Your avatar complexity is blah blah blah - Page 2 - SLUniverse Forums
Navigation » SLUniverse Forums > Virtual World Discussion > General SL Discussion » Your avatar complexity is blah blah blah


General SL Discussion Discuss topics related to Second Life

Sponsor:
LIONHEART - We Have Your Land
Reply
 
LinkBack Thread Tools Display Modes
Old 09-12-2018, 12:31 AM   #26 (permalink)
Senior Member
 
Natsuki Morigi's Avatar
Is this live?
 
Join Date: Nov 2009
Location: Tokyo ,Japan
Posts: 427
My Mood:
SL Join Date: 7 April 2007
Client: Black Dragon
Also , note, Black Dragon calculates Avatar rendering cost differently to the official LL viewer Firestorm. I believe the numbers come out quite a bit larger in Black Dragon

see for a bit of an explanation:
https://niranv-sl.blogspot.com/2018/...25-caring.html
__________________
---------------------------------
my latest Snapzilla pics

Last edited by Natsuki Morigi; 09-12-2018 at 01:27 AM.
Natsuki Morigi is offline   Reply With Quote
Old 09-12-2018, 01:44 AM   #27 (permalink)
The Purple
 
Chalice Yao's Avatar
Adoring goats.
 
Join Date: Dec 2007
Location: Somewhere purple, Germany
Posts: 9,108
My Mood:
SL Join Date: 20. January 2007
Client: NaCl
Quote:
Originally Posted by Natsuki Morigi View Post
Also , note, Black Dragon calculates Avatar rendering cost differently to the official LL viewer Firestorm. I believe the numbers come out quite a bit larger in Black Dragon

What Black Dragon does is put a lot more weight on the amount of triangles/geometry a mesh has - which actually makes a *lot* of sense, and is a thing LL (supposedly) wants to fix in their new calculations. Someday. Maybe. Which brings me to...

@Val: It explains why your numbers are way higher. It's hard to compare Black Dragon's to any other client's.

Also,
Quote:
Originally Posted by Adeon Writer View Post
That said, the rating itself leans on the leinient side, for example, the biggest source of abuse, excessive texture use, has almost no effect on complexity.
This is definitely one of the two big problems with the current calculations. Large textures barely nudge it higher than smalelr textures with efficient UVs.

The other problem is the aformentioned behavior regading triangles - current calculations by LL strongly punish complex far-away LODs, which is fine....but they downright ignore gargantuan geometries in the close-up LODs - which is the reason why you can have a buhzillion triangle close-up piece with a low render weight and a sane triangle piece with a high render weight - the former has its far-away LODs send to '1' and the other has a few triangles more further away. It is a big oversight.



EDIT:
It's also a reason why I've pretty much condensed my decision wether or not to get a piece down to two things: Texture VRAM usage and triangle count. I have tossed out a LOT of hair I used regularly and adored, simply because I noticed that they had horrid triangle counts. It hurt to clear that trash bin. :|



.
__________________
"Have you ever noticed that anybody driving slower than you is an idiot, and anyone going faster than you is a maniac?" - George Carlin
Chalice Yao is offline   Reply With Quote
1 User Said Thanks:
Old 09-12-2018, 03:57 AM   #28 (permalink)
Senior Member
 
ChinRey's Avatar
Status symbol
 
Join Date: Jun 2014
Posts: 170
My Mood:
Business: OPQ
Quote:
Originally Posted by Chalice Yao View Post
Also,This is definitely one of the two big problems with the current calculations. Large textures barely nudge it higher than smalelr textures with efficient UVs.
Do you have any data about that? I've always suspected the render cost calculation underesimates the impact of high resolution textures but I haven't been able to test it.

For those who are interested, here is the formula:
http://wiki.secondlife.com/wiki/Mesh/Rendering_weight
The page only mentions mesh and avatars but the same formula is used for all objects - prims, sculpts and meshes - and it also applies to rezzed objects. It's old info but there is no indcation that there have been any significant updates to it.

