Here's my own skin (lacking texture2)...
...and smoth's variant of it.
Now stop bitching. :P
Excessive detail
Moderator: Moderators
-
BrainDamage
- Lobby Developer
- Posts: 1164
- Joined: 25 Sep 2006, 13:56
http://spring.clan-sy.com/phpbb/viewtop ... 3&start=20Sleksa wrote: WHERE CAN I FIND WALLPAPER-SIZE PICTURES OF THESE O__O
Ahoy, haven't time to read the whole thing, just wanted to point out:
Those are peewees and aren't the sort of unit I'm really talking about for a current project.
Eh, er.. what I'm getting at is, peewees are freaking everywhere and in huge numbers. They're not units you'll see a handful of in a rare situation.
Still, I believe if you intend to release those as part of a larger package which will likely not reach completion any time soon realistically, you're wasting effort trying to optimize them so they'll run on outdated computers which will only become more outdated.
I feel like making a chart, and hencenforthen, I shall.
Notice how the world goes through quantum phase shift trilogy XXTREME-ROFL when 8800 is released, thereby necessitating that everything that comes after it is at least somewhat competitive, while everything that came before it looks like concentrated suckery.
2007 marks:
-Core 2, meaning CPU limitations are 20% reduced
-Nv8800, meaning GPU limitations are basically gone for the time being
-Cheap RAM
-Cheap everything else
-"everything else" becoming cheap due to the first two things, since "everything else" is now totally obsolete.
You cannot buy a new computer that doesn't have a core 2 chip in it, and you can't buy a new gaming computer that doesn't have a GF8 series GPU in it. This means exponential growth of people who have PCs that don't completely suck.
There's a completely new standard now. GF8 and Core 2 being relased makes everything that came before them cheap as dirt, so even if you don't feel like blowing $400 or whatever on a massive quantum leap of computing power, you can spend half (or less) and still get a massive upgrade. You HAVE TO upgrade your computer from time to time.
I think the reason nobody here can face this fact is because virtually the entire TA modding community has devoted their life to finding ways to reduce their polycount by 3% without anyone noticing. Big whoop. 3%, that means you can now put 3% more units out there. Except your CPU can't handle that many units' pathfinding so it's irrelevant anyway.
Charts RULE!
Here's another one!
I couldn't decide what the Y axis should represent. I just really enjoyed putting that hard limit on 500 units, approximately where it belongs even with a high end processor. It's far far far far far far far far far far lower if you don't have a high end CPU.
But here's where I start really shitting bricks. If someone needs a high end CPU before they can even START requiring a high end GPU, why the heck are we spending all time time making optimizations for the GPU, when it's not even the problem!? Nobody will have a high end-CPU unless they also have a high end GPU, there's no such thing as a "hardcore word/powerpoint user"! It's completely beside the real problem!
This is basically the biggest reason I thought GEM would be a good idea here - the ships are super-detailed and few in number, meaning that I'm not hitting any artificial limits, I'm using both GPU and CPU in about the same amount.
Heck, even your standard common cheapass onboard GPU fare can handle a handful of high detail models.
OMFG ANoThEr ChArTTT!!!! :OJ :OJSKS
Notice how I avoid bottlenecks altogether by designing the whole game around the reality of things: CPU cannot handle 30 thrillion (thrillion. Yes, I thought it sounded awesome too) low poly units, but the GPU can. Oh but crap. That means you can handle roughly 500 tops, and you're wasting effort on unnecessary optimization. However, any GPU can handle 30 Gigawattjillion low poly units. Oh the irony. You've optimized for the wrong thing! Arrgh! Oh the spearheaded moses beak!
However, I design the entire game around the fact that cheap GPUs can handle, let's pick a random figure, 100 high detail units easily, and so can cheap GPUs. So I have a balanced load across CPU and GPU. Roughly 50% utilization in common circumstances on common computers, since the entire game was designed with the real-world limitations in mind, not fantasy "well if I go with 100 polys per unit I'll have lots woot!".
(BTW, the last chart on the bottom right says "low poly stuff". I ran out of room!
