Blog Entries
The Myth of Avatar Complexity Tags: myth avatar complexity



It started on Second Life years ago as a virtual witch hunt that pointed the blame for "lag" at users rather than at poor world foundation software and servers.  It wasn't long before the debunking started and that attempt pretty much died.


Realizing that attempt failed, it wasn't long before they came up with a new, better, modified AVATAR COMPLEXITY figure, which is of course touted as being more accurate, useful, and definitely not a witch hunt looking to mis-place blame for system-created lag. 

But is there any validity and truth behind avatar complexity figures?   The answer is both yes and no.  The first thing we should notice is those complexity figures are excessively high. They are (supposedly) based on accurate rendering figures... but savvy users are aware that rendering figures do not equate to lag.   For those unaware of this principle, let's consider a basic example.

Create a simple box prim.  Wear it.   Your AC will jump a surprising 405 points... for a simple, non-functional cube!

Now, go to the Features tab of that cube and turn it FLEXIBLE and re-wear it.  Your AC will now jump to 409... just 4 points over a basic cube.

So what does this mean in actual daily use?  We established very long ago that simple prim count does not "lag", whether that prim is stable on a sim or worn on an avatar.   It did at one time, but that was code-fixed years ago.   The reality is that simple prims don't lag, at all.   An oft-quoted experiment was when 140,000 prims were rezzed on an empty Open Sim region and the testers observed zero "lag" of any sort.  

Elf Clan ran its own tests on a brand new, empty SL region on which we rezzed 12,000 prims, then put sit scripts in 5,000 of them.  Neither situation presented any discernible, measurable "lag" whatsoever.  So the prim-lag myth was busted, totally, by two separate documented tests.


As stated above, the Flexi shows a mere 4 points above a non-flexi prim.  Hardly worth noting.   However it has been long-established that flexible prims are one of the most lag-inducing items on virtual worlds.  How lag-inducing are they?  On a sim where they conducted flexi tests, they set out 12 "blankets" hanging from a clothes line.   At a touch the blankets could be turned rigid, or flexi and blowing in the wind.   The result:  When all 12 blankets were turned to flexi the sim lagged almost to a standstill. 

Now admittedly Flexi handling has improved somewhat... but it is still one of the most lag-causing things on VR.   The point is that the Avatar Complexity figure does not reflect this factor in its ratings.  It counts the flexi factor a mere 5 points out of 409... almost identical to a non-flexi cube.   According to "Avatar Complexity", they are both pretty much the same when it comes to avatars causing "lag" on a region.  However we know for a fact that the inclusion of flexis on an avatar are heavily lag inducing.   Obviously Avatar Complexity comes several cards short when it comes to actually measuring lag-causing factors.


We were curious, so we took some ratings on three different avatars, on two different grids:

1)  The basic Dwagon created by Snoots Dwagon, a completely primmed avatar head to toe.  This avatar consists of a total of 427 tortured prims. 

2)  A fully-armored warrior, the biggest of the baddest.  Wearing highly-detailed armor head to toe, this avatar comes in at over 1,700 prims.

3) A standard female avatar with flexi hair, skirt and reasonable jewelry.   Nothing special... just the kind you'd find at most clubs (in fact, clubs were where we took multiple ratings).


The following were single avatar tests. In cases of numerous avatars, lag will usually be very discernible.  Simple truth:  avatar presence lags, regardless of what they're wearing.

The dwagon consisting of 427 almot constantly-moving prims, displayed an AC of 70,155.  That's our base-line figure. The avatar produced zero discernible lag.

The warrior consisting of over 1,700 prims displayed an AC of 203,919... roughly 2.9 times higher AC than a dwagon.  This would be expected; the warrior contained more than 3 times the prims.  Almost a direct correlation.  This avatar produced zero discernible lag.

Standard female avatars dressed in regular clothing with flexi hair and dresses almost across the board hit "Jelly Doll" stage... which means their AC count was so high their avatars became shaped mono-color blobs to onlookers.  When the Jelly Doll function was turned off and AC checked it was discovered these avatars commonly exceeded an AC rating of 300,000 and more.  As there were multiple such avatars on the region, lag was definitely discernible.  But was this because of AC ratings... or simply because of region servers inefficiently handling avatar presence?



We created an "insane" avatar by building a 1000-prim linked object, duplicating it and attaching it to head, arms, chest and legs, with the result of an avatar wearing thousands of cubed prims.  The result?  Zero discernible lag.



(First, let me encourage readers to check out the comments section following for a realistic viewpoint of using avatars in certain situations.  Quite valid.)

The Avatar Complexity system is as bogus a witch hunt as it's always been.   It does not determine the actual affect an avatar has on sim server performance, and does not take into account standard, everyday, common usage of virtual world systems. 

Do avatars "lag" virtual worlds?  It is safe to state that avatars are the most lag-inducing item on VR (besides server issues).  There is no doubt of that.  But it's not because of avatar complexity; it's because the way in which Second Life style virtual reality coding handles avatar rendering.  The primary source of lag is inefficient server and viewer code, not the avatar you choose to wear.  However, the avatar we choose to wear can affect that poor server and viewer code.  It's a Catch-22.

Now yes, if we all took off our hair and clothing and walked around naked, the servers wouldn't lag as badly.  But they'd still lag.  Every time someone teleports in or out, every time the wind blows a flexi-based tree, every time the system glitches for no discernible reason, there would still be lag.  

But the reality is virtual worlds revolve around avatars.   To try and limit those avatars by using an inaccurate rating system is goofy.  Basically, it is perfectly okay to totally ignore the Avatar Complexity readings.  They have very little to do with the reality of lag on virtual worlds. 

The reality is this:  avatars with far lower AC ratings can "lag" far worse than avatars with high AC rating.   That fact alone invalidates the AC system as a useful tool.



Many, many years ago, AO devices lagged.  Since then the server software has been updated and the AO devices have been improved.  Tests have been run, extensively.   So we can stop that witch hunt as well.  Tinies require AOs to function.  So do many other avatars.  The best thing to do to prevent an event from lagging:   reset the region prior to the event, and make sure region scripts are low-lag designed.  Use textures efficiently. In the case of an event that is expected to have a large audience... if possible put the performance group on an adjoining region.



Ignore Avatar Complexity. Ignore AOs. Just enjoy yourself in your activities.  Avatar Complexity was started by Linden Lab and foisted off on other virtual worlds.  It is a bogus "tool" that does not correlate to region lag issues.

Sim content and avatars are the life of virtual worlds.  "An empty sim is a fast sim" may be true... but the people on that lifeless sim won't be having much fun.  Avatars will dress according to their preferences. (See addendum below in comments.)  This is not to say event hosts can't request avatar simplification for the good of the event;  I just recommend not using the questionable and irrelevant readings of "Avatar Complexity" as a reason. 

-- Wayfinder




This website is powered by Spruz