Designing for a low Download Impact?

[Pandora Wrigglesworth]

I' very confused as to how download impact is calculated.

I created a mesh with 916 triangles. I made LOD models with 771, 516, and 309 triangles each.

When I upload my mesh, using all four LOD models, I wind up with a download impact of 15.6.

If I switch to using just my low and lowest LODs (516 and 309 triangles), I still get the same download impact of 15.6.

I tried removing all sharp edges and completely flattening the UV map to eliminate seams and the download impact actually went up a little.

If reducing the number of vertices does not reduce the download impact, what does?

[Kaluura Boa]

You are not the only one to be confused by the uploader. From what I observed through various experiments, lowering the number of triangles of the lowest LOD has an enornous impact... I would even say "insane"!

To have some numbers to put in this post, I fed the uploader with a high-poly model and played with the triangle limits for the various LODs.

The model is 13376 triangles and the uploader "proposes" me LODs of 3344, 836 and 419 triangles, for a download (DL) of 44.7.

If I double the medium and low LODs to 6687 and 3343 triangles, the DL barely climbs to 50.9.

If I double the lowest LOD too to 836, the DL doubles too: 86.5.

If, on the contrary, I divide the lowest LOD by 2 at 213, while keeping the higher numbers for medium and low LODs, the DL looses roughly 1/4 at 31.8.

And, in the end, if there is a formula behind these numbers, it makes no sense to me. My wild guess is that to increase the number of triangles in medium and low LODs has a moderate impact on the final numbers but it is best to stick to whatever number the uploader "proposes" for the lowest LOD.

Any way, since yesterday, I decided that I refuse to even try to understand the uploader. I moved 4 triangles... Just that! Somewhere in the file, the definitions of the triangles changed to use some other already existing vertices... and the land impact was divided by 2! So... YMMV and YGIAGAM...

[Drongle McMahon]

The effect of each LOD on the download weight depends on the size of the object. This is because larger objects have to be rendered at higher LODs more of the time. The details are described in this thread, The useful numbers to remember are about 6.25, 12.5 and 50 meters. At these dimensions, the lowest, low and then medium LODs cease to have any effect on the download weight of a mesh with a cube bounding box. For a flat square, the critical sizes are about 7.5m 15 and 61 meters, and for a pole, 11, 22 and 88 meters (the last of which you can't reach of course). Around 1 meter, the lowest LOD has an overwhelming influence. Between these sizes, the influence gradually shifts from the lower to the higher LODs.

[Winter DiPrima]

From my experience it's mostly the physics shape that  really affect LI. Try using just a 3 vertex polygon as a physics shape and see how that goes.

[Medhue Simoni]

It's your lowest LOD. 309 is kind of alot for the Lowest LOD, depending on what it is. Plus, a proper physics shape. You can't ever just use 1 of your LODs for the physics, It will always be more than you need. Making a physics shape for every mesh is a must. The good part is, they are easy.