Understand full sim mesh builds/city architecture

[Voodoo Schnyder]

So I open my autodesk maya and start doing something I've want to for a while, a city for the size of a full SL region, but before I go on and develop most of it, I want to know if anyone can give me some pointers on how do poly faces affect prim count, by my understanding, is that mesh builds don't take as much prims/land objects as building everything in standard prim cubes? Is this right, or will I be disappointed when I finish a 65k sqm build and can't have it in SL because of the prim limits?

I ask mostly, because one time I bought a full perm mesh foot, it had around, I dont know maybe 100 to 300 faces and it was only taking 1 prim in the sim! Most cubes and planes used for sim building have even less faces than that, so this will make a good result? Full mesh sim, good idea?

 

Thanks in advance,

Voodoo Schnyder

[Rahkis Andel]

It depends.

 

My understanding is that overall there is more potential for mesh to be optimized, but it isn't intrinsically going to mean lower land impact. Particularly with mesh not being as stable and mature as other methods. I have to wonder if the scale of what you want will be a factor.

 

Sorry, the wasn't a helpful answer. I'm curious too.

[Jenni Darkwatch]

The "it depends" answer is pretty much spot on. Mesh can have more detail than normal builds but it gets adversely affected by size in many cases. If you plan on building large structures like houses you may sometimes get ridiculously high LI if you build it in one piece, but significantly lower LI if you build high-vertex details in smaller pieces and only low-poly parts in large single sets.

This is because of LOD switching. Mesh penalizes high detail builds if they're visible far away. The bigger something is, the further it can be seen, and the further away LOD switching happens.

Taking the house example, you could go ahead and build the basic box with very few details out of one mesh. Then add windowsills, frames and such as individual mesh parts. Because they switch LODs faster and are smaller, they'll be lower LI than if you integrate them into the house mesh.

Another "trick" that helps is to build the insides of buildings as a separate mesh with no lower LOD. Effectively it means that the inside disappears if people are too far away. Depending on windows and whatnot this may or may not be feasible, but it does certainly help with LI and also with viewer framerate.

The beauty of mesh is that it allows more freedom. The bane of mesh is that it requires a lot more skill and knowledge.

[Drongle McMahon]

Jenni's advice about choosing the right sizes for different parts, and insides vs outsides is exactly right. You need to understand these issues, and you will probably need to experiment. It would be a big mistake to build the whole thing before uploading any of it. You will need to keep track of LI consumption as you go, This will be especially true when it comes to physics weights, if you intend the insides of buildings to be accessible. Also remember that curves are much more expensive than sharp edges, for both download and physics components of the land impact.

[Medhue Simoni]

First let me say, I've never built for a whole sim. There are many tricks to building large, as stated above. What you will need to do is totally dependent on what the situation is with the sim, and what you want to accomplish with your builds there. If you are talking about having free reign on a whole sim, you might go about it all a bit differently. Prim limits are only an issue when other people are dealing with your builds. As far as I understand it, a sim has more than enough prims, for a sensible creator to build a whole mesh sim without having too many issues, depending on what you are trying to create. Large builds are going to be the most challenging part. Smaller builds will be so efficient, if done right, you will save a massive amount of prims that you can use on larger builds. Physics is the most important part of saving prims, and the proper LODs. You will want to check out some of the other threads on advice about large mesh buildings.

[Kwakkelde Kwak]

I've never done a full sim mesh build, but I did upload quite a bunch of buildings.

Getting the LI down with less geometry than normal prims shouldn't be too much of an issue, besides (like others mentioned already) keeping an eye on the LI as you go.

My experience is a LI of 5-10 times lower than prims is nothing to be surprised about.

What I am wondering about is the network load and viewers capabilities. A prim and even a sculpt is nowhere near as big as a mesh file I think. Mesh files have to cover all individual vertices, faces, normals, UV data etc and that for every LoD.

Only one way to find out how a full mesh sim performs, good luck and please keep us informed...

