limit avatars to 1000 prims

apples and oranges. 3d prerendering has much more leeway in optimization, and now - uses of bumpmaps, shaders, etc that SL doesn't have.

Indeed, we could strip everyone's avatars down. For such a change to be widely effective, it would have to, by definition make a large volume of existing content obsolete. Otherwise, the results would be negligible.

I do understand that the exciting prospect of "ruining some peoples day" on such a widespread scale makes some folks here positively giddy, People here who are shouting out limits and numbers base their 'limits of what is ok and what isn't' solely on *their* preferences.

There's nothing guaranteeing that "limiting avatars to 1000 prims" will do a thing except -- limit avatars to 1000 prims. I don't see any solid evidence that the world's framerate would go up.

Furthermore, I have a hard time believing that it's the only way to increase client framerate. Maybe the easiest, dirtiest, and quickest -- and maybe a method that'd make some people here feel 'big' and vindicated, but it's like using a shotgun to unlock a door. Yeah, it'll do the job, but it'll also do a lot of damage and maybe hurt someone in the process.

To some, I guess, that's their style anyhow.

Note that there are folks who use the default Linden avs who probably think that 50 attached prims are excessive.

Not at the whim of the individual exactly, but the possibility should exist. It exists with land, to a very large extent - buy a sim, buy four sims and put them together, whatever.



To an extent, prim limits are arbitrary, yes.



Actually, it's nowhere near that simple.



Logic. My personal avatar is far less than 1000, or even 500, prims.

QFT. Exactly.

Eltee, as to your post reguarding system performance: My system is about two years old now, it's a P4-3ghz, 1 gig of DDR ram, and an nVidia 5900. This is NOT a powerhouse system by modern standards - it wasn't even THAT great when it was new. I hang around luskwood all the time, with furries all around, decked out in jewelry, hair, weapons, shoes, you name it, and I still get ~8 fps or better most times. And as you say, while this may not be silk, this isn't a twitch game - WHAT I see is far more important than how fast I see it.

By the time you get enough furries in one spot to lag me to the point it's unplayable, as long as the sim around me isn't terribly designed, there are enough avatars under all the prims that I would be lagging anyhow.

LOL sometimes I just have to chuckle.

While you all are busy patting each other on the backs for making the same points already made-- by the opposition-- several times in fact-- have you stopped to consider that every bit of lag reduced on SL helps reduce overall lag? Or have you considered that relatively very few people on SL wear avatars containing over 500 prims? Or that if SL passed a "500 prim limit"... very few avatars being sold would even be affected.

I'm not saying there won't be any loss at all. Every major change requires some compromise (we have seen dwell and stippends both go that route, despite some major whining). The benefit is the gain from such a move. If that gain is not sufficient... then it's a bad move. If it is... I don't think anyone is going to keel over and die because LL puts a 500 prim limit on avatars.

But... as I said before -- at this point in time that would be generally ineffective and a useless step. Because avatars can create significant lag on far less than 500 prims.

So before getting all up in arms and emotional... you might consider the possibility that this concept has some validity. Although it may not happen now, there is a very real possibility that as time passes, LL will eventually make some significant decisions to help reduce lag. That may include placing a "prim weight" rating on avs and limiting the total effect an individual avatar can have on a server. Ranting won't stop that. Exaggeration won't stop that.

LOL, "hundreds of thousands"... why not "billions and billions"??? It's good for a grin, I guess.


On a different note... I did appreciate the comment about the total triangles on a sphere vs a cube vs a torus. (Where do folks find that info anyway?). Still, not being an extreme graphics tech myself, I have only one comment to make: I've never seen a set of wings crash a sim. I *have* seen a single hoochie hairdo lag a sim to a standstill, and I have a friend that has a wig she personally made that was a sim crasher. I've seen sims crashed by prim activity. (I think that's been reduced to an extent over the past several months). So.... tech and stats set aside... there appears to be a significant difference between prim wings spheres and hoochie hair torus. I don't know exactly what that is, but whatever it is... it glitches the dickens out of both server and client systems.

I would be curious as to how much such things still do so. I know some alterations have been made in how SL controls graphics (such as culling). But when it comes to general lag and graphic rendering, I don't really need to know the precise tech reasons and specs; I can see the very real effects and results and through simple process of elimination narrow things down to what in general is causing the problem. That's simple business 101: leave the tech to the techs. Identify the trouble and let them figure out the solution.

I do know one thing: standing around and saying, "OH NO! We don't want to do that!" is never going to solve or accomplish anything. Have we never heard the phrase, "To make an omlet you have to break a few eggs?"