)
"But Caydr," you say, "You've just pwnt yourself! You're remodelling TA units with like 1000 polies apiece! Holy COW!" "Ah, my sexy young steed man," I respond, "You forget that I'm in the process of remodelling/retexturing about 400 units singlehandedly whilst simultaneously working on 5 mods that nobody believes exist due to the fact that I'm working on 5 of them whilst remodelling/retexturing 400 units singlehandedly. Do you figure I'll be done those by year-end?"
I am after all a talentless hack. It'll probably be a few years at least. Given that even now you can buy a GF8 which is capable of rendering - easily - 500,000 polygons, without breaking 60 FPS, it's my belief that by the time I finish, such GPUs will be old technology and anyone will be able to handle 1000-poly units (1000x500=500,000), except for the same group of geriatric stick-in-the-muds that have Pentium IIIs who will probably never upgrade, spending their working hours doing nothing but complain about how their prebyzantine computers cannot handle all the unnecessary eyecandy in Windows 98.
Roflies, remember how I'm an idiot and GEM will never work because there's too many moving pieces and all my tests involve static units? The battle I did last week involving 300 8,000-triangle ships each armed with 16 independently-targeting turrets, 32 barrels, and rotating radar dishes, with 400 1,000-triangle fighters buzzing around, with all features except shadows (which look crap on large models) turned on, which produced a battle with a low limit of 20 FPS and took over 10 minutes to finish, on a 20x20 map, while I accidentally forgot to shut down the resource-hogging Azureus bt client, which was apparently downloading at about 150 KB/s, seems to have disproved that. Especially since the entire game has been designed around the fact that you'll see perhaps 30 of them in a battle, tops.
Those are peewees and aren't the sort of unit I'm really talking about for a current project.
Eh, er.. what I'm getting at is, peewees are freaking everywhere and in huge numbers. They're not units you'll see a handful of in a rare situation.
Still, I believe if you intend to release those as part of a larger package which will likely not reach completion any time soon realistically, you're wasting effort trying to optimize them so they'll run on outdated computers which will only become more outdated.
I feel like making a chart, and hencenforthen, I shall.
Notice how the world goes through quantum phase shift trilogy XXTREME-ROFL when 8800 is released, thereby necessitating that everything that comes after it is at least somewhat competitive, while everything that came before it looks like concentrated suckery.
2007 marks:
-Core 2, meaning CPU limitations are 20% reduced
-Nv8800, meaning GPU limitations are basically gone for the time being
-Cheap RAM
-Cheap everything else
-"everything else" becoming cheap due to the first two things, since "everything else" is now totally obsolete.
You cannot buy a new computer that doesn't have a core 2 chip in it, and you can't buy a new gaming computer that doesn't have a GF8 series GPU in it. This means exponential growth of people who have PCs that don't completely suck.
There's a completely new standard now. GF8 and Core 2 being relased makes everything that came before them cheap as dirt, so even if you don't feel like blowing $400 or whatever on a massive quantum leap of computing power, you can spend half (or less) and still get a massive upgrade. You HAVE TO upgrade your computer from time to time.
I think the reason nobody here can face this fact is because virtually the entire TA modding community has devoted their life to finding ways to reduce their polycount by 3% without anyone noticing. Big whoop. 3%, that means you can now put 3% more units out there. Except your CPU can't handle that many units' pathfinding so it's irrelevant anyway.
Charts RULE!
Here's another one!
I couldn't decide what the Y axis should represent. I just really enjoyed putting that hard limit on 500 units, approximately where it belongs even with a high end processor. It's far far far far far far far far far far lower if you don't have a high end CPU.
But here's where I start really shitting bricks. If someone needs a high end CPU before they can even START requiring a high end GPU, why the heck are we spending all time time making optimizations for the GPU, when it's not even the problem!? Nobody will have a high end-CPU unless they also have a high end GPU, there's no such thing as a "hardcore word/powerpoint user"! It's completely beside the real problem!
This is basically the biggest reason I thought GEM would be a good idea here - the ships are super-detailed and few in number, meaning that I'm not hitting any artificial limits, I'm using both GPU and CPU in about the same amount.
Heck, even your standard common cheapass onboard GPU fare can handle a handful of high detail models.
OMFG ANoThEr ChArTTT!!!! :OJ :OJSKS
Notice how I avoid bottlenecks altogether by designing the whole game around the reality of things: CPU cannot handle 30 thrillion (thrillion. Yes, I thought it sounded awesome too) low poly units, but the GPU can. Oh but crap. That means you can handle roughly 500 tops, and you're wasting effort on unnecessary optimization. However, any GPU can handle 30 Gigawattjillion low poly units. Oh the irony. You've optimized for the wrong thing! Arrgh! Oh the spearheaded moses beak!
