Penny Patton

It would be more correct to say that textures and geometry impact performance in different ways. Excessive use of textures can have a significant (far more than LL's complexity calculations account for) impact on performance if they fill up your VRAM. Textures may be "rendered once" but they have to remain in memory to remain visible on your screen. When you have avatars wandering around with literally half a Gigabyte of textures (or more, I've seen avatars using well over a gig) while enjoying a low complexity score, it indicates a major disconnect between the complexity calculations and performance impact. LL is absolutely undervaluing textures in these calculations. This is demonstrably incorrect. I've been building sims for years and reducing the texture load of those sims has resulted in significant performance gains. Not just increased FPS (and I'm talking doubling or even tripling my framerates) for myself and all visitors to my sims, but also faster rez times and the complete absence of "texture thrashing".

I usually tell people "high complexity is always bad, low complexity is not always good." Unless LL has changed the formula recently it largely undervalues things like VRAM use. An avatar with a lot of large textures can tank framerates just as badly as one with a high poly count. High VRAM use also sucks up bandwidth. It would be more useful if LL displayed VRAM use and triangle count. Even better if the viewer gave people that info on their own individual attachments, so residents could pinpoint problematic attachments and optimize their appearance.

Just moving the height slider squashes and stretches your avatar. If you make a shorter shape and also want to be thin and lanky then you need to adjust other sliders to achieve that.

I'm not sure what sort of problem you're seeing with PNG files. Screenshots might be helpful. As for exporting TGA files, it's different in that you don't make the pixels themselves transparent, you create an alpha mask channel and work with the transparency that way. I do find that this gives me more control over how the transparency looks once imported into SL and it's non-destructive, leaving the texture intact, so I tend to prefer that method. Once you've made the alpha mask, you do have to make sure that Photoshop is saving the alpha channel along with the file. There's a checkbox for it when you're entering the file name, and after that you have to save it as a 32bit texture.

Oh, I agree, but it doesn't change the fact that this is the only way to stop the out of control content that is spiking bandwidth use and killing framerates. They will have to do this eventually if they ever hope to solve the problem. Doesn't mean they will. I could totally see LL kicking this can down the road indefinitely. As for whether or not the users would tolerate it, well that depends on how LL handled it. Most of the user fear about placing caps on avatar resource use comes from two beliefs: The belief that it would render the content they own unusable. The belief that it would force people to have lower quality avatars. LL could get a handle on the first issue by taking things slowly. By introducing features that help highlight poorly optimized content, guiding content creators to create better optimized work, well ahead of any announcements about avatar rendering caps. And also giving a lengthy grace period. Old content, in general, tends to get phased out with time. The people with the more resource intensive avatars typically aren't wearing the same old content 3 or so years down the line. They move on to shiny new content. As for the second concern, better optimized content doesn't mean worse looking content. Much of the resource bloat you find in content isn't anything you can see. It's huge texture maps that are mostly blank, unused space. It's mesh heads with complex mesh logos or fully meshed and textured brains hidden inside, that you never see. It's enormous textures on the tiniest details, where you literally cannot zoom in close enough to see the full resolution of the textures. Get content creators optimizing your work and right away you'll see content that looks every bit as good as most of the content you see today. Your average SL user will see no loss in quality. And LL could even grandfather in some old content, like pre-mesh content. The few people who still look fondly on that content would account for minimal performance issues for others. I'm sure even with all that there will be those who complain, but it wouldn't be a complete riot by the userbase, and the grumbling would pass with time.

Unfortunately, no amount of overhauling is going to make SL run better with these extremely resource heavy avatars. The only solution is to cap the resources an avatar can use . Obviously I'm not suggesting LL render the majority of content unusable overnight by suddenly dropping such a cap on us, they need to think long term. Take measures that will curb the use of resource intensive attachments over time, give users and content creators alike the tools and guidelines to help bring avatar resource use down while also making it easy for content creators to produce more resource friendly content. LL needs to show people that optimization doesn't always mean sacrificing visual quality. They need to encourage content creators to put the effort in towards optimizing their creations. If LL does this right then by the time they announce such a cap is coming, the average rendering cost of an individual avatar will be trending downward, and by the time such a cap actually arrives, then hopefully most of the userbase won't feel affected by it, or will at least find adapting to be be easy.