I've seen thread after thread after thread dealing with lag, and there are always the people who say, "I NEVER LAG!" (pardon me, but yeah, right). And there are those who say, "It's all client side!!!"-- while seemingly oblivious to the fact that LL also wrote the client-side software. There are people who talk about their super-mega $3000 computers, or their having only 50 scripts on their whole sim, or "less than 4000 prims on my sim and it runs just fine"... and they seem to have the delusion that such statements actually have significance for the average SL user (like we're all going to go buy a $3000 computer and limit our sims to 4000 prims and 50 scripts). LOL. Some folks just don't get it.

There are always the nay-sayers... and they accomplish absolutely ZIP... because they're not looking for solutions... they're looking to argue on forums.

The only ones who ever seem to spark progress are those who are willing to stir the pot. Arguing in favor of something that's obviously malfunctioning and glitchy is just sad. LOL

Over the past months I and others have made several forum propositions. Those same things were attacked, personal aspersions made, insults proffered... pretty much by the same people who still do so. Emotionalistic response is always present. Yet as the months have passed, those suggestions have been implemented by Linden Lab and those who were against such, vehemently and quite vocally, failed to really have any impact... because they were so busy insisting on their opinions and conducting forum battles-- that they ignored the ever-developing facts.

Bewail all you want folks, if SL is going to work better, things are going to have to be done to MAKE it work better. You can't accomplish that by bemoaning every potential change to the system... or by decrying as bogus every hypothesis that people put forward as possible solutions.

To find out the geometry details, you can grab the scene data from SL client with intercept program and load it into 3d application of choice (very helpful for things like adding radiosity shading to existing SL builds, or precision texture work tailored for shape of one's avatar)

That said, since some hair use more than 50 tori with multiple twists to them, so these can have definitely more lethal performance than the example given. But then i think most of the crashing comes rather from bad code that deals with level of detail/geometry updates and gets really confused when you get close to these tortured prim things... rather than plain triangle amounts. It's hard to tell really o.O;

in my formation i have been teached stuffs about polygon reduction, basically when you are in a game development studio, the coders evaluate what the engine can bear, how open might be the landscapes, how many potential users we might see at a given time, and after this you get an estimation of how many trangles you are allowed to use per character, how many per object, etc

then you on the artist side have to do with it, lets say you have 3000 triangles, you can use 3000 or 2500, maybe 3010 but never ever 3500, if you are told to work with 3000 you will work with 3000, it is what make the difference between an average modeler and a good modeler.

If the game industry impose limits, you might ask why, well because if there is no limit it will be unplayable, sure "sl isn't a twitcher" and other snarky comments like this will fall from some fucking snobs. for an fps player, the framerate they feel allright is 60fps, of course we aren't in an fps, so lets say we want to achieve 25, wich is the movie picture standard for an illusion of movement (some animes go as low as 10 but it really start to show)

okay so 25fps is the target, now lets see... average fps... 10/15 , on a top class computer...
and i am sorry but unless you are masochist, 10 fps in a game is ugly and tiring to look at for your brain.
Solution? lets see a bit the limits in place to prevent abuses:
-limited number of prims per sqm
-limited number of sounds played (if i remember only 8 are mixed)
-limited number of textures visible (based on the number of prim faces)
-avatars: no real limit, exepted max link on each attachment point...

i don't know for you but to me the problem is pretty obvious

of course we will put all the blame on Linden Labs, they have a bad engine, the prims aren't optimized, their 3D coders suck, etc...

HOWEVER unless Linden labs take counseling from id software (quake , quake2,quake3,quake4,doom ,doom2, doom3,...) or Epic (unreal serie)

we have very little chances that this 3D engine perform better, AND EVEN, in all these games, the polygon count is LIMITED.

now other facts:
-most sl builders haven't worked for any game companies
-most builders in sl don't know what is a polygon
-some builders in sl are probably egoist and don't give a shit of the impact of their "art" on the surrounding
-most builders of sl don't know a thing in 3D engines and optimisation

of course, sl isn't an environement of professionals, but with proper limits we would encourage the use of good textures (where someone put fullprims nails and bolts), we would make the "massive" side of the game more enjoyable (higher fps, not needed to drive to the radar in a club)

even if the limit go to 500 and below i won't whine and i will simply try to do the same work i used to do, with less prims

Then why don't you simply work with less prims now, and try to encourage others to do so volentarily, rather than trying to enforce it on others who may actually have a reason to need the other prims?

