PE = tier = real dollars

[Drongle McMahon]

We are talking a lot here about prim costs, henceforth to be called PEwts, in a rather abstract fashion, as perhaps befits a virtual world. However, it may be a reaonable to look at this with a bit of hard-nosed realism.

What do prims really cost? To most residents, their tier payment, whether that be direct or through rent pid to a landlord, is for buying prims. Granted there is a little more to land than that, but that does seem to be the overriding consideration. People even buy tiny parcels to add to their prims and nothing else. Now the marginal monthly tier per prim varies between 1.3 cents, for the owner of a full mainland region, to 4.2, for a humble pemium to expand his free 512. For estates, which are the vast majority of land, it is 2 cents for the landlord and therefore more for his tennants. So it is conservative to use the 2 cent figure*. Most will be paying more, and we will let the difference cover the other land related benefits.

That means that my gallery, at 250 PEwts, is going to cost me  (or you!) 5 US$ per month to keep on the main grid. That's 60 US$ a year, before I put anything in it. I find that a rather sobering calculation.

*Of course, as far as resource consumption is concerned this really includes a rather large subsidy for everyone else's attachments, but that is a different question.


[Gaia Clary]

That is the front side of the coin. On the backside of that coin i still have to undertand why i would want to use meshes when they are so expensive by all means ?

The front side of the coin is understandable to me. The owners of Second Life may want to earn more money. Nobody can blaim them for doing that. We the users only can decide to not use their offer...

But it is the backside of the coin which makes me a lot of trouble. Many have meanwhile asked the question what basic ideas are behind the decisions to make the less resource intensive meshes become more expensive. 

So i can understand that i have to work much harder to get more efficient meshes just because i have to define the LOD-meshes, the physics mesh, the texture faces. I have to reduce poly count as much as possible to optimize my meshes.

But after i have done all this and contributed to optimizing everything, then i have to pay more for uploading (well, we do still not know what surprise we will get there) and more for rezzing. So as a content creator i have not only to sell for higher price to cover the builders costs, but also my customers have to pay more in terms of rezz units (prim costs).

I think it is time to get an explanation where all that makes sense and where we all are thinking wrong.

Or is this all just the first step to raise the price for everything including regular prims and sculpted prims ? I mean that would make sense after all... ;-(

[DanielRavenNest Noe]

My feeling is the programmer Linden staff are looking at it from a technical point of view.  They do not want complex 3D models to slow down the server and people who have slow computers.  But this ignores the customer side of things.  Mesh is a "new product" along with prims and sculpts.  If it is much more expensive to make an object with mesh, nobody will use it.

I have an example terrain model I made which is 2200 triangles.  It could be done with 2-3 sculpts, which would cost 2-3 prims.  The mesh version costs 56 prims.  So why would anyone use mesh when the cost is 20x higher?  I have a small palm tree which is 900 triangles.  That could be 1 sculpt, but it's 11 prims as a mesh.  So again, the cost is much higher.

It is the wrong approach to penalize everyone for the people with slow computers.  For those people, simply do not send the highest LOD, or restrict the LOD distance for each level.  They will see fewer triangles, and have better frames per second, and the server will send less data.  But as the costs stand now, mesh will be useless for land-based building, it will only work for avatar wearables.

[Drongle McMahon]

"If it is much more expensive to make an object with mesh, nobody will use it." 

And the consequence will be that the people with slow computers willhave to look at sculpties instead, which will make their experience worse when it could have been improved by sensibly priced meshes. I really don't see any more how this can work at all unless there is a plan to penalise sculpties the same way.  They'll also have to look at super high poly mesh attachments because for them there is no restraint whatsoever. 

[Gaia Clary]

 

---

DanielRavenNest Noe wrote:

My feeling is the programmer Linden staff are looking at it from a technical point of view.  They do not want complex 3D models to slow down the server and people who have slow computers. 

---

But how does this match up with the fact that i can make meshes which use much less resources in terms of triangles, LOD, etc...  compared to equivalent builds with regular prims/sculpted prims.

But when i have to decide what i will actually use in my builds i will always end up with the version which yield the smallest costs and the least effort for me and the customer.

If creating low resource builds would be rewarded by lower prim costs, then we would do all our best and find optimal solutions for our builds. But as long as whatever we do we will end up with higher costs, then what reason will make us still try to go along that way except experimental curiousity and enthuisiasm ? 

Well, one positive remark: I must admit that i still have not digged as deep as possible into mesh optimizations. There may be still a lot of potential to get resource costs down... Thinking of my kettle i can see at least that (according to resource costs if those are still what we will have to count) i can get into the same cost range as with sculpties, but i can get better LOD behaviour and better texture behaviour. Maybe we just have not digged deep enough ...

