Jump to content

WTF? Avatar Draw Weight noncommutative (or: Pick a number, any number)


You are about to reply to a thread that has been inactive for 2170 days.

Please take a moment to consider if this thread is worth bumping.

Recommended Posts

Back in the day there was "Avatar Rendering Cost" but that flawed metric was replaced and got a new name, "Avatar Draw Weight".

In the past few months I've been trying to pay attention to how difficult it is for others in a crowd to render my avatar, so I've been idly checking my Avatar Draw Weight. I noticed it fluctuating wildly for no apparent reason. Finally I spent some time trying to track down what is happening.

There are several ways to monitor Avatar Draw Weight, including "Advanced / Performance Tools / Show Draw Weight for Avatars", "Develop / Show Info / Show Avatar Render Info", and by script using OBJECT_RENDER_WEIGHT. The viewer settings appear to show the same numbers, but the script number is different, but displays the same weirdness that I'm carping about. To wit:

 

1Draw weight at login54,504
2after removing mesh hair18,924
3adding back the same hair47,928
4removing a mesh bow tie52,869
5removing the hair again17289
6adding back the bow tie18924
7adding back the hair51935
8removing the bow tie52,869
9removing the hair17289
10adding back the hair29935

 

Note that  steps # 1, 3, and 7 represent the same set of attachments (and yes, they're all fully rezzed), but the weight varies: 54.5K, 47.9K, 51.9K.  More dramatically, steps 4, 8, and 10 represent the hair without bow tie with weight varying between 52.9K (twice) and 29.9K. At this stage I can teleport all around and the weight stays 29.9K, but if I login again it's back to 52,869.

What's going on here? Is the difficulty of drawing my avatar really changing in this weird way -- in which case I could maybe discover an optimal order of adding and removing attachments for each outfit -- or is this metric just completely useless?

Link to post
Share on other sites


Qie Niangao wrote:

Back in the day there was "
" but that flawed metric was replaced and got a new name, "
".

In the past few months I've been
trying
to pay attention to how difficult it is for others in a crowd to render my avatar, so I've been idly checking my Avatar Draw Weight. I noticed it fluctuating wildly for no apparent reason. Finally I spent some time trying to track down what is happening.

There are several ways to monitor Avatar Draw Weight, including "Advanced / Performance Tools / Show Draw Weight for Avatars", "Develop / Show Info / Show Avatar Render Info", and by script using
. The viewer settings appear to show the same numbers, but the script number is different, but displays the same weirdness that I'm carping about. To wit:

 
1
Draw weight at login
54,504
2
after removing mesh hair
18,924
3
adding back the same hair
47,928
4
removing a mesh bow tie
52,869
5
removing the hair again
17289
6
adding back the bow tie
18924
7
adding back the hair
51935
8
removing the bow tie
52,869
9
removing the hair
17289
10
adding back the hair
29935

 

Note that  steps # 1, 3, and 7 represent the same set of attachments (and yes, they're all fully rezzed), but the weight varies: 54.5K, 47.9K, 51.9K.  More dramatically, steps 4, 8, and 10 represent the hair without bow tie with weight varying between 52.9K (twice) and 29.9K. At this stage I can teleport all around and the weight stays 29.9K, but if I login again it's back to 52,869.

What's going on here? Is the difficulty of drawing my avatar really changing in this weird way -- in which case I could maybe discover an optimal order of adding and removing attachments for each outfit -- or is this metric just completely useless?

How many other attached objects are you wearing? I've noticed similar effects depending on whether or not my shoes, etc. were visible on my screen. Also the number only updates when you change what you're wearing. It's a "snapshot" figure, basically.  If you get a value based on a zoomed-in view that doesn't render some objects it'll stay that way until you make a change. There may also be LOD issues at different camera distances too.

ETA - the real X factor seems to involve sculpts, which makes sense because they have to be calculated by the viewer instead of reading metadata.

Link to post
Share on other sites

Yeah, no, that's not the thing here. Yes, what's visible and zoomed-in-on at time of measurement does affect the results, but I've found that under conditions strictly controlled to be comparable, the metric can be as near to random as makes no difference.

The numbers below are with no other attachments at all, from a fixed (default) cam view. That html table was too difficult to do again, so here are the numbers for my bare avatar with no other attachments, no clothes, no tattoos:

  1. "Naked" login: 1,000
  2. + same mesh hair: 55,430
  3. + bow tie: 38,215
  4. - hair: 2,635
  5. + hair: 6,717

For the very same view of the very same attachments (steps 3 and 5), the weight varies by a factor of five. What is more, adding an attachment (in step 3) reduced the draw weight by 30%. I mean, that's worse than being noncommutative: as a metric it's not even ordinal.  

How is it possible to make any use of such numbers?

Link to post
Share on other sites


Qie Niangao wrote:

Yeah, no, that's not the thing here. Yes, what's visible and zoomed-in-on at time of measurement does affect the results, but I've found that under conditions strictly controlled to be comparable, the metric can be as near to random as makes no difference.

The numbers below are with no other attachments at all, from a fixed (default) cam view. That html table was too difficult to do again, so here are the numbers for my bare avatar with no other attachments, no clothes, no tattoos:
  1. "Naked" login: 1,000
  2. + same mesh hair: 55,430
  3. + bow tie: 38,215
  4. - hair: 2,635
  5. + hair: 6,717

For the very same view of the very same attachments (steps 3 and 5), the weight varies
by a factor of five
. What is more,
adding
an attachment (in step 3)
reduced
the draw weight by 30%. I mean, that's worse than being noncommutative: as a metric it's not even
ordinal.
  

How is it possible to make any use of such numbers?

What viewer are you using? I'm just not seeing anything close to that level of variation and have found values to be pretty consistent for similar outfits.

Link to post
Share on other sites

I'm using the new Linden viewer, 3.8.1 (303230), but I've been seeing whacky sequence effects ever since I started trying to watch my numbers, for the past few weeks, while I was using a version (or two) of the Experience release candidate viewer.

This hair, though supposedly "mesh", seems to have a heck of a lot of little sculpties in it (along with some flexi boxes and god only knows what all else) so the observation that sculpties are an "X factor" may be apt.

Link to post
Share on other sites

I've seen these kind of variations also in the past.  

Could some kind of wierd object occlusion be affecting the count which I wouldn't think could happen?

My guess actually is that the math being done by the Server is somehow failing to update in the viewer.

Link to post
Share on other sites
You are about to reply to a thread that has been inactive for 2170 days.

Please take a moment to consider if this thread is worth bumping.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...