Some of the flaws:
  • The fitmesh LoD bug. The LoD swap distances are determined by the object's size and the render weight calculation gets that completely wrong. Typically it means the calculation assumes the fitmesh is rendered at lowest LoD while it is in fact rendered at highest. Some fitmesh makers discovered this bug early and rather than report it they decided to take advantage of it to make their items look less laggy than they are. Typically a fitmesh that does not try to take advantage of the bug will have an actual render weight (if the formula had been applied correctly) 5-100 times as high as the calculated one. A mesh that takes advantage of the bug may well have an actuall render cost more than 1000 times higher than what the number says.
This is the big one of course and it must be a nightmare for LL developers. When Vir Linden was asked about it, all he could say was that they were aware of it. As far as I know, they haven't found a solution yet.

There's more though:
  • The view distance cap is too high. The triangle count the formula uses is supposed to be based on the number of triangles in each LoD model weighed towards how wide an area each model is visible at. In addition to getting that competely wrong with fitted mesh, it also overestimates the total view range - it is only capped at 181 m. This means that the lowest LoD model is given too high a weight compared to the high one. It's the combination of this and the fitmesh LoD bug Chalice is talking about when she says the calculaction "ignore gargantuan geometries in the close-up LODs".
  • The calculation doesn't use the actual triangle counts for meshes. No, seriously it doesn't. It estimates the triangle count based on the size of the downloaded file. That means it has a tendency to overestimate the triangle count of low poly meshes and underestimate high poly ones.
  • The calculation does not take cpu load into account. I'm not sure if that is a flaw because it is hard to see how this can be baked into the render cost. But CoffeeDujour, Animats and others have done some tests showing that cpu performance can be just as important as gpu for client performance. It is not unusual for the graphics processor to while away the milliseconds waiting to hear from the central processor. This may be especially significant for sculpts and fitted mesh since they seem to be more cpu heavy than other objects.
  • The calculation does not take LoD factor into account. There is a question whether it should since it's up to the user to decide whether they want to increase their LoD factor. But it does skew the balance between geometry and textures. Increasing the LoD factor will increase the geometry load but not the texture load.
  • The calculation seems to underestimate the significance of high resolution textures. As Chalice said. I hope she has more info about this.
  • The formula does not know about alpha masking. Alpha masking was introduced after the formula was created so it's completely ignored, faces with alpha masking are treated as if they were fully opaque. A face that is made completely invisible with alpha masking should not add any gpu load at all whle one that is partly masked should add slightly more weight than if it was opaque.
  • The formula does not know about normal and specular maps. Again, those are features added after the formula was created so they are ignored. The maps are not even included in the texture count. Whether it should take these into account is a good question of course. They only apply if you have ALM switched on and that's your choice.
  • The web page does not mention fitted mesh at all. But that is probably just the web page. Fitted mesh almost certainly uses the same multiplier as rigged mesh.
  • The balance between the rigged (and presumably fitted) mesh, the flexiprim and the animated textures multipliers seems to be off. 1.2, 5 and 4 respectively. There doesn't seem to be any reason why a path based flexible mesh (which is what a flexiprim is as far as the gpu is concerned) should be that much harder to handle than a skeleton based one and a simple cycling texture changer should be easier for the gpu to manage than ever changing mesh geometry. There must be something wrong either with the formula or the code handling those features or - most likely - both.

Last edited by ChinRey; 09-12-2018 at 04:34 AM. Reason: Correcting a few typos and adding to factors I forgot
ChinRey is offline   Reply With Quote
1 User Said Thanks:
Old 09-12-2018, 05:51 AM   #29 (permalink)
The Purple
 
Chalice Yao's Avatar
Adoring goats.
 
Join Date: Dec 2007
Location: Somewhere purple, Germany
Posts: 9,108
My Mood:
SL Join Date: 20. January 2007
Client: NaCl
Quote:
Originally Posted by ChinRey View Post
Do you have any data about that? I've always suspected the render cost calculation underesimates the impact of high resolution textures but I haven't been able to test it.
There is a formula on the wiki page you linked, and I just verified it in the viewer code.

The formula for textures is as follows:
Each *unique* texture in the linkset adds 256 + (16 * (width/128 + height/128)) render weight.

The problem is that this scales linearly with the texture width and height due to them being added instead of being multiplied.

Here are the resulting numbers (*Without* the baseline 256 weight added per texture):
32x32 = 8
64x64 = 16
128x128 = 32
256x256 = 64
512x512 = 128
1024x1024 = 256

You can see how a 4 times larger texture only incurs double the penalty - and a small penalty to boot. So far so good, not quite optimal due to linear scaling, but fixable by changing the addition into a multiplier. In theory.
HOWEVER:
These values get added on top of a base 256, which makes things horribly bad as it completely demolishes the already linear-instead-of-exponential scaling: A 32x32 texture has a weight of 264. a 1024x1024 incures a weight of 512. A little twice that of a 32x32, despite being 1024 times(!) larger in pixelcount and memory.