Or maybe a kettle is just one of those shapes which just can not be done well  (cost efficient) with mesh (as drongle mentioned) ?

[guY Ralior]

---

Drongle McMahon wrote:

We are talking a lot here about prim costs, henceforth to be called PEwts, in a rather abstract fashion, as perhaps befits a virtual world. However, it may be a reaonable to look at this with a bit of hard-nosed realism.

 

What do prims really cost? To most residents, their tier payment, whether that be direct or through rent pid to a landlord, is for buying prims. Granted there is a little more to land than that, but that does seem to be the overriding consideration. People even buy tiny parcels to add to their prims and nothing else. Now the marginal monthly tier per prim varies between 1.3 cents, for the owner of a full mainland region, to 4.2, for a humble pemium to expand his free 512. For estates, which are the vast majority of land, it is 2 cents for the landlord and therefore more for his tennants. So it is conservative to use the 2 cent figure*. Most will be paying more, and we will let the difference cover the other land related benefits.

 

That means that my gallery, at 250 PEwts, is going to cost me  (or you!) 5 US$ per month to keep on the main grid. That's 60 US$ a year, before I put anything in it. I find that a rather sobering calculation.

 

*Of course, as far as resource consumption is concerned this really includes a rather large subsidy for everyone else's attachments, but that is a different question.

---

Does this mean that with mesh items you pay a fee for them after the initial upload charge? If that is the case of course the answer is obvious on what to use no matter how much it strains the server to use sculpties.

 

 

 

 

[Drongle McMahon]

No extra fee. This is the tier you pay for owning land. You (or your landlord) have to pay tier to have the land to put your prims on, unless you are a squatter or live in sandboxes. If you just have a Linden Home, it's part of your premium subscription. It's the same for all prims, mesh or otherwise. Mesh heve a Prim Equivalent weight, which uses up some of the land owner's prim allowance. It is the calculation of that equivalence that we are concerned with.

[guY Ralior]

Ahh understood on how it weighs against the tier or land fee a person would pay I was just for a moment thinking you were meaning there was going to be a charge for them like each month somehow, that was going to stop me in my tracks from thinking any type of mesh was going to be viable

I work with a smaller prim allowance so for me it is very important how it all adds up as far as static inWorld objects

[Dain Shan]

Well Gaia

We tried that all i think.

The best example is still the one with the most primitive object we can do thats still 3D

A Cube with 12 Triangles.

 

The cost is always 2 prims at least. So we cant compete with an object that can have a mulitude of faces and still is 1 prim only.

Theres simply no way. Alone because of the fact that we cant go below 2 prims.

The so called ressouce talk and that mesh shouldnt burden low end PCs too much is .. well i dont know how to put it friendly, so i better spare that, .... because the low end PCs look at sculpties at any corner and any square meter they move.

So thats so not a good argument.

So what else there is?  I still cant come up with a resonable explanation.

 

I agree that the future of meshes seems to lie in Attachments and some cases of rigged clothes only. For tha broad audience at least.

 

 

 

[Gaia Clary]

---

Dain Shan wrote:

A Cube with 12 Triangles.

The cost is always 2 prims at least. So we cant compete with an object that can have a mulitude of faces and still is 1 prim only.

Theres simply no way. Alone because of the fact that we cant go below 2 prims.

 

---

Has the PE been acknowledged as the final value ? The resource cost for a <64,64,64> mesh cube is 0.78. The PE for the same cube is 2. As long as LL does not tell what we have to believe i continue looking at the  "Resource costs" . That is what Runitai has told us to do 10 days ago. And they told us on monday that we will "probably see consense between what the viewer thinks (PE?) and what the server thinks (RC?) by end of this week"

Since there is no consense at the moment, how can we tell what is right and what is wrong ?

 

[Zslash Cyberstar]

i will probably not use meshes for my creations because 1. costs too much, and 2, sculpties and prims are cheaper and it doesnt cost to make or "upload". except sculpts tho its 10l ....

[bodzette Coignet]

re: IFMesh is much more expensive then nobody will use it

That's not necessarily true, neither in RL nor in SL.

There are various factors including but not limited to: how much will people value mesh once it has been uploaded?

The two reasons why people would still upload mesh even though it's more expensive that immediately spring to mind are that creators can sell way more copies and that even if they are more expensive and demand is low, people will still buy them for builds because they effectively get more vertices for their buck.

 

[Maniac Choche]

---

Gaia Clary wrote:

But how does this match up with the fact that i can make meshes which use much less resources in terms of triangles, LOD, etc...  compared to equivalent builds with regular prims/sculpted prims.

