What is a good amount of triangles/vertices/faces for a mesh body?

[Penny Patton]

45 minutes ago, Gabriele Graves said:

These extra things mostly add extra mesh to achieve with the not-selected options being made invisible.

I'd still count them in the overall triangle count, where I would not with some other bodies, because they're non-negotiable. You HAVE to wear all those additional invisible pieces when using the Lara. That's one of the reasons why I find it easy to believe you could have a very similar body, with identical features, for a fraction of the triangles.

[ChinRey]

1 hour ago, Gabriele Graves said:

Sorry but if you think that the Kemono body proves that 30K looks as good as a M. Lara, you just don't see what myself and the rest of the market does.

The Kemono is a bit special but it certainly has smooth enough curves!

I was thinking more about Slink and Ruth 2. Slink has a higher tri ***** of course but that's because the hands and feet are old, dating back to the time before fitmesh lag was an issue only a few people were aware of.

Edited April 13, 2021 by ChinRey

[ChinRey]

2 hours ago, Gabriele Graves said:

I could try and guess but no, it isn't obvious.  However we only have just one static image to go on, cannot see everything and both are pretty rough and blocky looking.  I would think the real test is when you want to make an avatar that doesn't look blocky at all.  It is interesting though how bad one of them is compared to the other.  I guess that right hand avatar matches the left hand stats.

No, it's the other way round. That was the point of course: the least good looking NPC (they're not avatars but static rezzable figures - yes, they're that old) has by far the highest tri count

[Gabriele Graves]

43 minutes ago, Penny Patton said:

I'd still count them in the overall triangle count, where I would not with some other bodies, because they're non-negotiable. You HAVE to wear all those additional invisible pieces when using the Lara. That's one of the reasons why I find it easy to believe you could have a very similar body, with identical features, for a fraction of the triangles.

Yes, I had thought so too.  However, us M. Lara users for the most part do want the nail shapes, the feet positions, the different necks (due to different head requirements) and the alpha cuts.  So those features are a minimum and not negotiable either.

Though I have often thought it would be better to have animesh feet than multiple invisible copies but because LL decided that only Premium members get to have more than one animesh attachment, the creator has decided not to do it because for many, the feet would be all they are able to wear.

Not sure about the popularity of the HD nipples I have read mixed comments about those and I could happily live without them.

Edited April 13, 2021 by Gabriele Graves

[Penny Patton]

9 minutes ago, Gabriele Graves said:

However, us M. Lara users for the most part do want the nail shapes, the feet positions, the different necks (due to different head requirements) and the alpha cuts.  So those features are a minimum and not negotiable either.

Right. I'm just saying that as far as nail shapes, feet positions, neck sizes, etc could all be alternate attachments rather than every single version all worn at once.

The Kemono has like 3 or 4 different necks, two competing brands of feet that each come in like 3 different positions, and at least two different brands of nails (I think they come in different sizes but I'm not sure and not able to log in right at the moment to check) but they're all different attachments that can be worn individually. So I'd only include the variations I'm wearing at the time in my total triangle count, because that's all my hardware needs to worry about.

And, again, I'm not trying to prop the Kemono up as an alternative to the Lara, just using it as an example of a different way creators could approach building out their bodies and features in a more render-friendly way.

[Gabriele Graves]

6 minutes ago, Penny Patton said:

Right. I'm just saying that as far as nail shapes, feet positions, neck sizes, etc could all be alternate attachments rather than every single version all worn at once.

The Kemono has like 3 or 4 different necks, two competing brands of feet that each come in like 3 different positions, and at least two different brands of nails (I think they come in different sizes but I'm not sure and not able to log in right at the moment to check) but they're all different attachments that can be worn individually. So I'd only include the variations I'm wearing at the time in my total triangle count, because that's all my hardware needs to worry about.

And, again, I'm not trying to prop the Kemono up as an alternative to the Lara, just using it as an example of a different way creators could approach building out their bodies and features in a more render-friendly way.

Ahh right, I get you now.  Yes, I am a big fan of making all this stuff optional which is why I liked that Maitreya made their applier layers optional.

I didn't think of you as propping up the Kemono at all just as I expected that no one in this conversation would think that I am propping up the M.Lara, so we are hopefully all good over that.