Comparing this with game development is a flawed analogy - I've been in that field before, albiet in an amature way with game mods. You try to stick to poly count limits, but they aren't /fixed/, unless the engine simply chokes at a certain point. There are guidelines dependant on use: Weapon models have so many polys, "class a" enemies have so many polys, "class b" enemies have so many polys, deathmatch models have so many polys, boss monsters have so many polys, etc. These give a rough idea, but if an artist needs fifty more polys on a weapons model, he does it - then sees if the engine can support it. If the artist needs more polys to flesh out a monster design, it might get bumped to a different class of monster (depending on the situations it will be used in). On a boss type monster, you might rob Peter to pay Paul... go for a very complex boss monster, and make sure you fight him in a relatively simple environment, for example.

None of this is possible with a hard limit - but it IS possible in SL right now. Look at luskwood, for example - relatively few, simple textures and simple geometry keep performance to allow for more avatars. This is the same principle the "black box" clubs use. By contrast, a really intricate place can lag for people even if nobody else is around is around - the set of Golden Spurs, for example.

exepted in sl some peoples think they are entitled to all the rendering power of the client. In SL if there isn't a hard limit, nobody care.

But that isn't something that a simple prim count is going to do shit about. I could lag you with a couple dozen prims if I was a moron (How many 2096x2096 textures would you like today? ), while a hard limit shafts everyone.

Actully, in the long run, I'd far prefer to deal with prims that an inordinate number of textures used to compensate for prims.

That some sensible limitation isn't going to weed out *all* cases of stupidity isn't really good argument the idea should be abandoned altogether. There's no silver bullets, but it does not mean obvious problems should not be addressed at all if just to some degree.

As for "shafting everyone" with hard limits ... there already *is* hard limits -- you can't have more than 7.5 k prims on your avatar due to limitations already present. Does it mean everyone is shafted by it, and we should demand unlimted amount of attachments and/or unlimited number of prims per link set?

Somehow, i don't think so... and so it seems to be more of a question if the already existing limits shouldn't be lowered.

The problem as I see it is, a limit on prims is a limit that cannot be worked around beyond a certain point. If you /need/ a prim to make a gizmo, you need a prim. Textures, on the other hand, you can work all sorts of magic on - and I do, on a routine basis. Combine textures and using offsets, optimizing the size/detail ratio,

Reguarding the already existing hard limits, yes, there are already are limits, but they are mostly of a technical nature - there are X attachment points, and linksets are limited in nature for other technical reasons. That varies, IMO, from an arbitrary limit imposed over top of that.

Thing is, I remember what avatars used to look like when worn prims mattered, back with taxation. In a word, fugly. Even the designers that were capable of doing good work. I willingly accept a degredation of performance to get away from that.

Wait, are you saying you don't think it's possible to make an AV that isn't "fugly" in under 1000 prims, in current days when people have quite a bit more experience in how they make their items? Because i really don't see otherwise how you'd use this particular argument to oppose such suggested limit. o.O; (now that we've established there's already hard limits in place, so the issue isn't as much "should there be ones", but rather "how high should they be"

And well, what you say makes it sound it wasn't actual limits that made people look bad in the old days, but it was unwilingness to pay out of one's *own* pocket for better looks of their AV. As opposed to the convenient "oh just shut up and buy better computers you whiners" that doesn't cost one anything...

Well, I am truly envious of your graphics knowledge. Capturing scene data from a client and loading it into a 3d application for further work? Woot! Would I love to know the specifics of how to do that. Pretty awesome.

And yeah, I agree. I think there are some nasty code bugs in there somewhere (based on observable evidence) and some routines that may need fixed. I think they have actually fixed something somewhere, because it's been along time since I've seen hoochie hair actually crash a sim.

One thing that has seriously changed on SL is affects of avatars. When I first joined, there was no lag noticeable until at least 24 avatars were on a sim at the same time. Even then, you could get up to 30, 40 without serious syrup lag. Now, I often notice lag at 10 avatars... 24 is noticeable... and 40 is guaranteed syrup. This applies to just about any standard sim on the grid. But... there are times when this is not the case. To me, that points to asset servers, communication routines, deep-core stuff. Avatars are definitely the biggest individual customer-created lag agent on the grid. It appears that lag is far more complex graphically than just about anything else... but I think it's more than just graphics. It may be a shared bandwidth issue, excessive demands made on the server due to non-streamlined code, a cache/memory leak, no telling what. But even primless avs can lag a sim like mad. I could see that with 200 avs. But 20? Something is definitely wrong.