---

Maybe the problem is related in the amount of streamed information? For example:

- regular prims have fixed normals , fixed UVs , all those are not streamed

- all information related to a sculpty object is hold in a very compressed image file

I've made a very rude test with a sculpty:

In it's native JPEG2000 format the size is ~4k. Same sculpty in uncompressed .dae is 1.05Mb. Since i have no clue what type of compression LL will use internally I've used RAR just to see how much i can compress it. I came up with something like 57kb and best possible compression settings. The differences are quite big , but the amount of information is more too. And let's not forget custom created LODs and Physics all those have UVs , normals , vertex positions etc. It's not just one DAE.

 

Best regards

 p.s .dae file have tons of tags that are probably ignored and not streamed

[Dain Shan]

---

Gaia Clary wrote:

---

Dain Shan wrote:

A Cube with 12 Triangles.

The cost is always 2 prims at least. So we cant compete with an object that can have a mulitude of faces and still is 1 prim only.

Theres simply no way. Alone because of the fact that we cant go below 2 prims.

 

---

Has the PE been acknowledged as the final value ? The resource cost for a <64,64,64> mesh cube is 0.78. The PE for the same cube is 2. As long as LL does not tell what we have to believe i continue looking at the  "Resource costs" . That is what Runitai has told us to do 10 days ago. And they told us on monday that we will "probably see consense between what the viewer thinks (PE?) and what the server thinks (RC?) by end of this week"

Since there is no consense at the moment, how can we tell what is right and what is wrong ?

 

---

As i understood it in the last meeting were final with the values.

Same goes for adding scripts to a Mesh prim.

 

I took that from the lines

[12:13]	Nyx Linden	scripted and physical objects send out more messages, so they cost more for Server cost. If your object has a higher streaming cost, it won't have an effect

For additional Costs when you add an script. As far as we testedt it la low cost mesh was pumped up to 4 Prims when a script was added

 

And since we asked the combined question

[12:08]

Charlar Linden

#3 Update on prim cost. We now see streaming cost is prim cost. How will making the object scripted/physical affect the overall cost? Min stand alone prim cost? Any rough formulas would be helpful. (Gearsawe)



The answer was

[12:09]

Nyx Linden

basically "prim equivalence" is the highest for download, server, and streaming costs. Whichever factor is highest for your linkset, that's what we charge

So it seems to me .. ( i may be mistaken ) that this values are considered final.

 

[Dain Shan]

---

Maniac Choche wrote:

Maybe the problem is related in the amount of streamed information? For example:

- regular prims have fixed normals , fixed UVs , all those are not streamed

- all information related to a sculpty object is hold in a very compressed image file

I've made a very rude test with a sculpty:

In it's native JPEG2000 format the size is ~4k. Same sculpty in uncompressed .dae is 1.05Mb. Since i have no clue what type of compression LL will use internally I've used RAR just to see how much i can compress it. I came up with something like 57kb and best possible compression settings. The differences are quite big , but the amount of information is more too. And let's not forget custom created LODs and Physics all those have UVs , normals , vertex positions etc. It's not just one DAE.

 

Best regards

 p.s .dae file have tons of tags that are probably ignored and not streamed

---

The final datapakage is far below 57 kb.

All files will be merged into one file for the object, the uploader just need all of them to generate this one file.

 

Also the data file of the sculpt you have looked at is most likely not even 50% if optimized into mesh.

 

 

[Maniac Choche]

Well no matter what is merged, the streamed information for a 1024 vertex mesh is several times bigger than a 1024 vertex sculpty. With a sculpty you get only vertex positions for the highest LOD all rest is calculated by the viewer. That's why sculpties was invented to ease the streaming cost not the FPS. It's logical that LL cares more about their internet costs than our FPS. 

Best regards

 

 

[Lazy Spectre]

Whether a single sculptmap VS a mesh with the same amount of verts is more or less doesn't matter when the majority of sculptie builds are made up of multiple sculptmaps + baked textures,i'm pretty sure that in the end those kind of builds end up as much more data transfer plus more of an FPS hit @ HI LOD than the mesh equivalent.

[Feynt Mistral]

Ok, a few "this is probably whys" are in order here:

Firstly, costs.  It's better to start out high and then realize that "hey, we really could handle more of this, let's turn down the cost."  No one complains about a price decrease.  Everyone complains about price increases.  Based on a technical, and also theoretical basis, the costs and prim equivalencies will make sense once the client/server communication catches up with each other, and even if they don't the public beta testing on the main grid will show them under real circumstances the impact that streaming meshes will have.  If they find that overall the experience is improved and the streaming is managable, they can decrease the costs to match and encourage further mesh usage (so that those low end machines can enjoy a properly LoD'ed landscape).

