Secondlife declining player base

[Penny Patton]

5 minutes ago, moirakathleen said:

I am not a creator or a designer, so the first two points are ones that would be important to me.  Right now, I do not really understand why my onion skin mesh body is not OK

Several things affect performance (ie: framerates, lag, etc), I'll give you the super simplified version:

• The more polygons your computer has to render, the harder it is to render. This means lower framerates, you have to turn off more of the shiny features that make SL look better (like shadows) or you just need to buy a more expensive computer.
• Texture use memory. Memory is limited. Not only by the amount you have on your computer, but also in that SL itself can only really use 1024MB of video memory well. (It gets crashy trying to use more than that.)
• Mesh bodies in SL are, with a few exceptions, higher-poly than they should be. Add to that the "onion skin" layers which double the polygon count over the base body for each additional layer. (so if you have a body with 3 layers (tattoo, underwear, clothes) that's four complete mesh models all being rendered at once. A lot of unnecessary polygons.
• Even if you don't use those layers, you keep them hidden, their polygons are still being rendered by your hardware, and so are the textures. Assuming you have 1024x1024 textures on every layer that's eight 4MB textures right there for a total of 32MB used by your mesh body alone before we even add materials. That's an additional 8MB if you only have materials on your body, but some applier clothing also uses materials which can push an "onion skin" mesh body up to 128MB of memory all on its own.
• That's not including the textures on your mesh head (I think I recall that the Catwa mesh heads use like 60MB of texture or something around there, not entirely sure.) or whatever other attachments you're wearing.

Let's say you have a Catwa mesh head, we're already talking nearly 200MB of textures for your naked avatar. That's just you. Now if everyone around you has a similar setup (mesh body+mesh head) that means you're  maxing out SL's 1024 texture memory limit with five naked avatars on screen. And once you hit that memory limit, you start seeing a severe drop in framerates (especially if you have shadows on, because shadows also utilize that same 1024MB of memory), texture thrashing (where textures go blurry then rez again, over and over), and lag (some of those textures will end up in your cache, but SL also may need to redownload those textures again).

 A system avatar like Coffee describes would utilize just 3 textures for a total of 12MB of memory.

And let's be clear, none of this is a judgement on you for using a mesh body (I use one myself!). I'm not even criticizing the mesh body makers here. LL should have known better from the beginning and given everybody a better framework in which to create our content. It's just that things as they are make SL run worse and worse, hurting the experience for everybody and make it SL an even harder sell to potential new users, which doesn't bode well for SL's future.

15 minutes ago, Fionalein said:

Do you even realize most " complexity offenders" don't even wear mesh?

Citation needed. Also remember that Avatar Complexity doesn't properly take into account either triangles or texture use. I've seen plenty of low complexity avatars who were pushing extreme numbers of triangles and/or textures, making them a worst performance hit than any of the higher complexity avatars around them.

16 minutes ago, Fionalein said:

I'm outa here, this thread is getting ridicilous - if you all know better, please team up and build your own BETTER GRID™and leave us in peace...

No. I was here first.

9 minutes ago, CoffeeDujour said:

Must have doesn't have to be some extra shiny, it really can be .. you must have this because in 8 months, it's going to be impractical.

We're talking a difference of "stick versus carrot". I prefer the carrot. I think it would be SL's best option. But yeah, eventually I think such a cap would have to be gridwide, but the tools and options all need to be in place first, with a time frame that allows people to adapt naturally, without feeling like LL is forcing them to abandon content they paid for. 

9 minutes ago, Magnek Fang said:

You know damn well there's folk with a grudge against non-human avatars or a particular mesh body and will gladly try to shoehorn everybody else into a specific aesthetic if they can get away with it.

Numbers don't hold grudges. These days anthro, animal and mythical creature avatars are no worse (and often better) about texture and polygon use than many regular human avatars.

I can't stress enough that LL shouldn't rush into any sort of new caps. There are ways LL could implement them without disrupting the entire userbase, ways that would allow people to naturally adapt often without even realizing it, and I've always been clear that LL should take that kind of "long view" approach.

4 minutes ago, CoffeeDujour said:

A good complexity would be more like 30K (and that's still overly generous).

You're not wrong, but I'm still not sure SL's complexity calculation is the best way to do it. Unless there's something important complexity considers that a straight look at triangles and texture use does not?

In any case, whatever value LL aims for, they need to properly lay the groundwork, preparing the platform and the userbase (customers and content creators alike) so that it doesn't end up being so disruptive it pushes people out of SL altogether.

