Hard Light Productions Forums
Modding, Mission Design, and Coding => The Modding Workshop => Topic started by: Starman01 on July 16, 2005, 06:49:16 am
-
Hello,
just a short question, before I work on new models. What's an acceptable or the best efficient texturing for models:
Currently I try to make capships that use max 3 maps with 2048x2048.
But what is better for fightermodels (around 1500 polies) in a way of performance. 2 Maps with 1024x1024, or one with 2048x2048.
Afaik the less textures, the better it is. But isn't a 2048 map very heavy for a fighter ?
-
Uhm you can't compare it this way.
One 2048*2048 map has twice the ammount of pixels of two 1024*1024 maps.
It needs the double ammount of memory too. So it's worse for PCs without much ram/gfx ram.
-
I know, but isn't it better for the performance in a HT&L engine, when there is only one map that must be rendered ?
AFAIK, the HT&L engine renders the model completly for each map involved, so one map means only one rendering time ?
-
Das hängt nur vom PC ab der verwendet wird.
Wenn die Grafikkarte genug Ram hat, dürfte die Variante mit der 2048² Map schneller sein.
Wenn die Grafikkarte nicht soviel Ram hat, ist die zweite Lösung besser.
Ich halte zwei Maps bei einem Jäger noch für ok.
Wenn die Qualität von zwei 1024² Maps ausreicht, würde ich mich dafür entscheiden.
-
Originally posted by DaBrain
Uhm you can't compare it this way.
One 2048*2048 map has twice the ammount of pixels than two 1024*1024 maps.
It has 4 times the amount of pixels actually.
-
I'm not sure if Bobb has completed his overhaul of the tex system.
But before it would be the one map, as you'd only need 3 render passes (shine/glow maps) as opposed to six.
-
But the render passes for the bigger map would take longer, right?
And it needs more memory anyway.
... For TBP, I'd say "Go for the big texture", because TBP doesn't use very much memory, but I don't know about WC:Saga.
-
Originally posted by Nico
It has 4 times the amount of pixels actually.
Hehe nope.
2* 1024² = 2097152
1* 2048² = 4194304
4194304 / 2097152= 2
-
Originally posted by Nico
It has 4 times the amount of pixels actually.
One 2048 x 2048 has 4 times the amount of pixels as a 1024 X 1024.
Therefore one 2048 x 2048 will have twice the amount of pixels as 2 1024 X 1024s, which is what DaBrain said.
-
Umm, no... Each step up is a 4x increase. Because your essentially adding 3 more of the same image space.
See?
-
:rolleyes:
Raa... did you really read what I wrote... ? :rolleyes: ;7
Starman asked about two 1024² maps.
:lol:
-
Me? Read?
(BTW; yah, two files, I'm not disputing that. Alot of people still seem to think that the 2048 is only 2x more than the 1024, though. So whatever :p)
-
I know, but I've worked with maps long enough to know this stuff. :nod:
-
Then it wasn't for you :p
-
Originally posted by DaBrain
Hehe nope.
2* 1024² = 2097152
1* 2048² = 4194304
4194304 / 2097152= 2
yeah, sorry, missed the part where you said two 1024x1024 maps.
-
Thanks for the answers and explanations :)
So it looks like each version has it's up and downs, either effective rendering operation, or the lower memory usage.
I think, I will then decide it according to the models, though 2048 er maps seem to be more logical, since the older machines will die sooner or later :)
-
I'd say to put everything into a single 1024x1024 maps, which should be enough for almost any fighter, this way you won't have any doubts:p
BUT personally, I'd go in that situation for 2 1024 rather than a single 2048
it is true that the more textures the more passages, but with 2x1024x24bit you save around 10 mb compared to 1x2048 if I'm not wrong, which doesn't sound that few to me, talking of just a fighter. I may be wrong but to a common sense I'd say that gpu's calculation power grow more than gpu's memory in perspective, and that if you drain the memory, you slow down the calculation time too; it's not a problem if it happen for a single model, but if you do it regulary I dunno what could be the result. I'm thinking at how the high res packs for fso can be intense even for non old PCs, sometimes. Well I'm curious to see what bob will say
-
I'd say, stick with one 1024x1024 map. With fighters, higher resolution is really not needed, unless you spend more time in external view looking at your ship from every angle. The original map for the ezechiel was a whooping 4096² map, that I cut down to 2048² for ingame, but eventually, I reduced it to 1024 (and iirc I joined a 512x512 map for lower specs), and it's as good, considering you can't stick your nose that close to the ship anyway.
-
:nod:
I don't see a need for anything about 1024 for a fighter, to be honest. Even then a lot of the detail is pretty much invisible to the player fighting at combat range.