Final numbers:
32x32 = 264
64x64 = 270
128x128 = 288
256x256 = 320
512x512 = 384
1024x1024 = 512





.

Last edited by Chalice Yao; 09-12-2018 at 06:00 AM.
Chalice Yao is offline   Reply With Quote
Old 09-12-2018, 05:57 AM   #30 (permalink)
Senior Member
 
Da5id Weatherwax's Avatar
I have no status
 
Join Date: Aug 2010
Location: Edinburgh
Posts: 1,467
Business: No Escape
Client: Official, Various RCs, Marines,Firestorm - for testing stuff. BlackDragon - For everything else.
Quote:
Originally Posted by ChinRey View Post
  • The web page does not mention fitted mesh at all. But that is probably just the web page. Fitted mesh almost certainly uses the same multiplier as rigged mesh.

This one actually makes sense since to any system resource there would be no difference. You're loading the same skeleton, the same mesh and its set of bone weights. The system doesn't care that the weights in mesh A are to one subset of bones in the skeleton and the weights in geometry-identical mesh B are to a different subset of bones in the same skeleton, and that's the only difference between vanilla-rigged and fitmesh-rigged.
Da5id Weatherwax is offline   Reply With Quote
Old 09-12-2018, 07:04 AM   #31 (permalink)
pushes vertices
 
Raindrop Drinkwater's Avatar
Shinyyyyy!
 
Join Date: Jul 2007
Location: London
Posts: 4,297
My Mood:
SL Join Date: 666
Business: Timetwist by A2NZ
Client: Official Viewer
Quote:
Originally Posted by Chalice Yao View Post
Also,This is definitely one of the two big problems with the current calculations. Large textures barely nudge it higher than smalelr textures with efficient UVs.

The other problem is the aformentioned behavior regading triangles - current calculations by LL strongly punish complex far-away LODs, which is fine....but they downright ignore gargantuan geometries in the close-up LODs - which is the reason why you can have a buhzillion triangle close-up piece with a low render weight and a sane triangle piece with a high render weight - the former has its far-away LODs send to '1' and the other has a few triangles more further away. It is a big oversight.
This is particularly aggravating. Objects that cheat the LODs with over-detailed highest LOD and one measly triangle at the lowest -- basically, objects done the lazy way -- end up having less land impact than objects with better optimisation (as few polygons as needed at highest LOD, while also retaining their rough shape at lowest LOD).

Not only do we spend more time and effort optimising our models, UV maps and texture sizes, we're actually punished for it.

So the grid ends up clogged with mesh pulled straight out of zBrush or Marvelous Designer with horrible topography, or "conversions" from high-poly objects designed for static renders that were never designed to be used as game assets.
__________________
Quote:
Originally Posted by Humps View Post
Your name is Rain, and that's amazing / At your posts I'm always gazing
I see your work and think 'she'll go far' which is why theres no need of an aide memoire
xxxxxx
My instagram
Raindrop Drinkwater is offline   Reply With Quote
1 User Agreed:
Old 09-12-2018, 08:34 AM   #32 (permalink)
Just another muppet
 
Sid II's Avatar
Eh.. mi mi mi mi mi?
 
Join Date: Jun 2015
Location: In the south of The Netherlands.
Posts: 1,751
My Mood:
SL Join Date: Spring 2007
Client: standard
Who cares about the calculations in the first place?
If people find that my avatar is to rich for their findings, they miss seeing me as I am.
I could not care less.
To me, it is mostly drama lama stuff to be honest.
Sid II is offline   Reply With Quote
1 User Agreed:
1 User Likes This:
Old 09-12-2018, 09:49 AM   #33 (permalink)
Senior Member
 
ChinRey's Avatar
Status symbol
 
Join Date: Jun 2014
Posts: 170
My Mood:
Business: OPQ
Quote:
Originally Posted by Da5id Weatherwax View Post
This one actually makes sense since to any system resource there would be no difference.

Yes it does makes sense. But the formula and the code handling it was made long before fitted mesh even existed and it is extremely unlikely LL updated it to account for fitmesh. So it will only have a multiplier at all if the code happens to recognise it as the same as the rigged mesh it knows about.


