Different PEs

[Pamela Galli]

I have had the same mesh with the same settings upload at costs varying from 3-6.

Now, I have a pretty detailed chair arm upload at .575 with 746 faces.

And when I upload the simple upholstered back, with 376 faces, it is a PE of 3.

 

They have the same LOD and physics settings.

 

Why?

 

 

ETA:  Just tried the back again with the same settings, it came in at 1.48.  Still a lot higher compared to the arm, which I don't understand, but half what it was.

[Drongle McMahon]

I have had the same mesh with the same settings upload at costs varying from 3-6.

Thuis can happen for at least two reasons. (a) using auto-LOD, the generator is non-deterministic and can generate differenty LOD meshes each time (usuall small differences). (b) If you don't have a UV map, it appears the uploader uses uninitialised data, which can make at least two-fold differences in download weight. See this jira for details.

Now, I have a pretty detailed chair arm upload at .575 with 746 faces.

And when I upload the simple upholstered back, with 376 faces, it is a PE of 3.

This sort of difference can be because of sharp versus smooth edges. Sharp edges cost more because the vertices have to be duplicated with the different normals. UV seams can have the same effect.

[Drongle McMahon]

I have had the same mesh with the same settings upload at costs varying from 3-6.

Thuis can happen for at least two reasons. (a) using auto-LOD, the generator is non-deterministic and can generate differenty LOD meshes each time (usuall small differences). (b) If you don't have a UV map, it appears the uploader uses uninitialised data, which can make at least two-fold differences in download weight. See this jira for details.

Now, I have a pretty detailed chair arm upload at .575 with 746 faces.

And when I upload the simple upholstered back, with 376 faces, it is a PE of 3.

This sort of difference can be because of sharp versus smooth edges. Sharp edges cost more because the vertices have to be duplicated with the different normals. UV seams can have the same effect.

[Pamela Galli]

Thanks Drongle. Still working on it, trying to isolate variables but all I come up with is time of day.  Maybe it's the sim.

 

 

 

[Pamela Galli]

I uploaded an upholstered chair back, it was about half a prim in LI. Today I tried several times to upload it, and the smallest LI was 28. 

28!

Yes, it was the only mesh exported.

No it was the same dimensions as before.

 

it's not just this mesh, though it is the most extreme example.  The same mesh is often 2-3X larger than on a previous upload.

[Pamela Galli]

Tried again with the chair back.  I could never get the button to calculate LI to be available.  I don't see many threads reporting these bizarre upload issues, tho, so guessing it this is a more isolated thing.

[Maeve Balfour]

A longshot in the dark here, but I'll mention it anyways just in case...

When you say your PE is fluctuating, am I correct in assuming that you are referring to the numbers WITHIN the uploader window itself? (without actually uploading it to rez inworld?). The reason I ask, is that at times for me, the tab where you set the upload's size parameters (I can't recall its exact name (typing this post at work)), where you set the import size percentages, I can sometimes get random imputations there in the uploader. Normally it is a standard default, but I get caught sometimes with it jumping around and giving me strange PE results. Since this affects the actual default rezzed size of your meshes, IF it's changing without your knowledge, it can potentially skew your PE figures by a big degree.

So if you aren't already doing so, try keeping an eye on that setting... just in case it is changing without your knowledge.

:matte-motes-smile:

[Pamela Galli]

Thanks for your answer Maeve - yes I notice that size number fluctuates.  In this case, the PE is radically different even when that number is the same.