-
you know the dimentions of the texture need to be power of two but the texture does not need to be square so 2048*1024 is just fine, so if you have two 1024 textures, just squash the first texture's coords to one side and the other texture's coords to the other, then just expand the canvas size of the first texture and paste the second onto the side of it, and tada! done.
and yes larger textures meen slower passes, it's like this, a render pass takes about five years, if you double the size of the map it takes six years, so what's slower six years or ten years?
-
Try DDS textures as well, for fighters they're supposed to be pretty good.
-
Originally posted by Bobboau
you know the dimentions of the texture need to be power of two but the texture does not need to be square so 2048*1024 is just fine, so if you have two 1024 textures, just squash the first texture's coords to one side and the other texture's coords to the other, then just expand the canvas size of the first texture and paste the second onto the side of it, and tada! done.
and yes larger textures meen slower passes, it's like this, a render pass takes about five years, if you double the size of the map it takes six years, so what's slower six years or ten years?
Damn.... I said I knew enough about maps... but I wasn't able to find this simple solution. :mad:
Well I have no choice: "I have to learn more about maps." :ick:
-
And 2 maps should be no problem for HT&L.
-
Originally posted by Bobboau
then just expand the canvas size of the first texture and paste the second onto the side of it, and tada! done.
And tada! Redo your UV mapping :D
-
Originally posted by Roanoke
And 2 maps should be no problem for HT&L.
And who cares about FPS? Let's just max out all our ships to 8 textures...
-
Originally posted by Nico
And tada! Redo your UV mapping :D
well it's trivial to adjust the uvmaps when they are "regular", like in this case. You can do that in 2 minutes in lithunwrap and, i guess, in any other uveditor.
BTW, bob, my question is if the render time is mmm addictive or linear, I mean: we know that fso likes few high res maps, and for each single ship it is better to have a single map, but what happen if you overuse the ram memory with various too much high res textures? In other words, isn't there the risk, when the game has to work with some dozens of high res maps,to have your system going choppyer due to consumed memory?the difference in terms of mem usage isn't that small between a 2048 and a 1024 maps
-
"fso likes few high res maps"
close, FSO likes few maps, it consiters one high res map a lesser evil than two low res maps.
with fewer maps you are less likely to have a swapout problem, and when you do swap out a texture you do it all at once, it's all about makeing the smallest number posable of big batches, and to a lesser degree changeing the smallest number of things posable.
-
Well, if you have two 1024x1024 maps, and a 2048x1024, the latter will actually use somewhat less RAM as the overhead of loading one texture is less than that of loading two textures.
-
Originally posted by KARMA
well it's trivial to adjust the uvmaps when they are "regular", like in this case. You can do that in 2 minutes in lithunwrap and, i guess, in any other uveditor.
What do you mean by regular? I´m really interested to learn anything I can about UVmapping, since it seems everyday I´ve got to do more of it... I´ve done UVmapping mostly by selecting faces and applying a uvmap modifier( in most cases planar) and then doing the same for all the other polys...
-
Uhhh. Not to derail the thread or anything, but does anyone know where I can get a copy of Lithium Unwrap? I googled and found nothing but bleedy reviews.
-
Check the modding section of my FAQ.
-
Originally posted by General Kazooie
What do you mean by regular? I´m really interested to learn anything I can about UVmapping, since it seems everyday I´ve got to do more of it... I´ve done UVmapping mostly by selecting faces and applying a uvmap modifier( in most cases planar) and then doing the same for all the other polys...
this is what I do in lithunwrap. There are other unwrapers which do almost the same, and I know that max has one internal. If you want lithunwrap, follow karajoma's instruction above.
1- create the desired texture in photoshop out of all the smaller textures and load the model in lithunwrap
2-create a new lith material and assign the new texture to it
3-for each original material, I select all the uvs and assign em to the new material
4-each time I do so, I scale the uvs by the same scale of the original texture compared to the new one. I mean that if the original texture is 1/4 of the new texture, I scale the corresponding uvs to 25%, then I move them in the correct position (either manually or automatically, depending by the situation)
5-I continue until all the faces are assigned to the new material, then I delete the old ones, save, save a backup, save a second backup:p, load the model in tuespace and make the smoothgroups.
Note about smoothgroups: I can't remember now if ts see as different materials the different lith materials, in that case it'd be easyer in lith to make many materials (one for each group) with the same new texture. Elseway you'd have to make many materials anyway, but before saving substitute from each material the texture with a solid colour or with some casual textures: this way, when you load the model in TS, you won't have to select manually the various faces to create the smoothgroup
Of course, baking textures in max is faster, if you can do it, and all of this procedure is faster and has a sense if you can make some regular subdivision of space (like a single 1024x1024 out of 4 512x512, or 1 1024x512 out of 8 256x256 or something like that)