However, I design the entire game around the fact that cheap GPUs can handle, let's pick a random figure, 100 high detail units easily, and so can cheap GPUs. So I have a balanced load across CPU and GPU. Roughly 50% utilization in common circumstances on common computers, since the entire game was designed with the real-world limitations in mind, not fantasy "well if I go with 100 polys per unit I'll have lots woot!".
(BTW, the last chart on the bottom right says "low poly stuff". I ran out of room!
)"But Caydr," you say, "You've just pwnt yourself! You're remodelling TA units with like 1000 polies apiece! Holy COW!" "Ah, my sexy young steed man," I respond, "You forget that I'm in the process of remodelling/retexturing about 400 units singlehandedly whilst simultaneously working on 5 mods that nobody believes exist due to the fact that I'm working on 5 of them whilst remodelling/retexturing 400 units singlehandedly. Do you figure I'll be done those by year-end?"
I am after all a talentless hack. It'll probably be a few years at least. Given that even now you can buy a GF8 which is capable of rendering - easily - 500,000 polygons, without breaking 60 FPS, it's my belief that by the time I finish, such GPUs will be old technology and anyone will be able to handle 1000-poly units (1000x500=500,000), except for the same group of geriatric stick-in-the-muds that have Pentium IIIs who will probably never upgrade, spending their working hours doing nothing but complain about how their prebyzantine computers cannot handle all the unnecessary eyecandy in Windows 98.
Roflies, remember how I'm an idiot and GEM will never work because there's too many moving pieces and all my tests involve static units? The battle I did last week involving 300 8,000-triangle ships each armed with 16 independently-targeting turrets, 32 barrels, and rotating radar dishes, with 400 1,000-triangle fighters buzzing around, with all features except shadows (which look crap on large models) turned on, which produced a battle with a low limit of 20 FPS and took over 10 minutes to finish, on a 20x20 map, while I accidentally forgot to shut down the resource-hogging Azureus bt client, which was apparently downloading at about 150 KB/s, seems to have disproved that. Especially since the entire game has been designed around the fact that you'll see perhaps 30 of them in a battle, tops.
truths,
in terms of RMD (rapid mod development
) and using your time efficiently, caydr has some really valid points here.
However sometimes people just like to optimize, whether its modelling or coding or whatever. Making one thing really perfect is often more satisfying than making 3 just as functional things that have imperfections.
in terms of RMD (rapid mod development
) and using your time efficiently, caydr has some really valid points here. However sometimes people just like to optimize, whether its modelling or coding or whatever. Making one thing really perfect is often more satisfying than making 3 just as functional things that have imperfections.
err.... if we're talking about purtify-ing spring, i agree that model detail and texture quality is quite important
but the tank or bot or whatever has to look good in the first place, so design and artwork really beats out the above
for example, kernel panic is just damn spiffeh imo
another thought of mine is that best bling-for-buck atm might best be achieved by things outside of modeling
once you get to a certain point, it really doesnt matter that you can make out the serial number etched on to the back of a peewee's head, as your view of the battlefield is gonna be pretty distant while playing anyways
stuff like realistic/cool animations (for movement, and firing, and death and such) and souped up graphical effects would leave a much greater impression
i'm talking about things like weapons effects, explosions, and environmental effects, like tracks and dust clouds
i really hate the way flashes "wiggle", "vibrate", and zigzag as they traverse bumpy terrain which is FREAKING LAME, and zpock and argh's suspension animation work is a good example of what i would like to see in future unit animation
i remember that in OTA, smoke (from damaged units, geos, fires, etc.) would drift in the direction of the wind along with the wind farms, something which is sadly lacking in spring
imagine the particulate cloud from a nuclear blast being slowly cleared by wind, showing the scarred terrain beneath
it's these little touches and details that leave a far greater impression on me of the polish of the game than the thread topic
imagine weapon dependent death animations, where a tank would simply stop and sit still when it was pierced by a laser, or be disintegrated spreading wreckage in a spray over the ground by a large explosion
the possibilities for a better-looking spring are endless outside the realm of model detail (though i have no clue as to how easy it would be to implement with the current engine)
but the tank or bot or whatever has to look good in the first place, so design and artwork really beats out the above
for example, kernel panic is just damn spiffeh imo
another thought of mine is that best bling-for-buck atm might best be achieved by things outside of modeling