I wasn't around at that time... but one must admit system graphics have increased significantly since then. I can absolutely guarantee that given 200-250 prims, I could design gorgeous avatars 'till the cows come home.

On ElvenMyst, we have a very unique merchant area. Instead of renting booths, we rent a 256m piece of land (nice... after pathway allowance, 12x12 instead of the standard 10x10). Merchants are allowed to build their own booths. The only limit: a total of 56 prims for both building and vendors.

You should see some of the amazing things these merchants have built using 56 prims. Bias aside... it's one of the most beautiful markets I've ever seen. Not only that.. it's on 2 levels; each merchant has an above-ground and an underground area interconnected by tunnels. It's awesome. 56 prims.

If avs were limited to 250 prims, we'd have amazing 250 prim avs. I'm not saying that should be the limit, because as pointed out already, a sim can be as effectively lagged with 50 prims as 250. All I'm saying is that people prove every day-- we can do amazing things with whatever tools at hand. I don't think avatars would be ugly just because we were limited to 250 prims. With 250 prims I could build a rocking set of armor and weaponry to boot. Agreed, 500 prims would allow a lot more lattitude with really fancy stuff. And that's where the Catch-22 comes in, because if everyone on the grid wore even 200 prims, the grid would come to a gridlock. The only thing that prevents such already is that most avatars wear very few prims. That covers for the occasional higher-prim av. But man, if everyone got the idea that it's ok to wear 200 prims... we wouldn't be able to move, at all. (Ever been to a club... with hoochie-hair dancers all over the floor?)

Not exactly. It's a little more complex then that.

Back then, nobody made any one attachment very detailed, because it was an added burden to the overall whole. A gun could look great in 100 prims, for example - but people will be reluctant to make 100 prim items, if people aren't going to want to wear them because they have prims commited to something else. So people simplify. How simple an item is damn near becomes it's selling point - and you wind up with really simplistic, boxy stuff that relies on texturing to look halfway decent - which introduces it's own problems.



Some people were willing - and you saw it happen. But it wasn't something that people who sold attachments wanted to gamble on. When you saw really really great prim avatars, they were almost always one-off custom jobs.

It's less complicated than i made it sound, i think ^^;; look up "ogle, openGL extractor" ... this is the 3d scene extractor. The data is saved in Wavefront .obj format which most 3d applications can read. From there, what you can do with it depends just on your 3d tool and the patience, pretty much.

If nothing else, it's useful for quick check just how heavy your AV really is, geometry-wise ^^;

I think we're getting off track here. This shouldn't be about people engaging in possibly antisocial behaviour, because there's no limit to that and nothing Linden Labs can really do about it in the software...

No, where I'm coming from here isn't "should people be allowed to do X" but "what is the best way to make everyone's experience the best".

That means considering things like:

* What is the actual reason why pushing the envelope causes problems for people.

* Where the edge of the envelope actually is.

* Where the best place for the limit is.

My comment about chi-chi hair versus primmy wings, for example, addresses the second point. If the edge of the envelope is purely prim counts, then the idea of limiting the number of prims makes sense. If there's something else involved (the type of prims, for example) then limiting it based on the number of prims will encourage the wrong behaviour... people will use fewer more complex prims that will likely lead to more lag than a larger collection of simpler prims.

Ah, now this is relevent to my first point. What is the actual reason why pushing the envelope causes problems. Do people running heavy-duty machines indeed suffer disproportionally? If a Mac mini running at 10 fps drops to 2 fps when a dragon shows up, and an Athlon 64 X2 with a PCI-E 7600 drops from 50 to 10... that would be a proportional impact. Is that the kind of effect you see? It's not what I see.

I'm saying that you shouldn't restrict them, unless that's the only way to stop them from lagging everyone around them.

I'm not putting the blame on the customer.

I'm not disputing that. We also deserve riches and long life, each one of us... but you don't get them by demanding them, you get your best chance at them by figuring out how, and then doing it. We need, here, to be discussing "how", not simply asserting what we want.

Actually, there kind of is. The problem is that sub-goal (limiting the number of prims) doesn't necessarily lead to the higher level goal (improving Second Life's frame rate) because it can motivate behaviour that is actually going to make things worse (using one truncated twisted tortured torus instead of three cylinders and a cone).

Indeed. But that's a sword that cuts both ways.

I challenge those numbers.

