as much as not having LODs at all. Now adays most of us have systems were the effects are not noticed nearly as much, and most high poly high res ships don't bother us. You can still a difference but it is only in the mission loading time as all the ships are loaded into memory. It is odd that you can see it flicker the first time LODs are loaded, that you don't see with none LOD ships, though the load time is less with the LOD ships. Once the ship is completeky in memeory it is pretty hard to see any differeces in game play.

So currently it really seems only load time is effected, as no mission I can thing of loads the maxium number ships playable. I tried it out with 36 different ships all with LODs, all with no LODs, All with low res textures, all with high res textures. Mostly it was all in the loading time for the mission that varied once the game got going it was not so noticeable, all the spare time cycles for the processor get taken up by the AI more than the models themselves. The Process was tested against a AI and a friend on a networked PC, the networked players used less processor time over the AI.

So in conclusion LODs still make a difference but no one is really using them anymore, kitbashers usually don't. It seems that there might be a maximum texture size allowed as well but none of my test ships had more than 12 textures and 3 or less at 1024 x 1024. Load time seemed to be the most effected area, about 5-8 seconds variation could not be discovered, depending on the models used and the texture sizes. Note on the variation it would change using the same models at times, no reason why.

So models with LODs and proper textures to go with them increase the folder size, but work better than none LOD ships, but with todays systems it really doesn't matter much.
People always said they wanted the government to listen to them and now the government is listening, taking notes and names...and coming to see you soon!

America-Not the land of the free anymore...
 Its the land of the freeloaders

Remember the axiom of big government bureaucrats: If it moves, tax it. If it keeps moving, regulate it. When, finally, under the crushing weight of taxes and regulation, it stops moving, subsidize it.

Offline Magnum357

  • Lt.
  • *
  • Posts: 641
Re: Poly Count versus Texture size...
« Reply #15 on: June 15, 2006, 05:06:07 pm »
GotAFarmYet?, that is very interesting info about .LOD files.  I have seen them but never what they were for until you explained it. 

Also, I'm not sure what you mean above about my Milkshape Export problems.  I actually do make my Textures a lot bigger then reduced them too 256 X 256.  Apparently, the plugin too allow Milkshape to export to .mod files will not allow you to use Textures greater then 256 X 256, unless there is a new Milkshape Plugin I'm not aware of. 

But this really doesn't bug me much with Texture Sizes.  I think I agree with Intersteller Machine, with my experience with SFC models, it seems the amount of Texturing and the Texture sizes make a bigger difference in performance then the Poly Count.  I have seen some fine TMP Connie and Excelcior Models in the past (superb work by a few great Modelers on this board) but some of these fine models have so many textures (sometimes a Large 1024 X 1024, and so many tiny little Textures too make a single Window or Photon Tube) that it can be hard for a low end machine to run the game well.  And its not always the Loading part of the game that gets effected, sometimes during Game play these High textured Models are unforgiving too Low End System users. 

Part of the reason why I brought this up is too find out which is more "Critical" to the Model Developer.  As I said before, I like making Low Poly Models, but if I can get away with it, it would be nice too give my Models a Poly Count boost if Low End System Users don't have too suffer for it.
"I sure am glad I like SFB!" - Magnum357 (me)

Offline GotAFarmYet?

  • Lt. Commander
  • *
  • Posts: 1189
Re: Poly Count versus Texture size...
« Reply #16 on: June 16, 2006, 12:47:50 am »
Magnum357,

If you want to lower the impact then you will have to make LODs, but only the max plugins and max allow you to add them to a mod file. That was the only purpose of them so that only close up screen shots would load LOD 1, well most game play would be with a lower poly/ resolution LOD 2. LOD 3 was good at lowering the impact of things off screen and far away, it was a good system they used. This was also in all three versions 1-3, LODs are used in the orginal models SFC 3 has a better loading system and polygon drawing that SFC 1 & 2 but I could not stand the game play of SFC 3.

About the exporting...
Save your orginal textures in a seperate folder, then make the smaller versions for exporting. Once exported move the mod file to the save folder. It should allow you to use those higher res textures that way instead of the 256 x 256 ones. the mapping to the model is the same reguardless of the texture size you use, just the names have to be the same (FCA_1 has to be named FCA_1 in both folders). I replace textures with other ones all the time nameing is important as then the same locations will be mapped to the new texture.
People always said they wanted the government to listen to them and now the government is listening, taking notes and names...and coming to see you soon!

America-Not the land of the free anymore...
 Its the land of the freeloaders

Remember the axiom of big government bureaucrats: If it moves, tax it. If it keeps moving, regulate it. When, finally, under the crushing weight of taxes and regulation, it stops moving, subsidize it.