once you get to a certain point, it really doesnt matter that you can make out the serial number etched on to the back of a peewee's head, as your view of the battlefield is gonna be pretty distant while playing anyways
stuff like realistic/cool animations (for movement, and firing, and death and such) and souped up graphical effects would leave a much greater impression
i'm talking about things like weapons effects, explosions, and environmental effects, like tracks and dust clouds
i really hate the way flashes "wiggle", "vibrate", and zigzag as they traverse bumpy terrain which is FREAKING LAME, and zpock and argh's suspension animation work is a good example of what i would like to see in future unit animation
i remember that in OTA, smoke (from damaged units, geos, fires, etc.) would drift in the direction of the wind along with the wind farms, something which is sadly lacking in spring
imagine the particulate cloud from a nuclear blast being slowly cleared by wind, showing the scarred terrain beneath
it's these little touches and details that leave a far greater impression on me of the polish of the game than the thread topic
imagine weapon dependent death animations, where a tank would simply stop and sit still when it was pierced by a laser, or be disintegrated spreading wreckage in a spray over the ground by a large explosion
the possibilities for a better-looking spring are endless outside the realm of model detail (though i have no clue as to how easy it would be to implement with the current engine
)-
BlackLiger
- Posts: 1371
- Joined: 05 Oct 2004, 21:58
-
SwiftSpear
- Classic Community Lead
- Posts: 7287
- Joined: 12 Aug 2005, 09:29
In terms of the modeling process... the UV mapping and texturing are probably the more involved process as apposed to throwing together the 3D mesh. Especially for spring where a 1000 poly 3D mesh is considered high end. If you laze on the modeling it will make the texturing alot harder. It's relatively easy to paint faux details on a flat surface and have them look good, it's much harder to get your head around making it look right for a dynamic surface.jcnossen wrote:truths,
in terms of RMD (rapid mod development
However sometimes people just like to optimize, whether its modelling or coding or whatever. Making one thing really perfect is often more satisfying than making 3 just as functional things that have imperfections.
Plus, we just hit the current gen tech recently. Less then 10% of our community will have been upgraded to gaming PC standard high end right now... Caydr's graphs are full of BS. You need to find the middle ground between obsessive optimization and lazy, at least to the point where it's efficient still. We're not even close to the point yet where you can just do it however you want that you believe will save you a few minutes and everyone will be happy with it.
part of the problem with poly waste is it often accompanies sloppy geomitry along with inconsistent detail levels throughout the model. No one here is arguing that the models need to be low poly. I think what most of us are saying is that if you want to use more polies fine but use them when needed for details and not jus 'cause. Sure use all the polies you want but don't expect people to get excited that there is an inset then an extruded making up those extra faces. That generally is met with frowns from experienced modelers hell even novice modelers like myself.
Use the extra polies in overall detail.
since caydr's models are the topic of discussion. He has ships with vast smooth hulls that have little guns sticking out like golf flags. It is inconsistent and you are left thinking, why did you not make some intruded faces for that window? Why is there no cargo or loading hatch intruded? Overall you are left with something that feels out of place.
Think of it this way...
imagine a map with sparse little super detailed people, and large blob like trees.
this will leave you thinking more could have been put into the trees. I am doing a bad job of getting the point across. Perhaps someone else could express it better.
Use the extra polies in overall detail.
since caydr's models are the topic of discussion. He has ships with vast smooth hulls that have little guns sticking out like golf flags. It is inconsistent and you are left thinking, why did you not make some intruded faces for that window? Why is there no cargo or loading hatch intruded? Overall you are left with something that feels out of place.
Think of it this way...
imagine a map with sparse little super detailed people, and large blob like trees.
this will leave you thinking more could have been put into the trees. I am doing a bad job of getting the point across. Perhaps someone else could express it better.
optimization is best suited for units that are built in groups of 5271346, so a flash is a pretty damn good unit to have optimized
more polys does not necessitate better looking, although there is a correlation
save ridiculous detail/unoptimized for shit like sumos where a group of 15 is a huge ass group, or for commanders where usually there's only one per team per game.
more polys does not necessitate better looking, although there is a correlation
save ridiculous detail/unoptimized for shit like sumos where a group of 15 is a huge ass group, or for commanders where usually there's only one per team per game.