Edited April 13, 2021 by Gabriele Graves

[Penny Patton]

Ditto. As an aside, hair that comes in multiple styles chosen through a HUD are real framerate killers. I tried out some hair this morning that, by itself, pushed my avatar over 1 million triangles. I literally detached everything else and just wore the hair and that alone was over a million tris. 

Luckily the hair was mod. I was able to unlink the 8 or so alternate hair meshes and only use the version I wanted to use and that was a much more reasonable amount of triangles.

I look at it this way. SL is nearly 20 years old. Mesh in SL is about ten years old now. If LL had capped us at half a million triangles back when mesh was introduced, most content at the time would have been unaffected and we'd be in better shape today. If LL capped us at a million triangles TODAY, 4/13/2021, then most content would be unaffected and we'd be in better shape ten years from now. Maybe not great shape, but our hardware would at least have ten years to catch up without the goal posts continuing to shift.

[Gabriele Graves]

Yes, some clothing is like this too now.  Not a big fan of those either.

It's a fair comment about the capping and I think I could live within a millions triangles budget but some of the people I see around...I am surprised the region doesn't just die when they arrive.

[Wulfie Reanimator]

Since we're still talking about the Kemono, I'd like to point out that the body itself (a fair comparison to other bodies like Maitreya -- excluding head or tail or anything else) is only 14'400 triangles, including two sets of legs from the knee-down.

Since the body is modifiable, I can remove the animal legs so that we're comparing two "fully human" bodies. That makes 11'100 triangles vs 176'300.

Granted, the Kemono body is not rigged to as many (read: most) shape-sliders as Maitreya, so it doesn't need to worry about not having enough topology to deal with the extremes.

Edit: Here's "Avatar 2.0" from the same creator as the Kemono. Rigged more completely.
17'200 triangles after I removed extra boobs. The butt is kinda lacking, though.

And while I am an advocate for better content, I don't think we need to set the bar around 10K. Making that amount look good does take a lot of skill, effort, and/or tradeoffs.

30-40K for a fully-rigged, human, headless bento body would be much more reasonable, especially for those who are already commercially successful in the current market.

We'll never hit numbers like we see in "real games" because of the kind of environment SL is, and how we interact with it. Humans will always compromise efficiency for convenience.

Edited April 13, 2021 by Wulfie Reanimator

[ChinRey]

5 hours ago, Penny Patton said:

If LL had capped us at half a million triangles back when mesh was introduced, most content at the time would have been unaffected and we'd be in better shape today.

Let's see... 50 avatars in a busy club, each with half a million tris, that adds up to 25 million.

But apart from that, I think you've put your finger at the crucial point, the root not only of the mesh body issue but most of the mesh problems we see in SL today. I have to get back to that later, it's a post that'll need quite a lot of preparation and I'm a bit busy with another project right now.

[Penny Patton]

1 hour ago, ChinRey said:

Let's see... 50 avatars in a busy club, each with half a million tris, that adds up to 25 million.

I'm not saying it would be perfect, but it would have been a helluva lot better than now where people are pushing a million tris each.

[Kyrah Abattoir]

It comes to my mind that another performance killer comes from sliceable bodies (increase in draw calls), which are "almost" not needed anymore since BoM rolled out. I'd argue that the reason only a handful of bodies embraced it fully is that most mesh body creators are unable/unwilling to upload a new slice-less version.

There could be many reasons, loss of originals, tool changes creating mismatches, it could also be the fear of changing something that is fully accepted by now (slice huds).

I'm not going to pretend that I understand how separate materials are rendered in SL but I'd wager is is more likely than not that no consolidation is done, and that a 50-80 slices body is rendered as 50-80 separate materials.

[Gabriele Graves]

7 hours ago, Kyrah Abattoir said:

It comes to my mind that another performance killer comes from sliceable bodies (increase in draw calls), which are "almost" not needed anymore since BoM rolled out. I'd argue that the reason only a handful of bodies embraced it fully is that most mesh body creators are unable/unwilling to upload a new slice-less version.

There could be many reasons, loss of originals, tool changes creating mismatches, it could also be the fear of changing something that is fully accepted by now (slice huds).

I'm not going to pretend that I understand how separate materials are rendered in SL but I'd wager is is more likely than not that no consolidation is done, and that a 50-80 slices body is rendered as 50-80 separate materials.

