Jump to content

What the fitmesh LoD bug actually means


ChinRey
 Share

You are about to reply to a thread that has been inactive for 2038 days.

Please take a moment to consider if this thread is worth bumping.

Recommended Posts

52 minutes ago, CoffeeDujour said:

I doubt it to be honest. The SL fashion industry runs on a very fast tick. Bodies and heads will be updated, clothing vendors will release new stuff and all the old stuff will end up in storage next to our prim shoes, flexi hair & skirts. Some wont make it, others will step up and take their place as has happened every time the content paradigm has moved forward.

Assuming you're talking about solution # 1, correcting the ARC calculation will not fix the core problem, only document it. It is not possible to make fitted mesh less laggy without a functional LoD system for them. Not unless we reduce the visual appearance and I think we agree we agree that's not an option (and it has nothing to do with optimising anyway). So there won't be any updated lower lag fitted mesh.

Solution #2 will force everybody to change to new LoD resistant fitted mesh. Everybody will have to do that overnight. There will be no transition time, the moment LoD models are introduced to fitted mesh, everything will be broken. Giving fair warning might help a little bit of course but how many will wait until the last moment? How many will even notice the warning? Don't forget that until the LoD system is upgraded, LoD resistant fitted meshes may actually show up with slightly higher render weight than non-resistant ones.

As for the combination of solution #1 and 2. I assume that means first correct the ARC calculation to fit the current situation, give merchants a bit of time to produce and distribute LoD resistant fitmesh and then introduce LoD models for it. That may soften the blow a little bit and it may be the only realistic solution. But people will still fail to notice or wait until the last minute before it hits them in the face. We will also need a way to test the LoD models before they are implemented and...

... and that may actually be the seed for the best solution. So, if anybody from LL happens to read this, here's an idea:

  • Introduce LoD models for fitted mesh. LoD swap distances should probably be based on the size the model was uploaded at since the size when worn is a bit too hard to predict.
  • Add a switch to the viewer to allow people to enable and disable LoD for fitted mesh (and fitted mesh only). Make it absolutely clear that this is a temporary one and that LoD models will be permanently enabled soon. When people disable LoD models, they get a popup message telling them this is just a temporary function and how long it will be until the disable function is disabled.
  • Make the ARC value depend on this setting. If LoD models are enabled, they get a base render weight according to the LoD swap distances, if it's disabled, they get a weight based only on the complexity of the high model.
  • Keep this for a few months - not as long as a year - and then remove the disable fitmesh LoD function.
  • Give TPV developers and users a little bit of time to upgrade. Block all viewers that don't comply to the new standard.
Edited by ChinRey
Link to comment
Share on other sites

10 minutes ago, ChinRey said:

Introduce LoD models for fitted mesh. LoD swap distances should probably be based on the size the model was uploaded at since the size when worn is a bit too hard to predict.

I think a set distance from the avatar would be more use. <10M - 15M - 20M - 40M+

That will give creators a solid on screen size to aim and its easy to them zoom the model out in blender (or whatever) to approx the same on screen size and see what it looks like.

Link to comment
Share on other sites

4 minutes ago, CoffeeDujour said:

I think a set distance from the avatar would be more use. <10M - 15M - 20M - 40M+

Maybe. The current LoD swap system is marvelously unsuitable for most meshes anyway. I'd be more than happy if LL can come up with something better.

LoD still has to be checked in-world though. We need a viewer that is able to do that in good time before fitmesh LoD is officially introduced. And once that function is in place, why not give it to everybody?