[Magnek Fang]

2 minutes ago, CoffeeDujour said:

An actual good complexity would be more like 30K (and that's still overly generous).

30K? Are you joking? I've been doing my damndest to keep myself below 100k with a Kemono or Kuroo with what I've accumulated over the years. You think I can do 30k with a more mainstream body?

This is exactly what I'm talking about here. I'm used to land prim limits because they've been there from the beginning. To suddenly slap on a ARC limit after so many years is just gonna drive people away. 

[Janet Voxel]

24 minutes ago, CoffeeDujour said:

 

A good complexity would be more like 30K (and that's still overly generous).

30k sounds like an arbitrarily low number. Why 30k?

[MoiraKathleen]

24 minutes ago, CoffeeDujour said:

I own shoes that contain more geometry than an entire scene in a triple A game title. This is simply unacceptable.

 

26 minutes ago, CoffeeDujour said:

Realistically, if LL were to announce "System avatar 2.0" today, it would be 8+ months away from deployment on the live grid with a significant amount of time dedicated to tuning the system hand in hand with content creators. That's your grace period.

Must have doesn't have to be some extra shiny, it really can be .. you must have this because in 8 months, it's going to be impractical.

 

27 minutes ago, CoffeeDujour said:

The replacement starter avatars was the Labs way of attempting to show us how they envisioned mesh bodies to be done. Much the same ways as done in commercial game titles, your body and clothing are one single asset. They really didn't plan on us making a million individually rigged clothing attachments.

If LL was to use a similar tone in making sweeping changes as suggested, which it sounds to me are meant to be pretty much mandatory (you must have this because in 8 months, it's going to be impractical) I would be out of here.  Only partly because of changes being forced on me,  but also because it would be a change to the character of the environment.  It doesn't seem to me that LL meant for SL to be like a triple A game title - if they did, they should have made the majority of the content and enforced limits for any allowed user-made content right from the beginning.   

I'm not against having better information and more education provided about how to create and/or choose assets that are better performing, but I think it still needs to be a choice - they need to be options rather than forced. 

[Solar Legion]

Point Blank: Want a hard limit on ARC? How about a mandatory system level avatar update/upgrade?

You do? Go make your own grid and systems then.

[Love Zhaoying]

31 minutes ago, Solar Legion said:

Point Blank: Want a hard limit on ARC? How about a mandatory system level avatar update/upgrade?

You do? Go make your own grid and systems then.

But but but..I’d rather whine!

[Solar Legion]

Just now, Love Zhaoying said:

But but but..I’d rather whine!

I'd offer a hug but I am unsure if you'd try to bite.

[Love Zhaoying]

3 minutes ago, Solar Legion said:

I'd offer a hug but I am unsure if you'd try to bite.

I am quite huggy. Can’t remember biting anyone on the forum.

[Solar Legion]

Just now, Love Zhaoying said:

I am quite huggy. Can’t remember biting anyone on the forum.

Well then, have a big, fuzzy, nine-tailed hug!

[Love Zhaoying]

3 minutes ago, Klytyna said:

*Rant*

Yay, a multi-rant! You rock!!

[animats]

1 hour ago, Penny Patton said:

We're talking a difference of "stick versus carrot". I prefer the carrot.

A nice carrot would be for Marketplace to display objects in 3D with WebGL. Users could rotate, pan, and zoom, as if in world. You'd be able to see the object at each level of detail if desired. As a side effect, Marketplace would know the complexity of the object and could display that. Optional feature for merchants, but your products will look dull by comparison if you don't use it.

[Klytyna]

1 hour ago, Penny Patton said:

Let's say you have a Catwa mesh head, we're already talking nearly 200MB of textures for your naked avatar.

Hmmm 1024 x 1024 pixel texture loads into vram as 4 mb if it's a 32 bit rgba...

Head texture, upper torso, lower torso, thats 12 mb, eyes another 4 mb, hairbase 4 mb, lets say a full body tattoo, 8 mb, lipstick 4 mb, eyeshadow 4 mb, blusher 4 mb, eyelashes and brows 8 mb...

That's 48 mb... add oh, say another 12 for normal maps for the skin, and 12 more for spec maps, assuming you find a skin maker who uses such things...

72mb...

Even adding mesh hair, you are hard pressed to find another 130 mb...

Where do you get this 200 mb textures figure from, for a naked avatar in a mesh body and a Catwa head, or did you just pull a number out of your ass?
 

