Jewelry workaround

[Rhys Goode]

I've been working on a necklace, ~ 20 cm diameter, and very thin, so the LOD changes to the lowest  level with the camera fairly close to the avatar.  Even using reduced vertex models for the lower LOD, the land impact gets pretty big if I want it to show decent detail from 1 or 2 meters away.

I just tried adding a long skinny rod to my model, with a different material type, so its easy to make it invisible.  With a ~75 cm long rod, the LOD does not change till I am far enough away that it does not matter.

Using reduced vertex models, I can't get the land impact below about 25 without this trick.  Using the trick, I can  let the system do whatever at lower LOD's and my land impact stays around 2.5.  I see great detail where ever I move the cam, moving back far enough to change the LOD, its too far away to see the crude models replacing my detailed model.

It seems like I am getting something for nothing here.  Is there some down side that I do not see?

 

 

[Drongle McMahon]

How are you arraging the thin rod relative to the ring? If it's passing through the ring and perpendicular to it, that is giving you about 6 times the radius for the comboned object. That means the LOD switch points (ignoring close-up adjustments) are 0.8, 3.2 and 6.4 meters instead of 0.14, 0.56 and 1.12 meters.

Now suppose you need to see the full detail at 2 meters, then without the rod, you will have to use the full detail at all four LODs. With the rod, you only need it for the first two LODs. Having the full detail at the lowest LOD is very bad because the gpu has to render all those triangles however far away the camera is. With the rod, the longer switch distances mean you can use the lower detail meshes, and the gpu can render the lower detail for most camera distances within viewing range.

So that is where the saving comes from. It does really save gpu load, and is therefore correctly rewarded with a lower download weight. What you describe is a legitimate way of reducing the resource consumption and receiving the appropriate lower land impact.

The exact numbers here will probably be inaccurate, because there are some adjustments in the code to the LOD oiints when you are close to an object, and because your geometry may be different, but the priciple is the same.

[Rhys Goode]

Oh, thank you for the numbers!

Yes, passing through and perpendicular, the notion was to increase the effective radius.  I knew that lod switch points are  (mostly) in relation to radius, but not the details. Your explanation not only tells me that the idea works, but makes it very clear how and why.  And why it makes sense.

Thanks again!