Link to comment
Share on other sites

4 hours ago, Fionalein said:

And @OptimoMaximo ... you claim you have a background in professional game design... let me put that severly into doubt with all the thoughtless demands you bring forth...

go and buy Skyrim, it's cheap now and comes with the Modding tools so you can see for yourself how game content is done. Even if it is old. Or get into UnrealEngine4 modding community. Put in doubt whatever you want, the thoughtless claims here are those demanding no limits. Show one of the avatars models from Belleza (the devkit comes without UVs only, but the model itself is the same as inworld) to any modder you may encounter, tell them that model goes to runtime as you see it and that clothing goes ON TOP without deleting invisible faces, also that the clothing is double sided by doubling the geometry... possibly show them also that clothing, a full outfit. In wireframe. Let's see who's the thoughtless then.

  • Like 1
Link to comment
Share on other sites

@Fionalein , @ChinRey , @CoffeeDujour , @Callum Meriman , @Nalates Urriah , @Penny Patton take a look at the developers talks on multiple platforms standards and on multiplayer for PC . You can find more yourself, all assessing similar numbers.

After that, take a look at the numbers in these two pictures below, from one of the most famous and widely used mesh body DevKit

MaitreyaLara_4_1.thumb.png.51593ac7d6e8d4d2124d135cf89cafae.png

This devkit comes with ALL the layers for precise rigging. "Yeah that's why BoM is on the works"... right...

MaitreyaLara_4_1_2.thumb.png.e4b8ba255e01a95fbcc5ab1b11a71260.png

This is the single layer for the actual body ALONE. Leaving alone the joints bending area's geometry butchery...which repeats also on the elbow, btw... Both pictures do not include the head, obviously, because that's another variable that doesn't depend from the same developer in most cases. Also not including the other two feet shapes, which we all know are always attached BUT the unused two are alpha'ed out. The numbers rise to ~16K more, in that case.

Read carefully those forum questions posted above, it specifies that characters getting to 100K include clothing, weapons, whistles and giggles (like in one of the exception-games "Ryse", well known for being quite cinematic which gets its main characters up to 150K, always including all the add-on stuff), otherwise 30K/40K for PC game or fewer on other platforms, always including clothing and stuff. This one is DOUBLE just naked.

Yeah, no avatar limits... these are just thoughtless considerations and ignorant babbling.

Edited by OptimoMaximo
  • Like 1
  • Thanks 4
  • Haha 1
Link to comment
Share on other sites

1 hour ago, OptimoMaximo said:

Yeah, no avatar limits... these are just thoughtless considerations and ignorant babbling.

Yes it is! It is out there, you cannot take it back, it maybe should have been prevented first place but taking it away again is not as easy as you think it is...

  • Haha 2
Link to comment
Share on other sites

3 minutes ago, Fionalein said:

Yes it is! It is out there, you cannot take it back, it maybe should have been prevented first place but taking it away again is not as easy as you think it is...

You're not entirely wrong. But there are ways LL can improve things moving forward. Like Coffee points out, old content can be grandfathered and new restrictions could only apply to content made going forward, or weighted to not have so much of a cost. And as I've pointed out, how LL introduces any new caps on avatars, and the steps they take leading up to that, can determine how much the average SL user is actually affected.

 It's something that needs to be approached with planning and consideration, but it can definitely be done without the worst case scenario you're talking about.

Not doing anything, on the other hand, will make it so hardware requirements continue to rise, performance continues to fall, and SL's decline will be that much faster.

  • Like 2
Link to comment
Share on other sites

25 minutes ago, Penny Patton said:

 It's something that needs to be approached with planning and consideration, but it can definitely be done without the worst case scenario you're talking about.

Yes indeed. The issue that "no-limits" people has is that they do not want it to happen, whether it's possible to do or not, they want feature film content "because SL is not a game and can't be like a game" . It is out already? Too bad, at some point the shepherd has to drive the herd back to the fold before it gets lost.

All it takes is one year at the very most, after a new LoD cost system is put in place, intimating creators to reduce their poly count and create proper LoDs. Along with proper LoD switching and render weight calculation, the transition won't be that harsh, provided that those creators would have the will and ability to learn how to optimize their content. Some shall fall, others will keep up. But the current situation is well beyond a triple A feature rich game and can't keep going like this.

