Does Linden Lab's SL Avatar Complexity Metric Make Any Sense?

SL graphics expert NiranV pointed out another mysterious finding in Cassie Middle's Mesh Body Survey (Google Spreadsheet available here):

"Vertices and Triangles do not change with neither skins nor alphas worn, unless alphas are separate wearable onion layers. Same for tattoos. Only VRAM changes," he begins, for background.

 

"But we already see something very important in this list:

 

"Lets Take Belleza, as example:

• 500,000 triangles
• 300,000 vertices
• Almost 60 megabytes VRAM

... but roughly 6000-7000 complexity.

 

"Now lets take Slink Physique Redux [pictured here]:

• 53,200 vertices
• 80,800 triangles
• 28 megabytes VRAM

Yet despite using far, far less resources than Belleza, the complexity total for Slink Physique Redux is 6,000 -- about the same as Belleza.

 

"So Linden Lab's complexity metric is bullshit?" I ask him.

 

"Exactly," says NiranV.

 

Maybe there's something he's missing, but I doubt it. Is there another explanation? Take a look at Cassie's spreadsheet, and see for yourself.

 

Update, 9:00pm: On the spreadsheet, the original vertice/triangle/complexity numbers reported for Slink Physique Redux were inaccurate, but have since been corrected. I updated them above and deleted a line by NiranV analyzing those old and incorrect (as it turns out) numbers. His basic analysis, that the complexity metric seems totally inconsistent, still stands.

Comments

[Galatea]

What about different LoD levels? I know when it comes to land impact for object meshes, you can have incredibly detailed max-level models, but with a much lower land impact, if they have lower complexity versions to fall back to. Do wearable bodies have a similar scheme? If so, it would make sense that the complexity rating would be lower if those less complex versions are there, even if the max LoD version is very complex.

[Pincess Venusia]

I believe that this kind of inquiry leads to nothing, except to prove that certain bodies are more beautiful than others, and that the most popular is not the most beautiful as I have always said, but otherwise if Second Life had stayed at their basic body, I think it would be a long time that second would be closed, it's precisely because the mesh bodies appeared that the world must remain economically profitable , and all of a sudden you would have to chase all that, have a mesh body, but not too much, in passing when I hear that wearing certain bodies would be like doing an attack ddos on the other, if I go in TESO, the bodies players are much more, beautiful, much more complex and my computer is still the

[Susan Wilson]

Hey Wagner, if there was a well-optimized body that looked as good as my Maitreya body and my full wardrobe of Maitreya clothing fit it I would buy it. You're right that there is a problem, but I don't experience lag in SL so if fixing that problem means I have to dump my wardrobe - it's not going to happen.

[Amanda Magick]

I think everyone needs to understand lag and optimization in SL. Then we all could try to do our part. But it seems to be out of our own control for the most part. Anyway I am always willing to learn.

[Soda Sullivan]

"I believe that this kind of inquiry leads to nothing, except to prove that certain bodies are more beautiful than other"

That makes absolutely no sense. What these numbers prove is that, with a little knowledge and work, you can make a very beautiful body that will use almost a third of the resources a poorly designed beautiful body will. Hats off to Slink and Signature. However all of these are going to give even the best gaming computer a hard punch right in the VRAM.

I would love to see a similar comparison done with mesh heads. I bet the results would be terrifying.

[Mondy]

Why do you always refer to Niran as a graphics expert when he's told you himself in replies to other aticles he isn't ?

Knowledgable in the area because of his work on viewers, sure. But you never seem to ask any viewer developers who use the same method of calculating things as Linden Lab does.

[Nalates]

Vir Linden is working on ARCTan... the coming ARC/ACI + Land-Impact revision. Everyone... well, some of us... know the current ARC/ACI + Land Impact weighted for older systems and days of a slower network needs updating.

Niran jumped the gun on ARC/ACIrevison and uses a different accounting system in Black Dragon. Something Niran thought was more realistic. It probably is. But, Niran doesn't have access to the data the Lab has been collecting on its users' machines over the past year+/-. So his system and the Lab's future system are very likely to be different.

However, since Niran's numbers are higher/bigger those regions that use complexity detectors tend to ban Black Dragon users because of the viewer reporting higher ACI.

If Niran's system were adopted, we would probably see loads of prims returned from regions. The Lab has to figure out a system that avoids that or implement a new one while simultaneously raising the region/parcel prim limits.

[TDGunner]

When will this topic go away? I think only Wagner and NiranV are the only ones that find this: a) interesting, b) relevant. The game works fine for me on my mid range laptop regardless of where I am in-world. On my desktop, its a blip.

The rest of us do not care about this so let's move on.

[Madeline blackbart]

Tbh the Arc system was never really accurate. Not 8 years ago when I first joined and not now either. Granted it's more accurate then it was in the past but not much more accurate.