[Voodoo Schnyder]

I've tried uploading a single building with almost 5k faces and 3,8k vertices and the result is 87 land impact (with no physics enabled), I tried enabling physics, just for a test and it delivered multiple errors, the building is complex so Im overall pleased with the result so far, but I will keep this posted on whatever was sucessful or not, in this full mesh sim adventure

 

Edit: there are tricks for large mesh buildings? I did a search on this topic on these forums, but no luck finding any solutions

[Nalates Urriah]

Did you find the wiki pages on best practices?  #SL Best Practices

If you are going to work with building this is an important article to check: Reducing Prim Count

[Medhue Simoni]

---

Kwakkelde Kwak wrote:

What I am wondering about is the network load and viewers capabilities. A prim and even a sculpt is nowhere near as big as a mesh file I think. Mesh files have to cover all individual vertices, faces, normals, UV data etc and that for every LoD.

---

I think you are not taking alot of data into consideration. I just checked a snowman mesh that I made the other day. It's a total of 600 kbs. Just the textures alone, on a normal prim build would likely be more than that. UV mapping saves a ton of data when compared to multiple textures on 1 build. Plus, when you consider that any detailed sculpty build uses many sculpts to create, sometimes dozens, especially avatars, you are talking a ton of data, and you still need textures for all those sculpts. There will always be cases where a prim or sculpty build will be a little less data, but overall, the mesh will almost always win, when made right. A smart creator could save massive amounts data when making a mesh build properly. Plus, LODs save a ton of data, and don't add much to the overall total. The results of LODs for the veiwer is even greater, because the pc is not having to download all the full models in the distance.

[Jenni Darkwatch]

Like everything, whether mesh is good or bad for performance depends entirely on the skill of the creator. Much of the bad rep of mesh is, IMO, due to the often horrendously done worn mesh - rigged or otherwise.

It's IMO always been the bane and lure of SL That everyone can create things means that there's a lot of very, very bad stuff in SL. Jewelry made out of hundreds of prims, with 1024x1024 textures on miniscule surfaces and so on. That's just one example.

LI is the (crude) indicator of efficiency. It doesn't take into account sim-wide efficiency management. If every mesh on the sim has a different texture, the sim will be hell for visitors because the textures alone will load forever. The same goes for re-using mesh (or sculpts). Re-use as much as you can and you can get away with very detailed builds while still giving visitors a pleasant experience.

Here's an easy set of example to try:

Rez a sim full of large sculpted trees - max out the LI of the sim, but only use ONE tree sculpt. It's going load extremely fast, but even modern computers will struggle rendering that with any decent FPS. Load time = low, render efficiency = lousy.

Now go ahead and create a mesh tree, but replace all LODs below the highest one with distance impostors (simple 4-surface tree surfaces). Plaster the sim full of those. Load time = low, render efficiency = great.

Last test, go ahead and use a different texture on every mesh tree. The sim will load forever, while still keeping FPS fairly high.

Mesh gives creators unprecedented control over the user experience. Whether creators can utilize that is a matter of skill and the willingness to acquire those skill. That's nothing new in SL. What is new is that building now largely happens outside of SL rather than in-world. That's my biggest gripe with mesh

[Kwakkelde Kwak]

---

Medhue Simoni wrote:

I think you are not taking alot of data into consideration. I just checked a snowman mesh that I made the other day. It's a total of 600 kbs.

---

I assume that 600 is the size of your dae file. That is not the size of the internal SL file (which is probably smaller). The 600 doesn't include the physics shape, it doesn't include any LoD data.

---

Just the textures alone, on a normal prim build would likely be more than that.

---

On a mesh you still need textures. Nothing to be gained by using mesh.

---

UV mapping saves a ton of data when compared to multiple textures on 1 build.

---

Prims and sculpts have UV data just as mesh does. Since they are fixed, you'll have to fiddle with the repeat and offset to make good use of it. This is often possible.