  • Like 5
Link to comment
Share on other sites

1 hour ago, Fionalein said:

it maybe should have been prevented first place

Ridiculous. When people does cheat the system on purpose, taking advantage of bugs without reporting the unwanted behavior, there is no prevention that could have been put in place. LL is responsible for the LoDs costs that, for inexperience or purposeful greed with higher costs for meshes that looked good on all LoDs, was there to be used, not to be gamed. A gaming supported by a bug which all creators well avoided to report for their dishonest convenience. It's easier to flood the market and cash out like crazy, instead of creating well crafted content. Takes too long to learn in the first place, and too long to release content. We all know about the well known brand who outed their source of "original content" (after years claiming it was their own) and others that are clearly doing the same. With the mentality you're showing, you're supporting ripped or purchased content instead of items crafted specifically for SL.

Edited by OptimoMaximo
  • Like 2
Link to comment
Share on other sites

2 hours ago, Penny Patton said:

ot doing anything, on the other hand, will make it so hardware requirements continue to rise, performance continues to fall, and SL's decline will be that much faster.

We're in a hole already.

High poly mesh is CPU intensive, SL is CPU intensive.

If you buy a shiny new graphics card it will sit half idle when running SL, your CPU can not pump enough data to fully use it.

CPU's are not getting faster, more and more cores do not help us.

We literally have nowhere to go from here and are already seeing the tide raise cutting people off. I'm not shouting about this for fun, there is no wiggle room in the current architecture. 

  • Like 4
Link to comment
Share on other sites

   The only real incentive creators will act on is consumers who stop buying mesh which isn't optimized for SL, takes advantage of a broken complexity reporting system, or otherwise causes an inordinate amount of system resource strain on a given machine to render. If a considerate, discriminating consumer doesn't know about the problem, or can't see it, they can't act on it.

Link to comment
Share on other sites

How can they tell.

You either don't know what you have till after you have paid for it. Demos are covered in additional geometry or merge multiple outfit parts together so aren't representative of what you actually buy. Rezzed copies are rarely the same model with the same LODs. ARC numbers are fudged and of little value.

Judging requires actual skill. What are people supposed to do .. wire-frame everything, ponder poly counts, access various LOD models., count textures, do a little math?

Altruistic thoughts about performance loose to vanity. Every single time.

  • Like 3
  • Thanks 1
Link to comment
Share on other sites

3 hours ago, OptimoMaximo said:

Ridiculous. When people does cheat the system on purpose, taking advantage of bugs without reporting the unwanted behavior, there is no prevention that could have been put in place.

I think what they mean is that avatars should have had limits in place from the beginning, before people ran wild with unoptimized content and using models not intended for realtime rendering. I think we can all agree that yeah, they really should have. There's a lot LL could have, and should have done from the beginning. But they didn't. I'm not sure that the Lindens working on SL now were in their current positions, or maybe even working for LL at all, at the time LL was making all these bad decisions, so I don't want to lay these problems at their feet, but I agree something needs to be done now.

1 hour ago, Ivanova Shostakovich said:

   The only real incentive creators will act on is consumers who stop buying mesh which isn't optimized for SL, takes advantage of a broken complexity reporting system, or otherwise causes an inordinate amount of system resource strain on a given machine to render. If a considerate, discriminating consumer doesn't know about the problem, or can't see it, they can't act on it.

 

1 hour ago, CoffeeDujour said:

How can they tell.

This is why I advocate the introduction of simple, easy to use tools residents can use to get all the information they need, in a form that makes sense to them.

Land Impact is not a bad system in itself. It's got a few things that need to be worked out, but the basic idea is fine. Land grants you X number of "Land Impact" points and everything you want to rez costs a certain number of points. Simple! Easy to understand. There is no reason a similar system couldn't exist for avatars. Add in tools to see the number of points or whatever an object costs, make it so creators are incentivized to provide that information on the marketplace, and make it so people can easily pinpoint attachments that are eating up a lot of their points. 