Quote:
Originally Posted by Chalice Yao View Post
The formula for textures is as follows:
Each *unique* texture in the linkset adds 256 + (16 * (width/128 + height/128)) render weight.

The weird thing there is that the x and y resolutions are added, not multiplied. I can't see how that can make any sense because it's the total pixel count that matters of course.


A credible formula should look something like this:


rw = o +(x*y*b)


(o is a constant for the "overhead" per texture, x and y are the resoultion, b a balancing factor to weight texture weight towards geometry weight.)


Quote:
Originally Posted by Sid II View Post
Who cares about the calculations in the first place?
...
To me, it is mostly drama lama stuff to be honest.

Render cost is a major factor for driving people away from SL. It may not matter to you but it's certainly very important to Second Life as a whole.


I'm not sure if you're serious or joking but you are actually right. As unreliable as the render weight is today, there's hardly any point for the average SL'er to bother about it.
ChinRey is offline   Reply With Quote
Old 09-12-2018, 09:53 AM   #34 (permalink)
Senior Member
 
Nika Talaj's Avatar
 
Join Date: Sep 2007
Posts: 3,581
Quote:
Originally Posted by Sid II View Post
To me, it is mostly drama lama stuff to be honest.
To each their own. But to anyone who does not have a monster rig and would like to go to events of any kind and actually SEE the other avatars there, it is very important.

I'm pretty appalled at the critiques above. Given SL's maturity, it is amazing that they don't stay on top of these numbers. No one could ever get them perfect, but this approach of just doing ARC once (or, actually, twice) and seemingly never touching the code again is ridiculous. If they had real expertise on board, that person could review the code regularly.

Also, if they were serious about performance, wouldn't there be both server-side numbers, relating to the properties of objects etc. rendered, and THEN a client side portion, that you could optionally turn on to see the cost given your machine's limitations?

For example, VRAM usage. The impact of this varies according to your graphics card, BUT also it seems to me that if your networking is not good, VRAM usage correlates to network traffic and can kill you. That's just a gut feel, tho. NUMBERS, WANT MORE NUMBERS!
Nika Talaj is offline   Reply With Quote
Old 09-12-2018, 10:24 AM   #35 (permalink)
Senior Member
 
Da5id Weatherwax's Avatar
I have no status
 
Join Date: Aug 2010
Location: Edinburgh
Posts: 1,467
Business: No Escape
Client: Official, Various RCs, Marines,Firestorm - for testing stuff. BlackDragon - For everything else.
Quote:
Originally Posted by ChinRey View Post
...So it will only have a multiplier at all if the code happens to recognise it as the same as the rigged mesh it knows about
(bolding mine)


If the code recognises rigged mesh it will recognise fitmesh too. fitmesh IS "rigged mesh" - it's just rigged to a different subset of bones in the same avatar skeleton.



mesh is weighted to the avatar skeleton -> it gets a multiplier.
Da5id Weatherwax is offline   Reply With Quote
Old 09-12-2018, 11:13 AM   #36 (permalink)
Just another muppet
 
Sid II's Avatar
Eh.. mi mi mi mi mi?
 
Join Date: Jun 2015
Location: In the south of The Netherlands.
Posts: 1,751
My Mood:
SL Join Date: Spring 2007
Client: standard
This rendering weight would make sense if LL would say: okay, at xxxx you're at the max, so it is impossible to ad more to your avatar.
Sid II is offline   Reply With Quote
1 User Disagreed:
1 User Likes This:
Old 09-12-2018, 11:21 AM   #37 (permalink)
Emergency Mustelid
 
Argent Stonecutter's Avatar
Please State the Nature of the Mustelid Emergency
 
Join Date: Sep 2009
Posts: 20,779
The whole point to jelly dolls is that it doesn't matter if you've got a heavy avatar, people who don't want to see heavy avatars won't see it. If you're hanging out with other people who like that kind of avatar, why should they not be able to create them and see each other... it won't impact anyone who doesn't want it.
__________________
Argent Stonecutter -- Skyhook Station -- Coonspiracy Store

"And now I'm going to show you something really cool."


