The importance of optimization

[ChinRey]

I know, I know, I keep plugging this and people are probably pretty bored by it by now! But this one has beautiful roses, maybe that helps?

 

Here are two rose bushes. They've been scaled down to 1x1x1 m as a handy reference size:

The one to the left is a traditional crossed sheets plant sculpt. The one to the right is optimized mesh. According to the calculated figures, the sculpt one takes 15 times as long to download and it gives the server six times as much work to handle the physics and the client computer four times as much work to handle the graphics.

 

Closer view of the sculpt bush:

and the mesh:

 

And then the LoD. This is how they look at 16 m distance with graphics prefs set to standard mid in the official SL viewer:

 

Even further away. Not sure what the actual distance is but it's right before they're eliminated by the 128 m draw distance:

 

As I said, these bushes are scaled down to 1x1x1 m. With a more realistic 1.5 m the download weight increases to 1.7 for the sculpt and 0.2 for the mesh. At 2 m size the sculpt's download weight is 2.0 while the mesh stays at 0.2.

A prim with he same shape as the sculpt - made the way they used to teach you at builder schools - would have 0.2 download weight, 0.4 physics weight, 2.0 server weight and 976 render (or display) weight and since it would be an old style build, the prim count would be 4.

A standard poorly optimized mesh could probably be done with the same download, physics and server weight as the optimized one but done the way they are usually done, the render weight would have been higher and of course, you'd have to compensate for the poor optimization by LoD butchery so it would look disastrous even at a moderate distance.

Edited December 13, 2017 by ChinRey
Typo in the title

[Penny Patton]

In summary; If content creators optimized their work we'd all be enjoying much better framerates or able to run SL at higher graphics settings. SL would look better AND run better, even on older/less powerful computers.

It sounds like LL actually plans to do more to discourage unoptimized content this year. They're introducing VRAM tools that will work like the render weight features (turning avatars with excessive VRAM consumption into jelly dolls) and also let you inspect avatars, attachments, and in-world objects to see their VRAM use.

There's also talk of redoing Land Impact calculations to encourage use of proper LOD models. A change that might cause a lot of short term headaches for people, but in the long term will lead to a much more enjoyable SL experience.

[ChinRey]

On 4.1.2018 at 6:18 AM, Penny Patton said:

There's also talk of redoing Land Impact calculations to encourage use of proper LOD models.

That's good news. I hope they don't throw out the baby with the bath water though. What we see on the hypergrid where LoD models don't affect the land impact at all, is that people tend to make too complex LoD models and that is just as bad or possibly even worse.

Edited January 5, 2018 by ChinRey