AATA Beta 0.9

SinbadEV · Post by **SinbadEV** » 21 Mar 2006, 20:24

the problem with Nano-Blobs slowdown is because the game requires the spamming of hundreds of units... almost all of which move around... you get slowdowns when you give an order to 100 cons at the same time... or send in a fleet of 50 planes... just happens

Drone_Fragger · Post by **Drone_Fragger** » 21 Mar 2006, 23:08

BTW, Caydr proved taht having high or low poly moderls doesn't affect game speed. Its the number of units that matter, not the tris.

Nemo · Post by **Nemo** » 22 Mar 2006, 00:59

This is true - pathfinding is the huge hog, not model rendering.

Then again, every bit of performance saved helps, because it means my computer has a little more power to deal with smoke. (A commander explosion brings me to 3-4 fps)

I'd say that take whatever the models need to be pretty, but try not to waste things (hollow barrels for each tank would be a bit much, for example)

Argh · Post by **Argh** » 22 Mar 2006, 04:03

@FLOZI, etc., etc. (gawd this is long):

1. Yes, people are probably lagging, in NanoBlobs, especially with the first beta, where the texture resolutions are as high as they are. If they're lagging with this last release, though, then their computers must suck pretty hardcore. Or they have all the particles and other crap turned all the way up, instead of at the midrange. Otherwise, I don't get why they're having problems, honestly.

I only have a 2500Mhz, 32-bit processor, a gig of 400Mhz DDR... and I run about 25% in multiplayer with full-on giant war going on, while hosting in this new Spring beta- they've obviously cleaned up pathfinding quite a bit, and it really shows! I am not making this mod for the low end of performance, though- if you don't have the power to handle the Blobs, you can always turn your settings down. I get over 350FPS when I turn everything down, and I get around 40 with FSAA 2X, AF 4X, shadows, reflectivity, and dynamic water. In short... if you lag, you probably should upgrade your video hardware to a 6600GT or thereabouts. They're cheap now.

That said- none of my models are over 1600 tris! Even the Knights, which have modeled, animated treads! I did not waste tris- I am, if anything, inclined to cut down tricounts whenever possible, and am a very ruthless modeler. For example, that Spire Rook looks really hi-poly, but it ain't. I used a 16-sided column (that'd be 32 tris) to make the main body section, another for the cone top, and bottom, used some cubes for the legs, etc. It's really not hard to achieve lots of detail without insane tricounts.

2. I would disagree pretty strongly with people who say, outright, that "polycount doesn't matter". This is a mistake, on a variety of levels:

A. Polycounts always matter, because even if the engine is efficient, and culls polygons that aren't being seen by the viewer, the engine still must do the culling in the first place! This is a fairly big step in the rendering phase, so polycounts matter here.

B. Polycounts always matter, when reflections or shadows are involved. While modern GPUs can do these transforms on the chip, requiring little/no CPU, the GPU still has to have the meshes, their UV normal coordinates, and other crap sent to the GPU. This is non-trivial, especially with shadowmaps.

So... don't say that "polycounts don't matter". A more accurate way to say things is, "if using S3O, you can minimize the amount of suck associated with higher-poly models".

I know that this whole S3O/uvmapping thing is taking a long time to sink in and become accepted by mod-makers for Spring- but let's look at things another way, for a second. If I'd gone out've my way to make NanoBlobs for high performance... I could have models with half the tricount of 3DO models which would have as many faces, come render time, and I could've kept the textures to a lower resolution (I suspect that you can make fairly good-looking skins at 256/256 for RTS units, if you're clever about it- there are plenty of older games that prove that is true) and had better performance than 3DO... without sacrificing anything.

But, all that said... 2500 tris, in an RTS, is excessive. Don't use that many tris unless you're building something really special, not something you're spamming. A really nice tank can be had for 500-800 tris, especially if you're not getting uber-fancy and making animated treads. I think that the Knights are maybe 350 tris before the treads, and they still look pretty detailed.

Also... the skins are everything. I mean that. They're what makes or breaks a unit, period. People who make high-poly models with crappy skins are amateur-night modelers. The pros know that the animation and the skin are most of what makes a model really shine. I have put up links to my uvmapping tutorials... just go through them with a simple tank, and you will find out that uvmapping models, especially boxy things like tanks, is actually quite painless and easy. I could do a reasonably realistic tank very quickly- it's really not that hard to find reference shots and work from them.