The previous is a cybernetic datum published - in direct contravention of DoD Regulation #229RR3X3 - as being conducive to the physical, psychological and/or social well-being of the population.
Argent Stonecutter is offline   Reply With Quote
2 Users Agreed:
2 Users Like This:
Old 09-12-2018, 01:23 PM   #38 (permalink)
Senior Member
 
ChinRey's Avatar
Status symbol
 
Join Date: Jun 2014
Posts: 170
My Mood:
Business: OPQ
Quote:
Originally Posted by Da5id Weatherwax View Post
If the code recognises rigged mesh it will recognise fitmesh too.
Yes, and as I said in my first post, it is almost certainly just the web page that hasn't been updated.
But I still keep that "almost" reservation until somebody has checked and confirmed. This is Linden Lab code, remember. Much of it was written during the darkest years of LL development by a programmer who didn't know the difference between diameter and radius. Who knows what kind of lunacies are lurking inside it. For all we know, the code for calculating rigged mesh render weight might be so hardwired to one specific set of bones it fails to recognise meshes with even the slightest deviation.
And before you ask, there is no reason to believe the code was updated or even checked when fitted mesh was introduced.

Last edited by ChinRey; 09-12-2018 at 01:46 PM.
ChinRey is offline   Reply With Quote
Old 09-12-2018, 02:17 PM   #39 (permalink)
Senior Member
 
Dakota Tebaldi's Avatar
Lurkin' and stuff
 
Join Date: Feb 2008
Location: Down by the bay, where the watermelons grow...
Posts: 16,791

Awards: 1
SLU Creepy Avatar Competition 2014 Participant 
This thread has made me sensitive to the complexity problem and I started paying attention to mine yesterday, and particularly Adeon's advice that trying to stay below 70k if possible is ideal. I had a look at the complexity of what I consider my "everyday" normal av, which is mesh-everything, including a new fitted body and head (which I explained in another thread). I was dismayed to see my number just a little under 110k, and wondered if the mesh body and head just weren't going to be practical. But after experimenting a little, I found out that my sneakers alone were worth 20k each, and my hair was 40k all by itself! What the heck? I've already found hair that looks great and is only like 5k! And I'll do some sneaker shopping tonight.

From the beginning, makers of mesh stuff have always been neglectful of things like mesh complexity, not using texture atlases when they really should, and not even bothering with creating proper LODs. I just didn't really realize how bad it was until now. When bump and spec maps exist and are useable in SL there's no excuse for having mesh that darn dense; but too many creators don't seem to approach SL mesh content from a game creation POV; they're amazing at modeling but aren't as well studied in the workflow of refining assets to be more game-ready. And maybe there's little incentive to, since the way the complexity information system is designed tends to put all the onus on users to police what they wear and not on the actual creators to properly optimize the stuff they sell.
__________________
Quote:
Originally Posted by Io Zeno
Cody, you are unusual.

Last edited by Dakota Tebaldi; 09-12-2018 at 02:26 PM.
Dakota Tebaldi is online now   Reply With Quote
Old 09-12-2018, 02:35 PM   #40 (permalink)
Emergency Mustelid
 
Argent Stonecutter's Avatar
Please State the Nature of the Mustelid Emergency
 
Join Date: Sep 2009
Posts: 20,779
One of my TWI rigged mesh animal avatars (wolf? fox? I don't recall which I noticed it on) was almost 200,000, when TWI updated it to bento that took it down to something like 40,000. And all I had was the base avatar and a prim hat. Taking the hat off had no noticeable effect.
Argent Stonecutter is offline   Reply With Quote
1 User Likes This:
Old 09-12-2018, 03:15 PM   #41 (permalink)
Senior Member
 
Dakota Tebaldi's Avatar
Lurkin' and stuff
 
Join Date: Feb 2008
Location: Down by the bay, where the watermelons grow...
Posts: 16,791

Awards: 1
SLU Creepy Avatar Competition 2014 Participant 
Quote:
Originally Posted by Argent Stonecutter View Post
One of my TWI rigged mesh animal avatars (wolf? fox? I don't recall which I noticed it on) was almost 200,000, when TWI updated it to bento that took it down to something like 40,000. And all I had was the base avatar and a prim hat. Taking the hat off had no noticeable effect.
That's awesome!