Sadly they are needed because the vast majority of users need a simple clickable way of controlling which areas are alpha'd.  It doesn't matter how easy people claim it is to live without them, it really isn't so and many people just will not transition to a body without them.  How does it compare to the other performance killers?  Is it number one or further down the list?

[ChinRey]

5 hours ago, Gabriele Graves said:

Sadly they are needed because the vast majority of users need a simple clickable way of controlling which areas are alpha'd.  It doesn't matter how easy people claim it is to live without them, it really isn't so and many people just will not transition to a body without them.  How does it compare to the other performance killers?  Is it number one or further down the list?

It's hard to say. Draw calls are one performance issue that has always been neglected in SL. Even the early Lindens who were always very performance concious didn't give it much attention.

It should be a problem that'll go away in the future though. BOM is supposed to make the alpha slices for fitmesh bodies redunant. Alpha layers not only performs better, they also allows for more precise masking so hopefully mesh clothes creators start to include them with their works.

Further down the line LL will have to replace OpenGL with Vulkan and Vulkan doesn't use draw calls as such at all.

[Gabriele Graves]

Thanks for the info about the draw calls.

10 minutes ago, ChinRey said:

It should be a problem that'll go away in the future though. BOM is supposed to make the alpha slices for fitmesh bodies redunant. Alpha layers not only performs better, they also allows for more precise masking so hopefully mesh clothes creators start to include them with their works.

That's my point, it isn't a problem that is going anywhere.  I am fully BoM and so are many other Maitreya users, however we still want the alpha cuts for reasons I have outlined in many other topics touching on this.  It doesn't matter how much better alpha layers perform, many, many people are not going to update to a body without alpha cuts.  That is the one of the big mistakes that Slink made and now support for their bodies is vanishing.

[ChinRey]

1 hour ago, Gabriele Graves said:

Thanks for the info about the draw calls.

That's my point, it isn't a problem that is going anywhere.

Well, regardless of whether the alpha cuts remain, there's still Vulkan coming up. That transition is going to be a big challenge for LL - maybe the biggest they've faced for years - but it's one they have to take on since Apple is planning to drop the little OpenGL support they still have. (Yes, I know Apple isn't going to support Vulkan at all but there are already ready made workarounds for that in place.)

Edited April 14, 2021 by ChinRey

[Gabriele Graves]

4 minutes ago, ChinRey said:

Well, regardless of whetehr the alpha cuts remain, there's still Vulkan coming up. That transition is going to be a big challenge for LL - maybe the biggest they've faced for years - but it's one they have to take on since Apple is planning to drop the little OpenGL support they still have. (Yes, I know Apple isn't going to support Vulkan at all but there are already ready made workarounds for that in place.)

I don't know very much about Vulkan.  I do know my Linux distribution has Vulkan support though.  Is that likely to provide a big boost in performance?

[Coffee Pancake]

3 minutes ago, Gabriele Graves said:

I don't know very much about Vulkan.  I do know my Linux distribution has Vulkan support though.  Is that likely to provide a big boost in performance?

Vulkan render engines can be built to take better advantage of multiple cpu cores and modern graphics cards. SL does on paper present a perfect use case for Vulkan and could benefit greatly, if the solution is built with this in mind from the ground up.

It's also possible to simply port an OpenGL application to Vulkan and get none of the benefits 🥺

A major issue for SL is that not all graphics cards in day to day use have drivers that support Vulkan.

[Mollymews]

16 minutes ago, Gabriele Graves said:

I don't know very much about Vulkan.  I do know my Linux distribution has Vulkan support though.  Is that likely to provide a big boost in performance?

the Naivi Habitat project was ported from OpenGL to Vulkan. More here: https://blog.napframework.com/2020/moving-habitat-to-nap-0.4/

[Kyrah Abattoir]

2 hours ago, Gabriele Graves said:

Thanks for the info about the draw calls.

That's my point, it isn't a problem that is going anywhere.  I am fully BoM and so are many other Maitreya users, however we still want the alpha cuts for reasons I have outlined in many other topics touching on this.  It doesn't matter how much better alpha layers perform, many, many people are not going to update to a body without alpha cuts.  That is the one of the big mistakes that Slink made and now support for their bodies is vanishing.