Felix the Cat · Post by **Felix the Cat** » 22 Mar 2006, 04:38

In response to Argh's NanoBlobs lag comments:

Here are my system specs:
-Intel Pentium 4 3.33 GHz
-512MB RAM
-ATI Mobility Radeon 9000 IGP
-128MB of RAM for it - haven't been able to figure out if it's directly interfaced with the GPU or if it's system RAM that's permanently assigned to graphics processing.

All sliders are at the mid-to-low range; no "extras" are turned on except for Fullscreen and Colorized Elevation Map. 1280x800 in-game screen resolution.

In NanoBlobs, during an intense battle, my FPS drops to the 4-8 range. Same if there's a commander explosion.

Do I have a crappy system in your estimation, or is there something wrong here?

Argh · Post by **Argh** » 22 Mar 2006, 04:58

-ATI Mobility Radeon 9000 IGP

Ok, that, by itself, is a major part of your problems. That's like... a weakened, slow GPU version of the crappier GeForce 4000+ series... online, I saw it compared with a 4200Go, which was weaker than a Geforce3, let alone the Geforce 4600Ti, which is what I had when Counter-Strike was one year old.

1280x800

I assume that's native rez, on a laptop LCD. Try lowering that.

-Intel Pentium 4 3.33 GHz

Look... my Athlon XP 2700+, running at a mere 2.5Ghz, runs circles around that thing- in real-world tests, it's about 30% faster, even cutting memory and other factors out've the picture. And even a weak Athlon 64, the current standard for hardware, eats my poor chip for lunch

And I doubt if you have DDR in that thing- it's probably still using SDRAM, given the age. So... you basically have a very slow video card, by modern standards... a slow processor... half the memory I have... a slower-bus motherboard... and everything's even slower than it would be, in a standard box, because the laptop engineers have to keep it all from draining the batteries too fast, or getting too hot.

In short... you're trying to tell me that a 3+ year-old LAPTOP is not a crappy gaming rig? Gimme a break

FLOZi · Post by **FLOZi** » 22 Mar 2006, 05:14

Argh wrote:@FLOZI, etc., etc. (gawd this is long):

But, all that said... 2500 tris, in an RTS, is excessive. Don't use that many tris unless you're building something really special, not something you're spamming. A really nice tank can be had for 500-800 tris, especially if you're not getting uber-fancy and making animated treads. I think that the Knights are maybe 350 tris before the treads, and they still look pretty detailed.

Also... the skins are everything. I mean that. They're what makes or breaks a unit, period. People who make high-poly models with crappy skins are amateur-night modelers. The pros know that the animation and the skin are most of what makes a model really shine. I have put up links to my uvmapping tutorials... just go through them with a simple tank, and you will find out that uvmapping models, especially boxy things like tanks, is actually quite painless and easy. I could do a reasonably realistic tank very quickly- it's really not that hard to find reference shots and work from them.

So, basically, at the end of the day, you agree 100% with my opinion, in regards to this mod which is what it was aimed at.

So please, no need to lecture all us mere mortals on the delights of S3O, we know! hence why we hope to go in that direction too. It doesn't need time to 'sink in' and frankly, i find it rather insulting that you think we are all so slow-witted.

Argh · Post by **Argh** » 22 Mar 2006, 05:37

First off, a quickie article, in regards to the 9200 Mobile vs. other GPUs:

http://www.notebookreview.com/default.asp?newsID=2526

And, um... no insult is intended, FLOZi. I'm just kind've frustrated with how many times I've had to repeat the same darn things over and over again, in regards to the performance advantages of S3O vs. 3DO. Every time I point towards the things we know, people are like, "but"... and I'm getting grumpy about it.

Maybe I should deliberately make a low-poly, low-skin-size mod for my next project, to end this nonsense once and for all. After all... I was thinking I'd do a psuedo WWI thing... so I could just make it deliberately ugly and then make performance the goal...

Warlord Zsinj · Post by **Warlord Zsinj** » 22 Mar 2006, 05:56

Note that the 2500 polies is after converted to triangles. It was about 1500 quads.

FLOZi · Post by **FLOZi** » 22 Mar 2006, 12:09

Argh wrote:First off, a quickie article, in regards to the 9200 Mobile vs. other GPUs:

http://www.notebookreview.com/default.asp?newsID=2526

And, um... no insult is intended, FLOZi. I'm just kind've frustrated with how many times I've had to repeat the same darn things over and over again, in regards to the performance advantages of S3O vs. 3DO. Every time I point towards the things we know, people are like, "but"... and I'm getting grumpy about it.

Maybe I should deliberately make a low-poly, low-skin-size mod for my next project, to end this nonsense once and for all. After all... I was thinking I'd do a psuedo WWI thing... so I could just make it deliberately ugly and then make performance the goal...