But giving the people tools to work with has to come before any limits are put in.

  • Like 1
Link to comment
Share on other sites

4 hours ago, CoffeeDujour said:

 High poly mesh is CPU intensive, SL is CPU intensive.

Most modern CPUs have more than one core... there is enough "wiggle room" find ways to access it instead of trying to shake the pillars of SL business, they might move, but they might as well just topple over, taking with them all who could have stabilized them...

Edited by Fionalein
Link to comment
Share on other sites

1 hour ago, Fionalein said:

Most modern CPUs have more than one core... there is enough "wiggle room" find ways to access it instead of trying to shake the pillars of SL business, they might move, but they might as well just topple over, taking with them all who could have stabilized them...

We have had multiple cores for many years now. Almost nothing with a real time variable load is done with it. The overhead and management take up more time than the problem that's attempting to be solved.

This isn't a failure of imagination or lack of sweat. You're hoping there is a rabbit in the hat and really .. there might be many rabbits.. but you can only pull one out at a time, and you have to wait for all the other rabbits to decide if it's ok.

Some problems scale really well, SL on the client is not one of those problems, neither are games in general. 

2 hours ago, Penny Patton said:

There is no reason a similar system couldn't exist for avatars. Add in tools to see the number of points or whatever an object costs, make it so creators are incentivized to provide that information on the marketplace, and make it so people can easily pinpoint attachments that are eating up a lot of their points. 

This is how it should have been from the start, increases could have come over the years as computers got better, but hindsight is 20/20 and back when avatars and attachments were designed there wasn't even a foreseeable need.

Adding them now with a years free-ride till they are in effect would inspire half of us to create compliant content, and half of us to hack in as much content as possible to ride well past the deadline, business as usual.

 

 

  • Like 1
Link to comment
Share on other sites

I am not arguing that things aren't a mess  -- and from my viewpoint seem to be getting worse rather than better (this judging from the amount of items I do NOT blog these days :D because of "issues"  -- AND the very obviously NOT original content at most of the top events). 

BUT for those of you that have been around for 10, 12, 13 years, SL  has ALWAYS been a platform that required an above average computer. When my RL friends joined 13 years ago -- all in the web, computer, multi-media field which already required fairly hefty computers -- they EACH had to invest in better graphics cards in order to get SL to run in a respectable way.  I had to also in order to see shadows LOL. 

So while mesh may CURRENTLY be the problem, this is not a new issue overall.  

I agree that things need to be throttled; to a small extent this has been done in Sansar where there ARE limits on mesh uploads (mostly wearables but some rezzed mesh also).   I doubt seriously that The Lab will do what any of you WANT them to do, though.  I suspect changes  will be fairly minor compared to your lists :D.   

It is pretty obvious that these themed threads that go on forever have pretty much said it all -- already.   The creators that are going to listen, have listened (and there are some). Those that don't care or have marginal skills to start out with aren't going to change. I personally don't think they would change if "forced" to by rules. 

This same outlook of "me first and *&%$ everyone else" isn't just in SL, it is all over the real world and we see it everyday.  So in my mind the best we can do is be true to ourselves and our beliefs and make an effort to do the best that we can in our creations. 

 

 

 

 

  • Like 5
Link to comment
Share on other sites

14 hours ago, Chic Aeon said:

So while mesh may CURRENTLY be the problem, this is not a new issue overall.  

You're absolutely right. Texture use and avatar attachments were both out of control pre-mesh. Mesh just made the problem worse. And mesh isn't the only problem, it's one of many.

14 hours ago, Chic Aeon said:

It is pretty obvious that these themed threads that go on forever have pretty much said it all -- already.   The creators that are going to listen, have listened (and there are some). Those that don't care or have marginal skills to start out with aren't going to change. I personally don't think they would change if "forced" to by rules. 