I do feel a little bad for furries in particular, because I can imagine the nature of their avatars makes it tough to keep that complexity number down too much. And there's the whole fursona idea where a lot of them are very closely attached to the avatars they've had since forever, and would find it very difficult to change - considering that mesh stuff by its nature can be a lot less customize-able and personalize-able that older prim stuff unless you are lucky enough to get a creator who's willing to go out of their way to accommodate that. And all of which I can understand, I mean look at me not wanting to give up a system skin I've had since 2011. For me, this just motivates me to do what I can to keep my own numbers down - take up a little of the slack, so to speak.
Dakota Tebaldi is online now   Reply With Quote
Old 09-12-2018, 07:21 PM   #42 (permalink)
Myf
McMahon
 
Myf's Avatar
Pop!
 
Join Date: Feb 2010
Location: Syd, Aust.
Posts: 4,075
My Mood:
SL Join Date: March '07
Business: The Kino Haus
Client: the one you hate

Awards: 2
SLU Creepy Avatar Competition 2014 Participant 
Quote:
Originally Posted by Sid II View Post
Who cares about the calculations in the first place?
If people find that my avatar is to rich for their findings, they miss seeing me as I am.
I could not care less.
To me, it is mostly drama lama stuff to be honest.
Because the situation quickly degenerates into a kind of Tragedy of the Commons, as more and more complex avatars bring everyone's viewers to their knees.
Myf is offline   Reply With Quote
Old 09-12-2018, 07:24 PM   #43 (permalink)
Emergency Mustelid
 
Argent Stonecutter's Avatar
Please State the Nature of the Mustelid Emergency
 
Join Date: Sep 2009
Posts: 20,779
Quote:
Originally Posted by Myf View Post
Because the situation quickly degenerates into a kind of Tragedy of the Commons, as more and more complex avatars bring everyone's viewers to their knees.
That used to be a problem. Now we have jellydolls.

Quote:
Originally Posted by me
The whole point to jelly dolls is that it doesn't matter if you've got a heavy avatar, people who don't want to see heavy avatars won't see it. If you're hanging out with other people who like that kind of avatar, why should they not be able to create them and see each other... it won't impact anyone who doesn't want it.
Argent Stonecutter is offline   Reply With Quote
Old 09-12-2018, 08:43 PM   #44 (permalink)
Under Construction
 
Jupiter Firelyte's Avatar
Figuring things out.
 
Join Date: Jul 2010
Location: Melbourne, Australia
Posts: 1,617
My Mood:
SL Join Date: 02/09/2010
Quote:
Originally Posted by Dakota Tebaldi View Post
This thread has made me sensitive to the complexity problem and I started paying attention to mine yesterday, and particularly Adeon's advice that trying to stay below 70k if possible is ideal. I had a look at the complexity of what I consider my "everyday" normal av, which is mesh-everything, including a new fitted body and head (which I explained in another thread). I was dismayed to see my number just a little under 110k, and wondered if the mesh body and head just weren't going to be practical. But after experimenting a little, I found out that my sneakers alone were worth 20k each, and my hair was 40k all by itself! What the heck? I've already found hair that looks great and is only like 5k! And I'll do some sneaker shopping tonight.

From the beginning, makers of mesh stuff have always been neglectful of things like mesh complexity, not using texture atlases when they really should, and not even bothering with creating proper LODs. I just didn't really realize how bad it was until now. When bump and spec maps exist and are useable in SL there's no excuse for having mesh that darn dense; but too many creators don't seem to approach SL mesh content from a game creation POV; they're amazing at modeling but aren't as well studied in the workflow of refining assets to be more game-ready. And maybe there's little incentive to, since the way the complexity information system is designed tends to put all the onus on users to police what they wear and not on the actual creators to properly optimize the stuff they sell.
When I returned to SL a few weeks ago, I bought mesh, and started clearing out my inventory of old things for the system body. All of my stuff was six years old because that's how long it had been since I was in SL. A whole bunch of it displayed much higher ARC than my new mesh stuff. Shoes and hair, in particular, were shockers. I assume that clothing made today is better than that made six years ago, as creators have learned more and gotten better at making stuff. But my mesh stuff today looks better and results in lower ARC than my system stuff from six years ago.

Also, people wear a lot of stuff! I wear a body, head, hair, clothes, shoes, and an AO. That's it. Occasionally I wear a hat and, if I'm in RP, some props. Blinking bling may be gone, but bling is still out there in the form of a dozen piercings, watches, jewellery, etc. There's no way you're going to have ARC under 100k with all that on. I also wonder how many people walk around with their mesh huds attached.