[Coffee Pancake]

1 hour ago, Penny Patton said:

We're talking a difference of "stick versus carrot". I prefer the carrot. I think it would be SL's best option. But yeah, eventually I think such a cap would have to be gridwide, but the tools and options all need to be in place first, with a time frame that allows people to adapt naturally, without feeling like LL is forcing them to abandon content they paid for. 

The carrot approach has clearly failed. Imagine the mess we would have if instead of LI, land capacity was only limited by it's sqm or ability to have a set number of things. Zero thought would be put into the complexity of rezzed objects.

1 hour ago, Magnek Fang said:

30K? Are you joking? I've been doing my damndest to keep myself below 100k with a Kemono or Kuroo with what I've accumulated over the years. You think I can do 30k with a more mainstream body?

This is exactly what I'm talking about here. I'm used to land prim limits because they've been there from the beginning. To suddenly slap on a ARC limit after so many years is just gonna drive people away. 

 

53 minutes ago, janetosilio said:

30k sounds like an arbitrarily low number. Why 30k?

Painful isn't it. Like .. OMG that's insanely low. But you have to remember the complexity of things is currently insanely high with huge amounts of invisible geometry that still has to be rendered. You aren't just one avatar, with a mesh body and mesh head, you are more like 4 naked avatars worth of geometry + your clothing and hair.

The ARC system we have is not an easy or perfect measurement, it does not correspond to a direct number of triangles or video ram use etc. Maybe linking it directly to VRAM use would be the fairest option.

30K is based on the intent for SL to be used as a social space with the current imperfect yard stick we have, you could have a busy event with 40 avatars at 30K and it wouldn't be the crippling single digit frame rate mess we have now. Get the average down to 15K and doing things that involve moving about will be significantly more feasible.

[Janet Voxel]

20 minutes ago, CoffeeDujour said:

 

 

Painful isn't it. Like .. OMG that's insanely low. But you have to remember the complexity of things is currently insanely high with huge amounts of invisible geometry that still has to be rendered. You aren't just one avatar, with a mesh body and mesh head, you are more like 4 naked avatars worth of geometry + your clothing and hair.

The ARC system we have is not an easy or perfect measurement, it does not correspond to a direct number of triangles or video ram use etc. Maybe linking it directly to VRAM use would be the fairest option.

30K is based on the intent for SL to be used as a social space with the current imperfect yard stick we have, you could have a busy event with 40 avatars at 30K and it wouldn't be the crippling single digit frame rate mess we have now. Get the average down to 15K and doing things that involve moving about will be significantly more feasible.

Not really painful. But uhhh

what happens to users’ objects when you set the cap to 30k?

[Coffee Pancake]

21 minutes ago, janetosilio said:

Not really painful. But uhhh

what happens to users’ objects when you set the cap to 30k?

We already have a soft ARC cap right now in the form of jellydolls. What happens to your stuff .. nothing, no one can see it. Any ARC changes on the part of LL will lower this ceiling. I think the hope was that this would impart pressure on creators to be ARC aware, but realistically that's not happening. Pressure is applied to individuals who have no idea how an attachment will affect their ARC till they wear it presenting two choices. Don't use the thing you just paid for. Don't care and to hell with everyone else.

Avatars are the most expensive thing SL has to render by a wide margin and the only answer TPV's have is to literally switch off avatars .. which is kind of self defeating for a social space.

The bottom line is that avatars are way too heavy and there is nothing developers can do to make that any better, avatar complexity has to change.

Edited August 17, 2018 by CoffeeDujour

[Penny Patton]

Yay double post.

Edited August 17, 2018 by Penny Patton

[MoiraKathleen]

8 minutes ago, Penny Patton said:

Several things affect performance (ie: framerates, lag, etc), I'll give you the super simplified version:

• The more polygons your computer has to render, the harder it is to render. This means lower framerates, you have to turn off more of the shiny features that make SL look better (like shadows) or you just need to buy a more expensive computer.
• Texture use memory. Memory is limited. Not only by the amount you have on your computer, but also in that SL itself can only really use 1024MB of video memory well. (It gets crashy trying to use more than that.)
• Mesh bodies in SL are, with a few exceptions, higher-poly than they should be. Add to that the "onion skin" layers which double the polygon count over the base body for each additional layer. (so if you have a body with 3 layers (tattoo, underwear, clothes) that's four complete mesh models all being rendered at once. A lot of unnecessary polygons.
• Even if you don't use those layers, you keep them hidden, their polygons are still being rendered by your hardware, and so are the textures. Assuming you have 1024x1024 textures on every layer that's eight 4MB textures right there for a total of 32MB used by your mesh body alone before we even add materials. That's an additional 8MB if you only have materials on your body, but some applier clothing also uses materials which can push an "onion skin" mesh body up to 128MB of memory all on its own.
• That's not including the textures on your mesh head (I think I recall that the Catwa mesh heads use like 60MB of texture or something around there, not entirely sure.) or whatever other attachments you're wearing.