 

Second, meshes vs. sculpties.  There's no comparison, sculpties win hands down on streaming.  ALL of their information is stored in a compact little 32x32 picture that is made for streaming.  The normals are calculated based on the surface generated from the colour channels.  There's no UV data transmitted either, it's generated as well.  That single 32x32 picture pulls triple duty.  With meshes you're passing along three values, certainly (x,y,z vertex info), but there's also individual normal and UV data sets for each vertex, so you've got two to three times the data.  And that's excluding possible weighting.



[leliel Mirihi]

---

DanielRavenNest Noe wrote:

 

It is the wrong approach to penalize everyone for the people with slow computers.  For those people, simply do not send the highest LOD, or restrict the LOD distance for each level.  They will see fewer triangles, and have better frames per second, and the server will send less data.  But as the costs stand now, mesh will be useless for land-based building, it will only work for avatar wearables.

---

The viewer already does this. You can control it with Preferences->Graphics->Object Mesh Detail which controls the debug setting RenderVolumeLODFactor.

[Scope]

Sounds like a vehicle may be a taller order than first imagined. Just 31 to play with but boy do they look good.

[Gaia Clary]

Lets put the costs aside for a moment and look what we will realy get with mesh:

 

feature	Sculpted Prim	Mesh
mesh topology	fixed topology 3 basic types plane,cylinder,torus	arbitrary meshes
facecount	16-1024	1-65536
shading	Smooth	Flat & Smooth (mixed)
texture faces	1	1-8
UV-map	fixed	user defined
LOD	4 automatic LOD	4 user definable LOD
what else ?	...	...

 

So if used with care i can imagine that we can create meshes which can compete with sculpties. For example despite the fact that i can not get my kettle below 2 PE, it still is superior to the Sculpted prim, because:

 

1. I can texturise the kettle interior individualy (With sculpties i'd need  one extra object to do that)
2. I can get sharp edges easily (no need to collapse edge loops (argh..) or bake sharp edges into the textures
3. I have full control over the texture resolution on each part of the object.
4. I can define optimal object specific LOD behaviour.

Right now my mesh kettle costs 4 PE ( i am still investigating why),  while my sculptie kettle costs 3 (one for the pot, one for the handle, one regular prim for the interior if i wanted that separately texturizable). Yes i know i could make the kettle out of one sculptie only, but i would loose all texturing gimicks then and needed to retexturize the entire object for any change inside.

So after thinking about it maybe it is wrong to try comparing single sculpted prims with meshes. Maybe we should look at multi sculpty builds and see how they compare to mesh. What immediately comes into my mind is :

 

1. shoes made of 60 sculpted prims and more
2. hair made of 50 sculpted prims
3. Jewelery made of 120-200 sculpted prims

I bet these builds can be made cheaper with Meshes, no ? hmmm, all that are attachments ... well, so then lets see what we can make out of what we got now with meshes... And maybe Feynt is right and we can see a decrease of PE per mesh in the future, who knows.

 

[Medhue Simoni]

lol, if sculpts are so efficient, then why do they take 30 or more seconds to load?

Does any1 remember Phillip saying, something to the affect, that if he could implement mesh for only attachments, he would, or he would do it now? Seems the LL staff might just all have the same mindset. Too bad I can't even upload a rigged pistol holster cause there is a bug or something related to 3ds max and partial rigs.

The problem, I think, is that we are all looking at it from the perspective of creators who all want to make things more efficient. While, on the other hand, LL might be looking at it from the point of stopping some1 from even being able to use a poorly made mesh, or a mesh not optimized for SL. IMHO tho, irreguardless of LL's reasoning, they will hurt all their good work on this project and seriously handcuff creators.

I would also warn LL that this project is a bit more important than most people think. Mesh could actually get people to use V2. If it is not implemented to it's maximum, and things stay as they are, this leaves the door open for other platforms to implement more reasonable restrictions. If that did happen, how many creators would then consider SL their main base for creative output?

Well, whatever tho, I just hope that if rigged is the only viable use, I can at least get it to upload.

[Ashasekayi Ra]

That isn't just happening with Max. You can't use partial rigs with Blender any longer either. Now, you have to include the whole list of bones.

[Medhue Simoni]

I was just uploading the whole scene as a workaround, which won't be a viable option when there are actually costs to uploading. Now, tho, I can't get the upload option on a rigged mesh, no matter what I do. I did just make a whole new mesh body and rerigged it tho, but it all looks right. Maybe, I will try to go back to a viewer that did allow me to upload them, if I can remember which 1 that was. Thanks for the info. I also tried adding all the bones to the holster, but that didn't work either. I think the max collada exporter will only record the parts that have weight on them. Might be the same for blender.

[Tapple Gao]

I needed to add all 21 or so vertex groups to a mesh in blender in order for it to export and upload correctly. Most of them were empty