It's true that you can use textures more efficiently on mesh in a lot of cases, especially compared to sculpts, which often need a very large texture just to get detail on one small section. This can save a lot of data traffic. For normal prims this issue is much much smaller, since shape and UV are pretty much a perfect match.

On the other hand, sculpt (and prims) have set UV maps, so this data doesn't have to be transferred over the network. All sculpts share their UV, so that data is in the viewer. All boxes share their UV, so that data is in the viewer, etc.

---

Plus, when you consider that any detailed sculpty build uses many sculpts to create, sometimes dozens, especially avatars, you are talking a ton of data, and you still need textures for all those sculpts.

---

A sculpt map holds all the data the viewer needs to build the shape. A lossless 64x64 jp2 file is only a couple of kb. (I'm looking at 4 kb file size and 8 kb size on disk here). All the other data for the shape is in the viewer.

All the viewer needs to build a prim are the numbers and text you can see in the edit menu. This should fit in 1 or 2 kb.

---

There will always be cases where a prim or sculpty build will be a little less data, but overall, the mesh will almost always win, when made right. A smart creator could save massive amounts data when making a mesh build properly.

---

Less data to be processed by your computer, yes, less data to be sent over the network, I don't think so. Even the textures which are a huge load on the graphics cards (a 24 bit 1024x1024 uses 3MB) aren't all that big on the network. At 85% compression (15% loss) a simple photo (24 bit 1024x1024) is only 340 kb. At 75% this is 195 kb.  It's very possible SL compresses even more than that.

Just like the UV maps, all normal maps are set for all prims and sculpts. Again no network traffic needed.

---

Plus, LODs save a ton of data, and don't add much to the overall total. The results of LODs for the veiwer is even greater, because the pc is not having to download all the full models in the distance.

---

Mesh with LoD saves data compared to mesh without LoD. But at the 2 kb for a prim and 4 kb for a sculpt, I'm sure even the lowest mesh LoD will be bigger.

As far as load on your system is concerned, it's pretty clear mesh is lightyears more friendly. For example, what you can build with 2048 faces (the amount of tris in a sculpt) in mesh is pretty stunning. I am talking entire buildings instead of a single column. Mesh doesn't need hidden faces, like pretty much all prim builds have.

I can think of more pro's for mesh, but I was talking about one big downside. Don't get me wrong, I'm all for mesh, I'm just not sure how a full mesh sim would perform. Time will tell.

[Baal Zobel]

If I were you I would check and see if there are any full mesh builds. If you find one, try IMing the builder. Maybe they have some real information that will help you. As far as I know the only person flush enough and “brave” enough thus far to attempt a full mesh build is Suzanna Soyinka at city of Lost angels, and even that has been a seeming very slow and I would imagine hard learning curve. If I wanted to know some facts from the horses mouth I would start there.
The rest is just speculation. Building a few Mesh buildings and dropping them in a sim is not a full mesh city. You can more ore less ignore any data coming from that kind of area.
Personally I think you would be pretty crazy to spend the time it would take to build a full Mesh city without some very serious reason for doing so. If you are doing it just for show then your time would be better spent building it for a game engine such as Unreal or Unity where at least your efforts will gain you some kind of realistic recognition in the real world.
Even if you can achieve your aim and pull off a full mesh city, you will be displaying it in an extremely outdated environment.
If LL were serious about full mesh builds they should have employed someone to create one , and then they could have written us all some useful documents and we would not be reduced to wild speculation. I have seriously considered a Mesh city for the past 2 full sim builds I have put together. But I do not have the time to waste messing with something so un tried and tested only to find that the tech cannot yet handle it. And neither would I want to be a guinea pig for LL.
It is true that most mesh builds perform bad because they are created by those who have little or no experience with mesh building in a professional capacity. There is a reason for this, anyone that good would not be building in SL.

Apoloogies if this sounds a little negative. I speak as someone who spends a massive amout of my time in SL. and most of that time is spent building complete sims some of which have been large complicated cities. But at this moment in time I would not even dream of attempting a complicated full Mesh sim.