Thanks, Penny, for the explanation.  

Are there ways to see texture usage, etc. inworld?  I'm just curious because I have several different bodies, and heads (though none are catwa) and if I did see significant differences between one over another than that might factor in to me deciding which ones to wear more often (rather than looking at script sizes and considering clothing  availability).

[Penny Patton]

1 hour ago, Klytyna said:

Where do you get this 200 mb textures figure from, for a naked avatar in a mesh body and a Catwa head, or did you just pull a number out of your ass?
 

4MB per texture.

With a mesh body you've got the base plus, in my example, three additional layers. Each uses two textures. As I said in my example, at a worst case scenario each layer also uses materials and those materials are also at 1024x1024. That's six textures per layer (6X4=24 textures). Twenty-four textures times 4M is 96MB plus my admitted guess at 60MB for a Catwa head (which, admittedly, I should double check, if that's wrong feel free to call me out on that) which comes to a total of 156 proving once and for all that I am not so good with the maths.

Where did I get 200? Hmm. Let's see...there was one shot in the study, one shot in the hall, one shot killed the singing telegram...or was it two shots in the hall? Ok, I see where I screwed up. Somehow I got 32*3=128 when it's only 96.

But yeah, congrats. You caught me in a small math error. 96MB worth of textures is still obscene for a naked avatar. We're talking "Daikatana arrow" crazy. Even if we assume a best case scenario, no materials on the clothing layers, just the body, we're still looking at six 1024x1024 textures on the body, plus let's even give Catwa the benefit of the doubt and cut my guess in half to 30MB. That's still 54MB of textures for one nude mesh avatar. Which is still crazy high where an identical avatar with the textures baked to a single layer is still only 9 textures, or 32MB worth of textures (assuming materials are being used and the materials are 1024x1024 which honestly they should not be). And that's also 1/3 the polygons. 

So what, exactly, is the point you're trying to make here? Because my point still stands, regardless.

2 hours ago, Magnek Fang said:

30K? Are you joking? I've been doing my damndest to keep myself below 100k with a Kemono or Kuroo with what I've accumulated over the years. You think I can do 30k with a more mainstream body?

You're right, 30K is ridiculously difficult to get below....using horribly unoptimized content. If content in SL were being made with optimization in mind, it would probably be extremely easy for even a highly detailed avatar to get their ARC below 30k. But that's why I keep saying you don't want to drop any sort of avatar rendering cap on people before

1. You've given everyone the tools they need to prepare.
2. You've given content creators all the information they need to work with the cap.
3. You've given enough of a grace period where most avatars will naturally fall below the cap long before the cap is enforced. The idea being you want the cap to have as little an impact on the average SL user as possible.

Honestly, while 30k is extremely reasonably given the existence of optimized content, I would aim more for something less restrictive, like 100-150k. It's excessively generous, yes, but given it would be years before we'd even feel the effects of it and hardware would continue to improve in that time period, plus the fact that merely having a cap that low would be enough of a push to give content creators all the motivation they need to aim for lowest cost+highest quality, and I suspect most people would fall well below the cap. After a few years most people would probably in the 30k neighborhood without even trying.

 That's the goal right there, despite other's trivializing of historical tragedy and staunchly determined ignorance; Reigning in laggy content without most people feeling they've been affected by the cap.

1 hour ago, CoffeeDujour said:

The carrot approach has clearly failed.

I disagree. LL has never offered any carrots for optimization. Not once. They've passed up every opportunity to do it that way. No, the approach that has failed is the "wild west, no rules" approach.

[Phorumities]

I guess they want us to lower our complexity so we can all have those cool animesh pets someone mentioned.

Why is it that every "great amazing new improvement" that LL introduces in in fact another step deeper into a lag black hole?

[Coffee Pancake]

1 minute ago, Phorumities said:

Why is it that every "great amazing new improvement" that LL introduces in in fact another step deeper into a lag black hole?