Well I was reading it at 4am, tends to make me grumpy.

Just that you first criticise me saying that 1-2k is to many polies (for this mod) and then pretty much agree.

Felix the Cat · Post by **Felix the Cat** » 23 Mar 2006, 02:10

Argh wrote:
In short... you're trying to tell me that a 3+ year-old LAPTOP is not a crappy gaming rig? Gimme a break

I bought the thing in October.

Regardless. Now we are required to have a gaming rig to play Spring? Perhaps you don't realize that some of us aren't rich (or, more likely, spoiled children of rich parents). It was all I could do to afford what I have - everything else in the price range had an Intel integrated graphics processor, something with which I had bad experiences in my last computer.

Ah well; I guess Spring is now a rich kid's toy. Oh well. (Hint: it could be easily fixed by allowing us to turn off smoke...)

Argh · Post by **Argh** » 23 Mar 2006, 02:22

Hey, now... I built my rig with my own money, and I am an adult with a real job. And it's not some uber-rig that cost $5,000, either- I built the entire thing for $1200 or so, and just upgrade a part at a time- I maybe put $450 into it every year, and lag about a year behind current technology, because I wait until it's cheap before upgrading. I'm not rich, and I have to budget, like a grownup.

So don't go around implying I'm some teenybopper brat or trust-fund baby

If you bought a laptop because you needed a mobile rig, that's fine, but don't say things like that- it's not nice or accurate. With the specs you cited, you probably have terrific problems with any modern game, like HL2, etc. Why should Spring modders cater to you when you don't represent mainstream specs?

Nemo · Post by **Nemo** » 23 Mar 2006, 02:48

Well...I think I really do have a crappy setup, and old to boot, and I can run spring on minimum settings without issue, in any mod (nanoblobs included), but smoke is the big performance killer. A commander explosion or any large amount of smoke near the screen drops me to 2-3 fps, but I run at 25+ fps otherwise (which is plenty playable)

Stats:
768mb SDRAM
1.8ghz Celeron
Geforce mx 400 (with no less than 64mb of kickass memory..<_<)
Screen res is 1024x768.

Dell dimension 2300, if it makes a difference what package it was.

As for other games, I ran CoD2 smoothly on midrange settings (again, smoke was an issue), UT04 without a hitch on low end settings, and BF1942 (no longer a standard, but it used to be the nasty one for rigs) on medium settings, since it runs a mite slowly on max ones.

I don't mind if I'm not what mods are designed for, but I'd be a bit annoyed if a mod was unplayable on minimum Spring settings on a computer like mine.

Argh · Post by **Argh** » 23 Mar 2006, 02:59

So, do you turn particle effects way down? I dunno what, if anything, I can do about that, especially since S3Os don't even obey standard Spring explosion script calls yet... if this is a problem now, then it's likely to get worse, and I should see if there's anything I can do to mitigate it. However, until/unless the SYs start giving modders more control over the FX sequences, the best way I can "control" things is by making the units simply spam a lot slower, which would kind've defeat the purpose and totally change the game design.

Nemo · Post by **Nemo** » 23 Mar 2006, 04:20

nanoblobs is playable - things start slowing down in lategame when there are a bazillion units around (sooner, of course, on larger maps or larger games), but other than that its fine.

I turn everything that will go down, down. I've had decals on recently because I worked on them for AATA, and was pleasently surprised that they didn't have much of an effect on performance (just a bitmap being applied..) and really added to the visuals.

Hopefully by the time the real next gen mods start rolling out (A&AS proper, the next mod that your crew works on, among others), I'll have upgraded, since I imagine spring is mighty pretty with all the bells and whistles turned on.

I imagine that Caydr's mod will run fine, despite very large polycounts, since it focuses on smaller numbers of units, since pathfinding for 300+ units at once, or large smokey explosions are really what make the game crawl for me.

In other, more AATA releated news:
I'm plugging away at infantry, trying to find a good balance for realism and gameplay with health and such, since soldiers have a strangely tough time taking cover in Spring <_<. Who would have imagined, giving units correctly scaled weapon ranges doesn't work at all, since a single rifleman can shoot most of the cold place (3/4ths of the map is in range with realistically scaled weapon ranges), nevermind an artillary piece or tank gun. We have a new scheme for tank armor and damage worked out, and the infantry weapons work (converting them from OTA shared weapons between teams to single, specific ones) is well on its way.

Argh · Post by **Argh** » 23 Mar 2006, 05:44

One way that you could simulate cover, pretty easily, is with the accuracy tag. The other, of course, is to have infantry be able to cross the kinds of terrain that they actually can, which is considerably steeper than anything wheeled can go (although some WWII tanks could do fairly steep slopes).