I don't believe you're thinking big picture enough. Back when LL was on the verge of releasing mesh, a number of us tried to convince LL that if they released rigged mesh and left it to content creators to create rigged mesh bodies, while they left the system avatar as it was, then it would completely fracture the clothing market. No longer would you be able to buy clothes and simply expect them to work with your body, you would need to shop exclusively for clothing designed to work with whatever rigged mesh body you were using. And that this would greatly increase the difficulty for new users who simply wanted their avatar to look decent. LL didn't believe us, but look at SL today and you'll see that's exactly what happened. LL can't undo that mistake, not overnight. But they could, finally, release a system av 2.0 that was on par with the mesh bodies on the market and then slowly, over years, we'd eventually start to see an SL were the bulk of mesh clothing would be rigged for the new system avatar. An SL where a new user could log in for the first time, go to a clothing store, buy something, and it would just fit.

What are you basing this on? We've never seen LL doing anywhere near "the best it can". LL has made many missteps over the years that have often made SL more difficult to get into rather than easier, and the presentation has always been an extremely low priority for LL. Not to mention LL has made costly mistakes with the content creation tools which have directly lead to all of SL's worst performance woes, which is why the general perception of SL is that it looks worse than a 12 year old game, but runs like crap even on modern hardware. I'm not saying SL could be as popular as Facebook, or that LL would have ever achieved their goals of a 3D internet, but it stands to reason that an SL that didn't suffer from so many critical, and avoidable, mistakes, a Second Life that looked better, ran better, and was far easier to get into and enjoy, could have been more popular than the SL we got.

As of the last Content Creation Group meeting I made it to (which was either this past one or the one before it) no ETA yet. But one of the new Lindens working on it did say they'd whittled down the bug list. Hopefully we'll get an ETA soon.

This is why I always say meaningful change needs to start with LL. There needs to be constraints on the rendering resources an avatar can eat up.

Especially when it comes to something like hair, alpha masking works best if you design the texture to be alpha masked. The less "fuzzy" the texture's alpha edges are, the smoother it will look when masked.

If you don't need them to move independently, you could make one large texture with both planets and use that for the moon texture. (although, being able to add more independently moving objects through the sky would be great.)

Its' much more than scripts and script memory. It's things like polygon counts and texture memory as well. And you actually have very limited control over the rendering impact of your environment (unless you make everything yourself). Sure, you can limit the max resolution of textures, but if the sim has 80GB worth of textures and you cut that in half or even down to a quarter by restricting max resolution, that's still more textures being crammed into your videocard than you have memory for. As for derendering avatars, that's hardly an optimal solution. If I have to derender nearly every single avatar in SL to make it run decently, then what's the point of having avatars at all? Should LL be honest and make that a part of their marketing? "Sure, you can make your avatar look however you like, but you're the only person who will be able to ever see it!" They're not going to draw in many new users with that. And no one is suggesting LL restrict content creation to pro-level creators. Optimization is pretty basic stuff. If you have the skills to create the content at all, then you have all the skills you need to optimize that content. People don't do it for one reason, and one reason only: LL has never incentivized optimization. There's no penalities for unoptimized content and most SL content creators do not understand how their content is directly impacting everyones' ability to render SL. If that ever changes and LL imposes penalities on unoptimized content (ie: higher land impact costs, restrictions on avatar attachments, etc) you will see every single SL content creator overnight suddenly discovering they had the ability to optimize their work all along.

Well, if you build your own environments in SL (whether it be your own small skybox, or a full RP sim) you also get to enjoy a relatively lag-less SL. I've been building sims for years and they all rez faster, have little or no texture thrashing, and don't freeze up my computer every time I move my camera. And I'm constantly getting messages from visitors gushing because they enjoy those benefits as well.That right there is more than enough incentive for me to keep doing what I do. Seriously, I enjoy better framerates and less lag in my own sims on a ten year old computer than most people with brand new computers are likely to ever experience in SL.

I've found that most SL content creators I encounter in-world don't know about emissive masks. Love it. Someone should make a guide like that for texture use, too. I rarely see avatars with under 200MB of textures on them, and I'm increasingly seeing avatars pushing 500MB to a Gigabyte of textures. And people wonder why SL runs as poorly as it does.

It's good to reduce the number of those as much as possible. As ChinRey pointed out, Blended alpha was originally the only alpha mode, and any texture you uploaded that had an alpha channel in the file would automatically be displayed in blended mode. (Which, as you can imagine, caused a lot of problems.) A few blended textures here and there shouldn't be a problem. Just use them sparingly. I took this to mean your sweaters. What about the store itself? Do you have any neighbors? The content you make does not exist in a vacuum. If your sweaters only use a single 512x512 texture, that's great, but it's not just the size of the textures, but the cumulative amount of memory SL allocates to the textures displayed on screen. And, unfortunately, unless you made everything in the entire sim by yourself and you know you're using good game design habits in your content creation, SL does not provide you the tools to manage this important resource. Firestorm has added some features, though, allowing you to inspect objects to see how much memory they use. That's a start.