I mean... LL could just "tell" those creators to make the move "or else".

Of course that is assuming LL has any kind of spine for this.

[Gabriele Graves]

27 minutes ago, Kyrah Abattoir said:

I mean... LL could just "tell" those creators to make the move "or else".

Of course that is assuming LL has any kind of spine for this.

It would be a pretty unprecedented move.  A technical limitation would be difficult to do without affecting lots of other content and has the added danger of killing the golden goose of the fashion economy.  I probably wouldn't stick around, there is only so much going back to square one a person can take and that would be a pretty massive one.

Doing it via policy would need policing and we have all seen how effective that is in other areas as it needs a serious investment to provide enough staff to make it effective.

If it didn't affect the bodies in people's inventories then people just wouldn't upgrade.  M. Lara still has lots of v4.1 people from a couple of years back, heck there are still people using v3.5 from at least 4 years ago.

So yeah they "could" do it.  I doubt that will happen personally considering the road behind us.

Edited April 14, 2021 by Gabriele Graves

[Coffee Pancake]

It's Lab policy that everything should work forever.

I would love to see their list of weirdo edge case objects that require special attention.

[Kyrah Abattoir]

10 hours ago, Coffee Pancake said:

It's Lab policy that everything should work forever.

I would love to see their list of weirdo edge case objects that require special attention.

Policies are just rough guidelines. At some point they won't be able to dig this hole deeper.

[Gabriele Graves]

At some point SL will close too but it's been wild ride so far!!!

[Cyrule Adder]

  

  

On 4/12/2021 at 10:29 PM, Gabriele Graves said:

Sorry but if you think that the Kemono body proves that 30K looks as good as a M. Lara, you just don't see what myself and the rest of the market does.  This is what I mean about the subjective difference between what one group of people would be happy with versus another.  I remain to be convinced it would be enough to recreate M. Lara indistinguishably from the 176K it is today.

The majority of AAA games will have hero characters of 15k to 40k polygons on average. This is for the entire body, including heads, hands, feet, and clothing. Higher polygon counts do not mean quality, nor does it make a mesh look "smoother". Especially when we concider that some MMOs like final fantasy, Guild wars, and so forth will have characters of polygonal coutn no higher than 20k in total. And despite the increase of technology, the polygon counts have not changed much. Not because GPUs cannot support them. But because the methodology of use them as you need them.

The issue with Maitreya's 176k polygon count (with no LODS) is that most of the polygons are absolutely not necessary.
Does it lend to the smoothness of the model? No. If I can draw a single line, and it's able to tangent 4 or 5 polygon loops with barely a noticable change, than the excessive polygons are unnecessary.

The second issue Maitreya's massive polygon count comes with the problem of micro-triangles. Where triangles can become smaller than a pixel. The problem with these is that they may have no impact on the final image or cause the same pixel to be redrawn multiple times, and the GPU still has to do all the math, and lighting for it. For a single player? This is nothing. But for how popular the body is. You can easily have 12 maitreya users on screen at a time... and in many cases going up to nearly 80 which will start dumpstering people's computers no matter how good the hardware is.

 

TLDR: a 30k model will look just as good as a Maitreya.

 

On 4/14/2021 at 3:45 AM, Gabriele Graves said:

It would be a pretty unprecedented move.  A technical limitation would be difficult to do without affecting lots of other content and has the added danger of killing the golden goose of the fashion economy.  I probably wouldn't stick around, there is only so much going back to square one a person can take and that would be a pretty massive one.

Doing it via policy would need policing and we have all seen how effective that is in other areas as it needs a serious investment to provide enough staff to make it effective.

If it didn't affect the bodies in people's inventories then people just wouldn't upgrade.  M. Lara still has lots of v4.1 people from a couple of years back, heck there are still people using v3.5 from at least 4 years ago.

So yeah they "could" do it.  I doubt that will happen personally considering the road behind us.

Yeeaaaaah. Unfortunately the most they can do is set some guidelines, but can't really enforce it without hurting themselves. And that might not change much. Especially since I can name at least 100 items on the market place where a screw no larger than 3cm has a 1024x1024 texture and roughly 3k polygons. These objects being advertised as HD.

Edited April 16, 2021 by Cyrule Adder