Also, just making infantry with true, long-range weapons (here, I'm talking specifically about battle rifles- anything autofire from WWII was firing pistol rounds, other than the '44, and should have pretty short ranges) have ballistic arcs would greatly simulate the real-world effects of firing battle rifles at people at the real ranges they can be used. Also, real-world scales of speed (i.e., the meters/sec, minus a fudge number to account for atmospheric scrub) would do wonders for things.

IRL... you rarely hit at the theoretical ranges of such weapons. Just lowering their range, though, quickly makes combat not true scale, and it would defeat a lot of things. Because if this is WWII, and infantry can charge across open fields of fire against machineguns, I am not going to play this, and will make a competing mod

In the real world, WWII infantry tended to die in droves, in open-field engagements (which, naturally, they tended to leave to armor and aircraft to decide). Most serious infantry engagements were either in closed terrain, such as forests or cities. There aren't a lot of maps for Spring that really do a good job of giving us either type of terrain... but one easy fix, short of making custom maps, is to make tanks utterly unable to run down trees. Which, in the real world, they DO NOT DO- do not believe idiots who tell you otherwise, they have not read about combat in the Bocage, or been in a real tank. Saplings, yes. Full-grown trees, no! So just assume that Spring's trees are full-grown

That'd give you a fairly large number of maps where infantry's real-life advantages would matter a lot.

Another thing you can do- and this sounds weird, but isn't, really... is take full advantage of how mobiles can now build mobiles... and make a series of "transports" that have a move-rate of 0, turn rate of 0, etc., etc. that units could shoot out've. What do you use these for? Why... bunkers, of course! Between that, and providing soldiers with some sort of "entrenching device" (say, a "building" that can be built fairly quickly, then self-destructed to leave a nice, neat crater)... you could very quickly set up things so that infantry could get nice and hidden over a reasonable time period, without being unrealistic.

Nemo · Post by **Nemo** » 23 Mar 2006, 06:57

Argh wrote:One way that you could simulate cover, pretty easily, is with the accuracy tag. The other, of course, is to have infantry be able to cross the kinds of terrain that they actually can, which is considerably steeper than anything wheeled can go (although some WWII tanks could do fairly steep slopes).

Well, yep. We might need to turn down the tanks a touch, but infantry can climb anything short of vertical.

Also, just making infantry with true, long-range weapons (here, I'm talking specifically about battle rifles- anything autofire from WWII was firing pistol rounds, other than the '44, and should have pretty short ranges) have ballistic arcs would greatly simulate the real-world effects of firing battle rifles at people at the real ranges they can be used. Also, real-world scales of speed (i.e., the meters/sec, minus a fudge number to account for atmospheric scrub) would do wonders for things.

Ballistics could help, but we use lasers as bullet fire - the tracer effect, and although you end up with a silly number of tracers, it makes battles easier to follow (and having bunches of spring default blobs fly around isn't all that great). The issue with real world numbers based on the scale we're using (which is already on the small side) is that a single guy with a K98K could hit 3/4ths of The Cold Place. As far as speed, I think we use a standard weapon speed for all bullets, since anything faster would start to break collision detection with our tiny units.

IRL... you rarely hit at the theoretical ranges of such weapons. Just lowering their range, though, quickly makes combat not true scale, and it would defeat a lot of things.

We take our weapon range stats from practical ranges where ever we can - obviously you could fire a M1 garand pretty damn far and still be accurate in a competition or shooting range, but in a combat situation its a fair bit different.
It was my idea to try realistically scaled weapons range - Flozi and Spiked already had a very workable system, abiet one that doesn't fit if you look *really* closely at things. The sheer spread of weapon range (from a M1911A1 Colt .45 to the 155mm M1 Long Tom) doesn't allow us to use a realistic scale, since anything beyond a 37mm gun would be able to hit the entire map on most maps.

Because if this is WWII, and infantry can charge across open fields of fire against machineguns, I am not going to play this, and will make a competing mod In the real world, WWII infantry tended to die in droves, in open-field engagements (which, naturally, they tended to leave to armor and aircraft to decide).

Don't you worry - infantry die in swarms, depending on how they're used. This is where cover comes back into play. Since cover is nearly impossible in Spring (even on an open field soldiers would go prone), they are what I've been calling Call of Duty soldiers - Ie, they're a bit tougher than realism would dictate. Realistically, any soldier that gets hit once by anything else is no longer going to be fighting. Gameplay wise, that leads to 100 infantry dying in three and a half seconds, and negates their usefulness almost entirely. So, we up that timeframe to perhaps 6 or 8 seconds. The damages go something like this currently (may vary from side to side, depending on equiv. weapon): Sniper=100% damage. Bolt action rifle = 66% damage. Carbine/semi auto rifle = 45% damage. Assault rifles and other LMGs = 38%. Heavy MGs (.50cal, MG42) = 75% damage. Pistols = 60% damage.

