Selling the complexity number

[arabellajones]

Look, I know the number for avatar complexity isn't a perfect measure, and might change with Project ArcTan and can vary enormously depending on whether an item is attached or rigged, but why does it never get mentioned? You can't search for a low-complexity version of an item, and hardly anyone bothers to mention a number in the item page they create.

Oh, and I know it's not the same as memory usage, which matters rather a lot for vehicles and smooth sim crossings. Some high-complexity components, such as flexiprims, can be pretty memory-efficient, but there's no way out of having both high complexity and high memory use from a mesh with a high triangle count. 

High-complexity outfits can be a nuisance for other users, but we can set the complexity limit in the viewer, and it can make a huge difference (I don't have such a slow system. but it doesn't take many avatars with over 200k complexity to slow things down). Anyway, it's our choice there.

But why do the Marketplace and the sellers make it so difficult for customers to choose?

[Alyona Su]

6 minutes ago, arabellajones said:

Look, I know the number for avatar complexity isn't a perfect measure, and might change with Project ArcTan and can vary enormously depending on whether an item is attached or rigged, but why does it never get mentioned? You can't search for a low-complexity version of an item, and hardly anyone bothers to mention a number in the item page they create.

You answer your own question: it's an unreliable measure. A friend and I were standing together, using the same viewer, the same and the same Windlight and graphic settings. We each saw different numbers for each of us. Not only is it a unreliable number for accuracy on it's own, it isn't even a reliable number between the same viewer and setting for different people.

[Bitsy Buccaneer]

Doesn't computer power figure into it too somehow, like the cost to render for your particular set-up? I think I remember reading somewhere that it wasn't a universal number, so comparisons between the numbers will be misleading if they come from different people.

(As well as the problems in the formula itself.).

[arabellajones]

1 hour ago, Alyona Su said:

You answer your own question: it's an unreliable measure. A friend and I were standing together, using the same viewer, the same and the same Windlight and graphic settings. We each saw different numbers for each of us. Not only is it a unreliable number for accuracy on it's own, it isn't even a reliable number between the same viewer and setting for different people.

Perfection and reliability aren't the same. And how high a number, how big a difference, are you talking about?

[arabellajones]

I remember, when advanced lighting came out, that the new code, we were warned, could slow rendering. Well, it depended on the hardware and graphics drivers. Switch on shadows, and you do get much slower rendering. I was using an already-old nVidia 750. and using ALM was faster.

But none of the documentation of the calculation appears to invoke the computer systems. It's about the data, things such as triangle counts and texture sizes and LOD distances. Some details of the calculation have changed, and the changes aren't well-documented. Long, narrow, triangles used to be a killer for complexity, now it uses a simple triangle count. At least, it seems to.

Some people don't seem to realize how LOD switching works for a rigged mesh, and how that can give different results, depending on the avatar wearing something. It's still nothing to do with graphics hardware and drivers. That will affect where you set the Max. Complexity so you get acceptable frame rates, but that's entirely your choice.

[Kyrah Abattoir]

1 hour ago, arabellajones said:

But why do the Marketplace and the sellers make it so difficult for customers to choose?

To put it simply, because it is not in their interest to create more ways for other creators to compete against them.

 

Creators who take the extra time to optimize their modems, produce good lod models, reduce texture usage, or try to script conservatively will usually pitch those aspect in their product description. And then you have those that will simply lie about it.

[Mollymews]

i use the various complexity numbers to compare like for like (relatively) as they render on my computer.  How they might render on some one else's computer I can't do much about

all I can hope for is that when I choose say trees that perform better than other trees on my computer then this will be true for other peoples computers, win some lose some, overall   

[Alyona Su]

1 hour ago, Bitsy Buccaneer said:

Doesn't computer power figure into it too somehow, like the cost to render for your particular set-up? I think I remember reading somewhere that it wasn't a universal number, so comparisons between the numbers will be misleading if they come from different people.

(As well as the problems in the formula itself.).

Which is another reason the OP suggestion is nonsensical: "Hey! I bought this and the Complexity Number you say it is is a LIE! EVERYONE DO NOT BUY THIS!!!". 

Adding complexity numbers to merchants presales information; what could possibly go wrong?

Edited January 20, 2020 by Alyona Su

[ChinRey]

1 hour ago, Bitsy Buccaneer said:

Doesn't computer power figure into it too somehow, like the cost to render for your particular set-up?

No, it does not take that into account. Here's the formula:

http://wiki.secondlife.com/wiki/Mesh/Rendering_weight

[Bitsy Buccaneer]

17 minutes ago, Alyona Su said:

Which is another reason the OP suggestion is nonsensical: "Hey! I bought this and the Complexity Number you say it is is a LIE! EVERYONE DO NOT BUY THIS!!!". 

Adding complexity numbers to merchants presales information; what could possibly go wrong?