People advise that residents remove all the extras when going to shopping and social events, but that's not going to happen. People want to see themselves and be seen wearing their best. Unfortunately, there's not a lot of consumer education about how these things affect the user experience.
Jupiter Firelyte is offline   Reply With Quote
1 User Said Thanks:
2 Users Agreed:
Old 09-12-2018, 08:47 PM   #45 (permalink)
Under Construction
 
Jupiter Firelyte's Avatar
Figuring things out.
 
Join Date: Jul 2010
Location: Melbourne, Australia
Posts: 1,617
My Mood:
SL Join Date: 02/09/2010
Quote:
Originally Posted by Argent Stonecutter View Post
That used to be a problem. Now we have jellydolls.
If I'm a new user with a midgrade computer and I see more jellydolls than rendered avatars, I don't think I'd stay in SL. I'd be like, wtf is this shit? and go back to my video games, which look awesome. I think people underestimate how important it is for SL to look good.
Jupiter Firelyte is offline   Reply With Quote
1 User Agreed:
1 User Likes This:
Old 09-12-2018, 10:20 PM   #46 (permalink)
Lost Bird
 
WolfEyes's Avatar
 
Join Date: Oct 2009
Posts: 14,042
SL Join Date: April 2004
Quote:
Originally Posted by Argent Stonecutter View Post
One of my TWI rigged mesh animal avatars (wolf? fox? I don't recall which I noticed it on) was almost 200,000, when TWI updated it to bento that took it down to something like 40,000. And all I had was the base avatar and a prim hat. Taking the hat off had no noticeable effect.
I heard my name...

__________________
This is the truth.
Mitakuye oyasin.
To all my relations, keep dancing.
Believe or you will not see.
WolfEyes is offline   Reply With Quote
3 Users Like This:
Old 09-13-2018, 12:50 AM   #47 (permalink)
Quietly watching
 
Vaelissa Cortes's Avatar
 
Join Date: Jan 2008
Location: Relaxing at home.
Posts: 3,538
My Mood:
SL Join Date: 4/24/2007
Client: PoundLife
Quote:
Originally Posted by Jupiter Firelyte View Post
Also, people wear a lot of stuff! I wear a body, head, hair, clothes, shoes, and an AO. That's it. Occasionally I wear a hat and, if I'm in RP, some props. Blinking bling may be gone, but bling is still out there in the form of a dozen piercings, watches, jewellery, etc. There's no way you're going to have ARC under 100k with all that on. I also wonder how many people walk around with their mesh huds attached.
I've been avoiding this thread but have to chime in here. It is very, very doable to have all of that stuff and more attached and stay under 100k complexity. I never reach my personal max limit of 80k, and am usually in the 40-50k-ish range, but complexity is a secondary metric that I rarely pay attention to, instead, I look at an item's VRAM usage and triangle count, which are far better indicators of how much of a performance impact an item has. Just paying attention to those two things tends to naturally result in an avatar with a lower complexity number since the items will be better made in the first place. When paired with some of the tips Chalice mentions above (like unlinking extra hidden geometry, and setting alpha to masked), along with self optimizing textures, even reaching 100k in the first place is an accomplishment.

It is especially important to pay attention to VRAM and triangles since, as mentioned earlier in this thread, the complexity system does not properly penalized those two things, even though they are the most important things that should be weighed. For example, you can have an avatar complexity under 100k and still be eating up half of SL's allowed video memory in textures, which is very bad.

The thing here is that most creators out there are simply awful at creating efficient content, so you need to both shop smart, and if it's in your ability, optimize the content yourself. Unfortunately this is not easy for the average consumer, especially with content getting exponentially worse efficiency-wise.

Second Life has its code problems for sure, but a huge bulk of the performance issues come down to creators not making efficient content due to having truly terrible creation habits. The aggrevating thing here is that most items and avatars out there could easily look the same as they do now, but use significantly less resources if only the creators of the items practiced some very basic things that are not only common, but absolutely required in literally the rest of the industry. This is one of the main reasons Second Life famously needs a more powerful computer just to get acceptable FPS, when you can use the same PC to render high end graphics at 60FPS all day in any other game.
Vaelissa Cortes is offline   Reply With Quote
1 User Said Thanks:
Old 09-13-2018, 02:18 AM   #48 (permalink)
Under Construction
 
Jupiter Firelyte's Avatar
Figuring things out.
 