This still leads to infantry dying *really* fast to direct fire, but lengthens infantry vs infantry combat a bit, and makes them more viable.

Most serious infantry engagements were either in closed terrain, such as forests or cities. There aren't a lot of maps for Spring that really do a good job of giving us either type of terrain... but one easy fix, short of making custom maps, is to make tanks utterly unable to run down trees. Which, in the real world, they DO NOT DO- do not believe idiots who tell you otherwise, they have not read about combat in the Bocage, or been in a real tank. Saplings, yes. Full-grown trees, no! So just assume that Spring's trees are full-grown That'd give you a fairly large number of maps where infantry's real-life advantages would matter a lot.

Check. No tanks knock down trees, as far as I know. Hiding infantry in them is very worthwhile, especially since there's no TA style radar equiv, so if they go unspotted, they're fine. There's a definate risk to this though, since I hid a handful of AT troops in a forest in a game the other day, and my opponent just fired a shell at a single tree. PFOOOF. Up they all go in flames, and my infantry with them. I laughed =)

Another thing you can do- and this sounds weird, but isn't, really... is take full advantage of how mobiles can now build mobiles... and make a series of "transports" that have a move-rate of 0, turn rate of 0, etc., etc. that units could shoot out've. What do you use these for? Why... bunkers, of course! Between that, and providing soldiers with some sort of "entrenching device" (say, a "building" that can be built fairly quickly, then self-destructed to leave a nice, neat crater)... you could very quickly set up things so that infantry could get nice and hidden over a reasonable time period, without being unrealistic.

Well, the German side already has several bunker buildings (command, barracks, resource) which are the very few buildings that can survive a direct hit from an aircraft bomb. That's an interesting idea for mobile entrenchment though - as a matter of fact, buildings can load units in Spring, so it wouldn't even have to be mobile.. If Isairbase=1; works again, they could even fire out, almost starcraft bunker style. I'm not sure it would fit with AATA (since infantry are super-expendable), but its a very interesting idea.

SpikedHelmet · Post by **SpikedHelmet** » 23 Mar 2006, 07:28

The problem lies with the fact that in-game, individual units do not know how to use the cover that can be provided via trenches and so forth. Sure you can blow a crater in the ground and pop infantry into it but most of the time it won't have any difference.

The best Spring can do is WW1-style mass rushes of bodies and metal. Unfortunately.

Argh · Post by **Argh** » 23 Mar 2006, 07:45

If only we could give units different speeds depeding on script states, we could fix a lot of this. Then prone soldiers could have a different movement speed, and through scripting tricks, you could have a different rate of movement- and different weapons performance, by using another weapon, to simulate the real-world advantages of prone/kneeling/standing fire.

But the AccuracyWhileMoving tag, all by itself, might be quite able to fix a lot of this.

As for the use of LineOfSight weapons with laser beams... check out the shots from my Wolf unit in NanoBlobs. I used a low-poly "shell" with a very strong glowmap, and it works great! Just a thought. Overall, though, this is sounding like it'll be a stronger and more realisitic experience than the previous "historical" mods and RTS games I've seen, which is what I'm hoping for. I don't want a tac-sim... there are plenty of those out there, and for the most part, they're dreadfully boring experiences. But something that has the feel of WWII, where tank designs became obsolete overnight, but infantry did not, would be most interesting to play.

Nemo · Post by **Nemo** » 23 Mar 2006, 08:11

SpikedHelmet wrote:The problem lies with the fact that in-game, individual units do not know how to use the cover that can be provided via trenches and so forth. Sure you can blow a crater in the ground and pop infantry into it but most of the time it won't have any difference.

The best Spring can do is WW1-style mass rushes of bodies and metal. Unfortunately.

That said, hiding tanks behind wreckage works very well, and digging foxholes for infantry works equally well, but you have to use ~3 AT mines to do it.

I'll see if I can create a worthwhile tracer bullet, since that would be pretty neat.

Edit: Or not. Here are what the Wolf firing and its projectiles look like for those of us who can't handle shiny glowy reflectiveness.

I still think that turning bullets into ballistic weapons would be good, but I'm not the person to be making a texture for them, since I can't even test how they look in game >_<.

Spring RTS Engine

AATA Beta 0.9