How is Avatar Complexity Doing?

More and more viewers are adopting the latest Linden additions to the viewer. The RLViewer just added the Avatar Complexity feature. So, more and more people are noticing the pop up messages generated by the Avatar Complexity feature. Merchants are starting to respond to the feature also. I suspect for now that concern is anticipatory rather than reactive.

Never Stop Dreaming.

Never Stop Dreaming

I have always suspected the majority of Second Life™ users have a level of concern about the amount of lag they add to the system and take some steps to reduce it. Of course most of us want to avoid lag. A few entitlement types put off on the Lab. But, I think that is a small percentage. 

A significant number of us balance lag against appearance, avatar and scene. I don’t mind pushing my frame rates into the low teens to have a pretty render. I will push my Avatar Complexity Information value above 200k for a picture. For daily wear I try to keep it under 100k.

The question is always how many care at what level? For now we don’t know. But, some merchants are already considering how the new feature may impact their sales. One of those is Marine Kelley, the developer of the RLV viewer and creator of the Maison de la Marquise (URL NSFW – requires marketplace adult setting) line of videos, clothes, and toys. The marketplace shop is rather limited and seems to be older items. You will find newer and more of her stuff in-world in regions named Pak and Lineside.

Marine has been examining the algorithm used to calculate ACI – Avatar Complexity Information, the numeric value. It seems transparent and sculpty items produce high ACI. But, according to Marine transparent items don’t really add to the render cost… I suspect it isn’t that simple. But, neither is Marine so, somethings up.

I also know the Lab spent a lot of the ACI development time tweaking the algorithm and considering the impact on the SL environment. So, I suspect where and how transparent textures are used is significantly complicated and while I accept Marine’s statement as accurate, the Lab also had reasons for building in the high ACI cost for transparency. See: RLV 2.9.18 and scroll down to On a side note… for more information.

From what she writes the primary reason for the interest is Marine’s concern about viewers giving her products high ACI values. Many mesh products use transparencies to hide parts not in use. For instance; clothing layers on mesh bodies, expressions on mesh heads, genitals worn in a hidden state and various levels of arousal… all are dependent on transparencies. Many of Marine’s products make use of transparency. So, the ACI algorithm may unnecessarily be putting her at a disadvantage… or not. But, I’ll take Marine’s statements at face value.

The real point here is that Marine is concerned about the ACI value her products have. She was already concerned about render efficiency well before ACI and creating render efficient products. I think Marine is representative of other merchants that are starting to think about how ACI will affect their business. Others will follow as more people learn what ACI means. At some point we will reach a ‘critical mass’ and start to see events that limit ACI just as they do scripts. Then most people will  know or learn about ACI and in some measure consider how to keep it low.

So, is ACI accomplishing its purpose? Too soon to tell. But, it looks like it will.

5 thoughts on “How is Avatar Complexity Doing?

  1. I agree with Marine that the cost of *fully* transparent prims is largely overly exaggerated in the rendering cost (since modern GPU hardware and drivers can easily take them off the equation).
    About *semi* transparent objects, things are however quite different. See this NVIDIA developer article, for example:

    There’s also the fact that other costs, such as wearing a “Physics” wearable (AKA wiggling boobs and buttocks) don’t count more than a simple legacy shirt wearable… The old ARC cost (extrapolated from old hardware performances by LL) was in fact directly used, unmodified, for the new “avatar complexity” calculation, which is definitely suboptimal and, in most cases, plain bogus…

    I’d also like the attachments “surface area” to be taken into account for a larger part (some hair, mesh or not, can be especially costly, in *actual* FPS drop measurement when worn, and the equivalent surface area seems to reflect this cost pretty accurately).

      • Nope: all TPVs shall keep the same formula, as per LL’s request (written in a comment in their code).
        However, the Cool VL Viewer uses higher limits for the various default graphics settings; it also allows to use the surface area limit independently of the avatar complexity (LL’s code requires that the complexity is set before it takes the surface area limit into account). Finally, it adds a setting for a limit on the attachments geometry bytes (memory consumption).

      • I’m sorry but arc and avatar complexity is calculated differently. If you try it you’ll see the numbers give different results.

        • Don’t confuse my use of the name of a value (the noun) with my use of the same words as a description of a process.

Leave a Reply

Your email address will not be published. Required fields are marked *