Possible, but it's also important to remember that we tend to use too many large textures in SL. It pays to be picky about the content you build with, using content with a low memory footprint (ie: fewer and smaller textures), and when making your own content be smart about texture use. You want to get the best possible results with the fewest and smallest textures possible. Since almost no one in SL does this, your viewer is quickly overwhelmed with texture fetch requests, and you run out of memory to keep them all rezzed.

Materials (spec and normal maps) can help with the illusion of fluffiness, too. So the OP doesn't think these answers are just being dismissive, the reason fuzzy layers way of doing it isn't ideal is twofold: The extra geometry adds that much more that needs to be rendered, becoming a drag on rendering performance. It may not seem like all that much on its own, but when everyone is doing it (as well as other bad creation habits) it adds up fast. The second issue is that blended alpha is very difficult to render. It has its uses, but it should be used sparingly. For that matter, hair, clothing, makeup and tattoo layers on mesh bodies, all become a major impact on rendering performance whenever they're on screen. The more we reign in these habits among content creators, the better SL will run for everyone.

Maybe but I wanted to answer the OP's question on why so few SL content creators do care to learn the basics of content creation. Sure, we might understand that relationship between content and performance, but the average SL content creator, just like the average SL consumer, does not. And even if you tell them, a lot of them are unlikely to believe it. Believe me, I've seen it constantly here on the SL forums. I've even seen viewer developers argue against the need for content optimization. For that matter, LL is sitting on some tools that were submitted to them that would make it a lot easier to see and manage your own avatar's triangle count and VRAM use, as well as jelly dolling other avatars based on those same aspects (which is way more useful than using ARC to determine which avatars are the problem).

"Why don't SL content creators learn the basics first?" Because there's nothing for them to overcome by doing so. Few SL users ever make that connection between poorly made content and the multitude of issues SL users complain about on a daily basis. So they don't see the incentive in putting in any effort at being less wasteful with textures and polygons. They don't see the downside in ramping object detail up as high as it will go so only the highest LOD models are ever displayed. The fact that this hurts everyone's framerates is just too abstract for them to grasp. There should be blocks in the way between a user deciding to make content, and actively selling that content to other SL users. Caps on avatar VRAM use and triangle counts. Having decently well made LOD models should not impede content creators with unreasonable LI costs, but gutting LOD should have some immediately recognizable negative effect that you can't get around by digging in the debug panel. These blocks shouldn't be insurmountable. It doesn't take that much effort to go from creating content, to optimizing that content. People just need to be incentivized to actually do it and that's only going to happen if unoptimized content is rendered undesirable to the userbase at large. And that is going to take action on LL's part. Getting from here to there isn't as easy as flipping a switch, of course. I'm not suggesting LL drop in new limits willy-nilly. But it's definitely possible for LL to work out a long term plan to get us where we need to be.

I'd be curious to see some examples, to get an idea on what we consider "decent". I try to retain the basic shape of an object at the lowest LOD, but it's fairly abstract as the lowest LOD isn't meant to be seen up close. For something like a milk can I'd have it be a box at the lowest LOD. For something more complicated like this sci-fi street lamp I try and make it so from a distance you can still make out the shape, but up close its a clump of triangles arranged in vaguely lamp-like fashion.

I usually have my LOD set to 2 or 2.5 when sim building. I've found that as long as an object holds up at those settings it's generally good enough. And if you can't make content with strong enough LOD to hold up at those settings, you're just not trying.

Latest update from Vir's office hours meeting this afternoon; LL just hired two new graphics people, one of which has already started and the other has yet to actually start, after their new employee training is over with they're going to be put to work on EEP.

So apparently the EEP project viewer is no longer coming. Instead it will be a Release Candidate, but Oz said that this might not be released for a while due to the fact that they just had a release and they want to fold in a bunch of other in development features for this coming RC.

Here's the TL:DR version: The average normie avatar, especially guys, tend to be 7-9' tall. This isn't a problem except that when they see avatars that are 6' or shorter, which many anime style avatars are (Furries, too, for that matter. Also, women in growing numbers. And landowners who want to get their money's worth.) too many of them cry "Child avatar!" and try to get shorter avatars banned simply for being shorter than they are. That's stupid. They should stop doing that. I'll add that there's a simple thing LL could do to begin clearing up this issue.