Caydr is completely right in that optimizing pathfinding or other CPU bound stuff is almost guaranteed to give you more performance increase, while reducing polies doesn't really help much on modern CPUs.
At the same time I think texturing is still more important then just throwing lots of polies at a model. If you can texture a 100 poly model very well, but you make a crap texture for the 1000 poly variant of the model, then the 100 poly model WILL look better, no matter how much more polies you throw at the 1000 poly one. (And I don't texture but I think it gets harder / takes longer if you have more polies?)
Usually you don't see all those polygons anyway because of the zoom level and the fact that most people use overhead camera. (Heck if we had good LOD they wouldn't even be rendered.)
At the same time I think texturing is still more important then just throwing lots of polies at a model. If you can texture a 100 poly model very well, but you make a crap texture for the 1000 poly variant of the model, then the 100 poly model WILL look better, no matter how much more polies you throw at the 1000 poly one. (And I don't texture but I think it gets harder / takes longer if you have more polies?)
Usually you don't see all those polygons anyway because of the zoom level and the fact that most people use overhead camera. (Heck if we had good LOD they wouldn't even be rendered.)
My only beef is that you cant justify something because the msot expensive video card of the day can do it. I buy a new PC that cna do it I've still got the old one sitting gathering dust. Ideally the not so out of date computer should be able to play against the uptodate one.
Core 2 duo and GF8 class cpu/gpu may be ideal but assuming all PCs will have them from now on or better is foolish. Integrated graphics shifts far greater quantities than graphics cards, and there are numerous people on sprign with these chipsets. Should we blame them for not having our wisdom on GPU technology? Should we blame them for having centrino 2 and AMD class cpus? Or being taken in by the dell PCs with the 7900s?
What about the guy who cant play with your super details because he hasnt an 8800 and has a slightly substandard system that only just misses the min spec, but he cant buy or upgrade because he's ind ebt and lives off a low income?
And what about the GF8800 and radeon HD2900 users suffering from repetitive BSOD and TDR crash syndrome?
When your finished making GEM 8800 drivers should be standard stable drivers. Well all be harping on how polycount isnt a direct issue its shaders. There'll be a chorus of people wondering why their not so old PCs cant run some mods getting disgruntled and leaving, and therell be a mass of elitist peopel sayign haha look at me!!!
Supreme commander followed your view caydr, and there are times when our lobby has more people than theres. OTA(not including us or 3DTA) has a larger community than Supcom.
As somebody said, the excessive detail is OK but it sint high poly models that make an RTS good, its houghtful details that make the difference. Wind blowing the trees, flames setting units on fire, better AI, more unusual tactics, environmental effects.
I get the impression now that you just want to argue.
Core 2 duo and GF8 class cpu/gpu may be ideal but assuming all PCs will have them from now on or better is foolish. Integrated graphics shifts far greater quantities than graphics cards, and there are numerous people on sprign with these chipsets. Should we blame them for not having our wisdom on GPU technology? Should we blame them for having centrino 2 and AMD class cpus? Or being taken in by the dell PCs with the 7900s?
What about the guy who cant play with your super details because he hasnt an 8800 and has a slightly substandard system that only just misses the min spec, but he cant buy or upgrade because he's ind ebt and lives off a low income?
And what about the GF8800 and radeon HD2900 users suffering from repetitive BSOD and TDR crash syndrome?
When your finished making GEM 8800 drivers should be standard stable drivers. Well all be harping on how polycount isnt a direct issue its shaders. There'll be a chorus of people wondering why their not so old PCs cant run some mods getting disgruntled and leaving, and therell be a mass of elitist peopel sayign haha look at me!!!
Supreme commander followed your view caydr, and there are times when our lobby has more people than theres. OTA(not including us or 3DTA) has a larger community than Supcom.
As somebody said, the excessive detail is OK but it sint high poly models that make an RTS good, its houghtful details that make the difference. Wind blowing the trees, flames setting units on fire, better AI, more unusual tactics, environmental effects.
I get the impression now that you just want to argue.