I just went in-world in a First Land sim, totally loaded with everyone's first-and-worst builds, flexiprims, particle emitters, you name it, and I'm getting 20 FPS on my Macbook pro which... while pretty good for a Mac... is hardly a "top class" computer by PC standards. And I've got all my graphics settings but draw distance maxed... draw distance is the default 128m so I'm only seeing 3/8th of the sim (but frankly, would you want to look at more than that?)

My home PC has an Athlon 64 X2 3800+ and a PCI-E 7600, and while it's not top of the line today (alas I bought it before the big cut in Athlon prices) it's a pretty good machine, and I get 25-30 in the same spot with draw distance 192+ and FSAA on as well.

At skybox altitude I get 50+ on my Macbook, and my Athlon's off-scale... literally. 80+ at peak.

The Welcome area? I count 38 avatars in view, and I've gone under 10 on my Macbook...

But... it's not prim wings causing the lag. There's only one set of those in view, even with "show transparent" to help. Lots of chichi hair, though.

That tells me that limiting the number of prims on an avatar is probably running off in the wrong direction. Improving the client application's ability to efficiently hide content is more likely to be useful.

And LL has been working on that. I turned off O-O occlusion in my first test and my FPS went from 20 to 14 immediately, and as I turned around it dropped below 10. I wish they'd bring back "reduce draw distance when FPS < [ ... ]", though...

I see your point about additional textures lagging more. However, once a texture is loaded... it's pretty simple to display thereafter, yes? In other words, a single texture is less laggy than say... 50 prims that would be required to do the same thing.

So since people would become prim-conscious on avatars, we'd wind up honed, streamlined attachments that still look great. Instead of someone making a 300-prim sword because they can, they'd start using textures or be more conservative with wasteful prim use. We'd still have great-looking, artfully-created avatars (just as we have great looking, artfully created buildings and furniture)... and the system would benefit from the reduced prims.

I'd say it's pretty safe to say... it is. Look at griefers. Nothing stops them except restrictions. The no-push setting is one of the best things LL ever did. Now they need to extend that concept to no-bullets as well (probably same setting would suffice for both). The could fire the bullet-- it just would not have any effect on any avatar it touched.

Sure we do. That happens all the time. People make demands to companies. They make demands to their congressmen. They make demands of their families. Life is all about stating needs and demands. That is even moreso in Second Life... where the average user has absolutely no control as to what LL does... except by feedback (which means ideas, suggestions and yes... making demands).

Whether I know how to do something or not is irrelevant. What is relevant is getting LL to act on it. Again, it's not our job to solve HOW to fix the lag. We don't have all the information to make such a call. It's our privilege and right, as paying customers, to note where lag occurs, report that to LL and demand that it be fixed. And if we can think of things that might help reduce lag.. we can throw that in too. That's how life works.


As a note, again-- no one (not me, nor anyone else I've seen) is at this point touting actually reducing prim count as a valid method of lag reduction. It would have to be far more involved than that. Prim composition (weight) would have to be calculated. Simple to do really; every time an avatar puts on or takes off something, the system should calculat a total "weight" of the avatar (ie, types of prims, "torture" value, and number of prims). I think probably this would be based on what.. total number of rendered triangles? Again, we don't have enough info about SL graphics to make that call. But simple prim count.. nah.. .that would likely be ineffective.

However, as far as using "simpler" prims and making avatars look less pretty... has anyone noticed what has happened to hoochie hair since flexi prims came out? Hair is now being made out of simpler prims. Now I don't know what all Flexi does to rendering and prim count. It can't be good. Still, rather than using toruses, people are now coming up with ways to use nice hair textures to accomplish the same thing. The point? Hair never needed to be made out of tortured toruses to look good. So all those hoochie-hairstyles that lagged sims to perdition were really unnecessary.

People can simplify avatar prim use and still have rocking avs. They just have to build smarter, lest wastefully prim-wise, and with a more considerate attitude.

Uh.. I challenge those numbers. Since from what I see, the max FPS showing is 45... 50? 80+???

Actually, textures are alot harder on the system than prims sometimes... because remember, if your moving around alot, or even LOOKING around alot sometimes, your system has to keep reloading the same textures, over, and over, and over.

The system caches textures. So once they're loaded, within the same sim they're pretty much loaded. If you're moving around so much that textures have to reload... prims do too.

FYI, I get 8-12 FPS in areas like Luskwood (tons of furrehs), on a two year old system, that wasn't top of the line /when I got it/. And it's perfectly playable at those speeds.

No offense, but I seriously question "perfectly playable" at 8-12 FPS. Maybe if you're standing around chatting. Not for much else.