Because nothing is free. If SL was just prims we would be laughing .... at SL users from some other platform.

[Penny Patton]

18 minutes ago, moirakathleen said:

Thanks, Penny, for the explanation.  

No problem. Someone else in the thread mentioned the importance of getting information out there and I agree.

19 minutes ago, moirakathleen said:

Are there ways to see texture usage, etc. inworld?

I believe both Firestorm and Black Dragon have "inspect" tools that show texture use and triangle counts. Firestorm was given a tool that would let you see it much more easily, and even jelly doll excessively render intensive avatars to give you a really nice performance boost, but they are currently withholding that tool from their users. Linden Lab has also been offered the same features, and they say they intend to include them in an upcoming viewer release, but this was over six months ago so don't hold your breath. It may come eventually but I wouldn't hazard to guess how long.

3 minutes ago, Phorumities said:

I guess they want us to lower our complexity so we can all have those cool animesh pets someone mentioned.

Why is it that every "great amazing new improvement" that LL introduces in in fact another step deeper into a lag black hole?

Like Coffee says, "nothing is free". You throw more complex, texture heavy content at your computer and it's going to choke on it. It's THE main reason SL runs as poorly as it does already, and it's only going to get worse as resource bloat becomes more and more excessive.

Keep in mind, optimized content looks just as good as laggy, bloated content. And it takes no more skills than those already required to create the content in the first place. The only reason content in SL is so unoptimized is that Linden Lab just assumed content creators would just know to optimize their work. Which is really ridiculous. There is absolutely no reason this should even be a topic anyone should need to care about. If LL were smart from the beginning you'd be enjoying content just as good looking as what you have now, but at higher framerates and possibly even higher graphical settings. That would be the only difference as far as most SLers would be concerned.

[Penny Patton]

3 hours ago, moirakathleen said:

If I have to be an expert in understanding how mesh is created, or triangle counts in mesh, or VRAM usage, just in order to choose furnishings and clothes for my avatar to go out exploring, etc., inworld, than that would make being inworld a chore for me, rather than being a relaxing enjoyment. 

I agree! The average SL user should not have to even think about this stuff. If LL had put better content creation tools in place, we wouldn't be in this situation. And the solutions I'm suggesting, as well as the way I suggest they be introduced, is meant to make it as easy as possible for the average SL user. Only content creators should be thinking about triangle counts and VRAM. And then, only insomuch as they already think about things like Land Impact.

That's a point I think I've so far failed to get across to people like Klytyna and Solar Legion. They've got it in their head that trying to solve this problem automatically means harsh, difficult to meet caps that will force them into ugly, undetailed avatars, and that's just not what we're talking about at all.

Because LL has done such a poor job so far it is probably impossible to get on a better path without hitting a few bumps, but there are ways to mitigate those bumps so that the average SL user will be relatively unaffected and have everything they need to get through the transition as painlessly as possible.

 For what it's worth, the current Second Life developers at LL do seem to recognize the issue (finally) and are talking about how best to address it. And they seem to have the same ideas (at least in terms of time frame and grace period) that I do, so fingers crossed they're able to finally do something about bloated content without disrupting the SL community any more than necessary.

Edited August 17, 2018 by Penny Patton

[Coffee Pancake]

Some people settle on a look or avatar years ago and refuse point blank to ever change.

Edited August 17, 2018 by CoffeeDujour

[Penny Patton]

5 minutes ago, CoffeeDujour said:

Some people settle on a look or avatar years ago and refuse point blank to ever change.

It's entirely anecdotal, but in my experience those avatars tend to rarely be a problem. Most of those I see in SL tend to fall well below the types of resource use we're talking about here.

[Janet Voxel]

1 hour ago, CoffeeDujour said:

We already have a soft ARC cap right now in the form of jellydolls. What happens to your stuff .. nothing, no one can see it. Any ARC changes on the part of LL will lower this ceiling. I think the hope was that this would impart pressure on creators to be ARC aware, but realistically that's not happening. Pressure is applied to individuals who have no idea how an attachment will affect their ARC till they wear it presenting two choices. Don't use the thing you just paid for. Don't care and to hell with everyone else.

Avatars are the most expensive thing SL has to render by a wide margin and the only answer TPV's have is to literally switch off avatars .. which is kind of self defeating for a social space.

The bottom line is that avatars are way too heavy and there is nothing developers can do to make that any better, avatar complexity has to change.

Ok, I can get behind that. I just think 30k is a little low for a cap.