Yeah, that would happen too. And probably more drama than we're able to imagine. No, than I'm able to imagine, you have more skill there than I do

I wonder if there is any number/sum of numbers which would work as a standardized indication of render weight? I'm not creative enough today to puzzle it out, but it would be useful.

 

1 hour ago, arabellajones said:

Long, narrow, triangles used to be a killer for complexity, now it uses a simple triangle count. At least, it seems to.

I'm used to hearing about long, narrow triangles as a physics issue, complicated by whether or not it's analysed or not analysed on upload. If no one addresses that here (I don't understand it well enough to venture), ask in the mesh creation forum. Someone there will likely be able to help.

[ChinRey]

1 hour ago, Mollymews said:

i use the various complexity numbers to compare like for like (relatively) as they render on my computer.  How they might render on some one else's computer I can't do much about

all I can hope for is that when I choose say trees that perform better than other trees on my computer then this will be true for other peoples computers, win some lose some, overall   

Display weight for rezzed objects and classic avatars is fairly reliable, or at least the errors are consistent enough it gives a reasonably good estimate of the relative difference between two items. It's only when fitted mesh is involved it is totally off.

[Dean Haystack]

Instead of Complexity, how about the the VRAM usage, number of Triangles, number of Scripts with their time and memory usage?

[ChinRey]

26 minutes ago, Dean Haystack said:

Instead of Complexity, how about the the VRAM usage, number of Triangles, number of Scripts with their time and memory usage?

Or complexity and all of those? Script time and memory usage are a bit tricky though since they aren't necessarily constant.

But the main problem with the idea is that they aren't selling points. I used to include much of that info in my MP descriptions but gave up since it turned out few people were itnerested and it didn't generate any extra sales worth speaking of.

[Wulfie Reanimator]

On 1/20/2020 at 5:45 PM, ChinRey said:

Display weight for rezzed objects and classic avatars is fairly reliable, or at least the errors are consistent enough it gives a reasonably good estimate of the relative difference between two items. It's only when fitted mesh is involved it is totally off.

This is still false.

A rezzed mesh object that has less textures, VRAM, links, and triangles can have higher complexity than other objectively worse items.

[OptimoMaximo]

6 minutes ago, Wulfie Reanimator said:

This is still false.

A rezzed mesh object that has less textures, VRAM, links, and triangles can have higher complexity than other objectively worse items.

Also, there is no difference whatsoever between a mesh rigged for a classic avatar and one using the fitted mesh system if not in the number of skinned joints. 

[ChinRey]

14 minutes ago, Wulfie Reanimator said:

This is still false.

A rezzed mesh object that has less textures, VRAM, links, and triangles can have higher complexity than other objectively worse items.

I suppose we can discuss what "reasonably good estimate" means.

[Dean Haystack]

On 2/3/2020 at 1:14 AM, ChinRey said:

(...)Script time and memory usage are a bit tricky though since they aren't necessarily constant.(...)

I know.
So let me elaborate.
To stay with forum rules I'll abstain from naming anything, but, a while ago I bought a vehicle.
This vaguely described vehicle made by an undisclosed maker is said, in the marketing material, to be built to withstand the worse lag of SL with super efficient scripts.

Vehicle A: [84/84] running scripts, 5328 KB allowed memory size limit, 0.385569 ms of CPU time consumed.]
You'll be lucky if you get 2~3 crosses before things go horribly wrong.

For reference, here's another vehicle:
Vehicle B: [25/25] running scripts, 1600 KB allowed memory size limit, 0.472982 ms of CPU time consumed.]

And another one:
Vehicle C: [11/11] running scripts, 176 KB allowed memory size limit, 0.142729 ms of CPU time consumed.]

Note: All measurements taken while moving.

Now, I know I can take those 10 times and get 10 different results.
But am I gonna come here as ask for just the number of scripts to be listed on the MP? No.
Because I can guarantee someone would immediately jump in here to "let me know" script performance is measured in units of time, not number of scripts, and I could have 1 script that was heavier than another 100 scripts combined...
Ence why I said:

On 2/3/2020 at 12:39 AM, Dean Haystack said:

number of Scripts with their time and memory usage

I believe the more information the better, since less isn't gonna be more reliable anyway.

[Semirans]

The big item for me on complexity is Hair.  I have some gorgeous flighty styles that I adore, but they are 45k-65k complexity.  Other styles that are great, but not as complex run about 2k (these are all mesh hair). On a side not and only somewhat related are the Li of mesh blanks.  I have FP items that are 100Li and others that are less than 10. I would love to rez the clothing in my store so people can see it on the actual garment before buying, but at 100Li each it is prohibitive

[Alwin Alcott]

11 minutes ago, Semirans said:

I would love to rez the clothing in my store so people can see it on the actual garment before buying, but at 100Li each it is prohibitive

don't rezz it but use limited permissioned demo's , people should WEAR clothes to try, with seeing it's never sure if the rigging is right.