I don't expect LL to change based on any forum posts, but simply spreading information around, even if it's all been repeated over and over, does have some small impact. Every so often I get a message from someone wanting to know more about how this or that SL feature works, or how to better optimize their work. Some of them content creators who ended up changing their approach to content creation for the better. That's all I really hope for when posting.

  • Like 2
Link to comment
Share on other sites

17 hours ago, Chic Aeon said:

So while mesh may CURRENTLY be the problem, this is not a new issue overall.  

However this problem is slightly different in that there isn't a natural escape. CPU's simply aren't getting faster, they haven't been for a while now.

We will not be able to buy our way out of this by upgrading our hardware. Put a brand new top of the line graphics card in a PC and run SL .. one core of your CPU is running flat out and your GPU is barely breaking a sweat. It's very badly bottlenecked.

The only software based solution is major architectural changes to the way the client works, major enough to be a whole new beast entirely.

The Lab chose to go make Sansar for lesser reasons than would be required to re-engineer this into SL2. So we're left with plan B, convince people to use fewer triangles.

  • Like 2
Link to comment
Share on other sites

  • 2 weeks later...

Sure, game models are made by professionals, but it doesn't mean they are perfect either, they have time constraints and sometimes, they are just sloppy. But despite that, they always look better than anything we make here, and at a fraction of the rendering cost.

WIP201.thumb.png.823a0407125046ae1c89161f3f88452e.png

Top picture is a dress + apron I made a while ago... And then we have the pixture below: (yes this is wireframe)

1466935343354.thumb.png.31789427ab8d3c8f17503c5fe438735b.pngjP6HLat.png.3458fe2021446c4ed8c99cd89e9888f1.png

If we get a complexity cap, this wouldn't be the death of detail, it would just mean people will have to try harder to get the details out.

"Detail greeble" Is not a sign of skill, talent or quality, hell people in the game industry spend a ridiculous amount of time automating them: city generators, monster generation, hardware tesselation.

Edited by Kyrah Abattoir
  • Like 2
  • Thanks 2
Link to comment
Share on other sites

What strikes me is that all of this is obvious to anyone with any 3D design experience outside of SL. Whether you're a student, professional, or a hobbyist that has been involved in game modding.

 I'm not sure LL has anyone like that on staff, not that they let anywhere near development of the platform at least. It's people like Oz and Vir who make decisions regarding content creation tools, Land Impact calculations, etcetera. And while they may be skilled in their area of expertise, they obviously don't know squat about the content creation side of things, leaving SL development entirely blind to entirely obvious issues like this and how to deal with them.

 LL needs an Art Director. Someone who can tell the software engineers how to deal with these problems, and when they say something needs to be done it is taken as gospel. Not tossed on some back burner until it's forgotten, or dismissed out of hand because Oz or whoever doesn't understand why it's important or see the problems they're causing by ignoring it.

  • Like 1
Link to comment
Share on other sites

14 minutes ago, Penny Patton said:

What strikes me is that all of this is obvious to anyone with any 3D design experience outside of SL. Whether you're a student, professional, or a hobbyist that has been involved in game modding.

And that is your problem, Penny. No they don't. If you do movies FPS don't count, If you do 3D printing FPSs do not count in fact in all freakin 3D design applications outside gaming... FPSs do not count - resolution however does. Do not assume you know everything, because as you just showed you have no idea for what 3D design can be and is used for...

Yes, SL would be better off with content optimized for gaming but not everyone in SL has a gaming background. At least accept the little fact that not every freakin 3D designer has to come from gaming... a lot don't, duh... a penny for your thoughts though (pun intended) as that might help me in understanding your little Jihad...

Link to comment
Share on other sites

You are about to reply to a thread that has been inactive for 2038 days.

Please take a moment to consider if this thread is worth bumping.

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
 Share

×
×
  • Create New...