Join Date: Jul 2010
Location: Melbourne, Australia
Posts: 1,617
My Mood:
SL Join Date: 02/09/2010
Quote:
Originally Posted by Vaelissa Cortes View Post
... but complexity is a secondary metric that I rarely pay attention to, instead, I look at an item's VRAM usage and triangle count, which are far better indicators of how much of a performance impact an item has... When paired with some of the tips Chalice mentions above (like unlinking extra hidden geometry, and setting alpha to masked), along with self optimizing textures, even reaching 100k in the first place is an accomplishment.

It is especially important to pay attention to VRAM and triangles since, as mentioned earlier in this thread, the complexity system does not properly penalized those two things...
I hate to sound like a dolt, but I'm an average consumer. I don't build anything. I don't script. I've never used 3D computer animation and modelling software. I just log in, go shopping, take photos, hang out. And although I kind of understand these issues, I have no idea how to do the things that are being described in this post and others like it and I wouldn't know what to look for when I'm trying on demos of clothing and accessories.
Jupiter Firelyte is offline   Reply With Quote
1 User Agreed:
Old 09-13-2018, 02:52 AM   #49 (permalink)
Senior Member
 
Dakota Tebaldi's Avatar
Lurkin' and stuff
 
Join Date: Feb 2008
Location: Down by the bay, where the watermelons grow...
Posts: 16,791

Awards: 1
SLU Creepy Avatar Competition 2014 Participant 
Quote:
Originally Posted by Jupiter Firelyte View Post
I hate to sound like a dolt, but I'm an average consumer. I don't build anything. I don't script. I've never used 3D computer animation and modelling software. I just log in, go shopping, take photos, hang out. And although I kind of understand these issues, I have no idea how to do the things that are being described in this post and others like it and I wouldn't know what to look for when I'm trying on demos of clothing and accessories.
Well that part's easy. Yeah, it's impossible by just looking at a picture of an item to tell how optimal it is, ARC-wise. But, that pop-up complexity message that you mentioned in your first post - that message pops up, with an updated total, every single time you attach or detach something to or from your avatar. So you can basically experiment by detaching things one at a time, and figure out which parts aren't so impactful and which are insanely impactful - like my 40,000-ARC hair that has now been replaced. And that's also true for any demo items you attach - so don't just look to make sure it fits and looks nice, check the ARC pop-up message when you put it on while you're at it. And you can use that to help you decide what to buy - or maybe even what to consider replacing.

I mean high ARC isn't like some kind of sin; there's gonna be times when you want to wear a special thing to a certain event and it's a little primmy and that just happens sometimes. But ARC is one of those things where the more people who are mindful of it generally, the better off everyone's experience will be generally.
Dakota Tebaldi is online now   Reply With Quote
1 User Said Thanks:
1 User Likes This:
Old 09-13-2018, 04:27 AM   #50 (permalink)
Quietly watching
 
Vaelissa Cortes's Avatar
 
Join Date: Jan 2008
Location: Relaxing at home.
Posts: 3,538
My Mood:
SL Join Date: 4/24/2007
Client: PoundLife
Quote:
Originally Posted by Jupiter Firelyte View Post
I hate to sound like a dolt, but I'm an average consumer. I don't build anything. I don't script. I've never used 3D computer animation and modelling software. I just log in, go shopping, take photos, hang out. And although I kind of understand these issues, I have no idea how to do the things that are being described in this post and others like it and I wouldn't know what to look for when I'm trying on demos of clothing and accessories.
I should emphasize that it's a problem creators need to deal with since the fault lies entirely with them. It is only due to creators doing a poor job at making content that any consumers even need to think about optimization in the first place, or do something about it.

The average consumer should not have to worry about resource usage at all, or be expected to learn any of the stuff you mentioned just so they can enjoy SL. Being able to just buy, wear and be done with something is a more than reasonable expectation. Sadly SL is far from a polished product and has a steep learning curve all around due to its nature.

As for things to look for, you can right click and inspect items to view VRAM and triangle counts. I put up some general guidelines in this post in another thread which should be helpful. I have some comprehensive guides in the works, but nothing complete enough to post yet.
Vaelissa Cortes is offline   Reply With Quote
1 User Agreed:
1 User Likes This:
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On




SEO by vBSEO