OK. What did you guys do to LOD in 1.10.2?

Right, 24, not 20... either way, it's not a power of two.

Well, other than for small values of 3.

well it doesn't absolutely need to be a power of 2 but powers of 2 tend to work the best simply because at different levels, the major vertices still line up, aka there is not very much mis-alignment between two shapes of the same size, one 16 segments, one 32. half the vertices still line up.

Now we have things that are not even multiples tesselating out so you can have for instance one shape be 24 while the cylander it touches be 16 and the two cannot possibly properly align. Also degrees of rotation can end up uneven and also impossible to properly align that used to be fine.

well, I think overall the maximum level of detail should simply be raised, and settable if the client/user so wishes. Detail shouldn't be capped.

Default can be lower and more low-end mid-end friendly, but designers simply need the detail so that their projects can scale.

Does "Object Mesh Detail" also deal with an avatar attachment's visual quality? Or is that a seperate slider?

Avatar attachments are objects, so object mesh detail affects this, not avatar mesh detail.

Meanwhile, can we bump this? Can we please look at having the option to have more than 24-sided round things in 1.12 if we choose to?

I still don't think that we need to be protected from our own frame rates if we choose to push the slider all the way up.

'24agons' are annoying. When I first arrived the most striking thing about SL was the quality of the 3D graphics.
I don't know what's going on, but there seems to be an ongoing incremental decrease in both graphic rendering quality and overall performance.
Fairly obviously something is messing things up, because for both graphics rendering and overall performance to be degrading there must be other factors responsible, 'something' is grabbing the lion's share of cpu, causing graphics performance to have to be throttled to unacceptable levels.

My big gripe, other than the blocky prims, is the distant landscape. Its no longer possible to see even half way across an island without the 'horizon' turning into a jagged mess!

Its a big problem. One I suspect will be ignored in favour of yet more gimmicky modifications of dubious worth to the UI.

Well according to robin's post, we very well may be stuck with this for a very long time, if not forever.

If we could get a groundswell of support, maybe. But there are less designers than non-designer residents. Course I guess now we don't have to worry about making stuff look as good anymore. <sigh>

If it is any consolation, it is sometimes possible to get the same torus shape in different ways using different combinations of x hole size and x size. Putting x hole size to 1.0 and altering the x size appropriately may help with the LOD. I guess I'll go back to my stuff when I get a chance and see what can be fixed this way.

More badness:



Thats a 15 sided polygon... 15 is a terrible number of sides - it doesnt divide by 4 so even 90 degree rotations will not line up. Just to reiterate, when trying to be precise with matching up vertices, this stuff makes it literally impossible to do.

Cylinders always have the old maximum of 24 sides, investigating with Lex Neva, we've found tori with cross-sections with 24, 15 and even 9 sides. It seems to depend on the x and y hole sizes somehow, and of course many shapes cannot be made without altering those parameters to the point that the number of sides goes down and vertices no longer match up.

If the X hole size is 0.4 or below, or the Y hole size is 0.2 or below, the maximum number of vertices in the cross-section drops to 15. If the X hole size 0.2 or below, or the Y hole size is 0.1 or below, the maximum number of vertices in the cross-section drops to 9. _9_. It's a really ugly looking 9, too.

I'm still here and observing. Also checking with Runitai for latest details. Thanks for the emails, Michi. Really appreciated.

[UPDATE] Runitai and Josh let me know that an improved LOD slider for Object Detail should be in the next release of Second Life which is 1.11.1 planned for this week. Please look for it soon.

Of course, it's hard to say how you feel about it until you see it in action--so we'll keep this thread ongoing for feedback.

Looks like its all back the way it should be from the preview! Thanks!

...almost...



This is a torus, parameters: Size: <0.5, 0.5, 0.5>, Twist begin 180, end 0 (this has a half twist in it, like a Mobius strip), hole size x: 0.60, y: 0.25. The cross section at maximum zoom (and my object detail slider is all the way to the right) is a 15-gon, which is why it doesn't line up with itself rotated 180 degrees. Very strangely, if the path cut is not 0.00 to 1.00 then we get the full detail 24-gon as expected.

With the x and y hole sizes small enough it will go to a 9-gon as well

{edit: There is a way round though - just use two half tori instead of one}

since the avatars arent prim capped (well they are but poorly) what about allowing a certain slice of rendering time to each avatar on the client so if an avatar need more time to be rendered than the rendering timeslice allow, the avatar loose details until it fit into the timeslice?

it would effectively penalise the narcissic torii fans , encourage peoples to build in an intelligent manner and use textures instead of stacking primitives like if it was nothing performance wise.

Interesting idea, Kyrah, but it gives a person very little information to go by. What may be very time-intensive to render for one person's system may happen very quickly on another's. If my system is relatively speedy, how am I gonna know what my av looks like for someone whose system renders my tori using 5 triangles? A limit like this is really difficult to work with, because the same avatar will look vastly different on many people's systems. The only real solution to produce an avatar that everyone would see the same would be to redo the av as very low-detail, or not at all.

Instead, triangle or prim count limits on avatars might be better... at least that way the maker of an av has half a chance of being able to deal gracefully with the limit.

I'm somewhat against arbitrarily introducing permanent system-wide limits to hamper creativity for years to come because a few people are trying to connect up on hardware that was already obsolete 3 years ago.

I'd say its probably a better guide to allow some form of detail preference so that objects could be given some type of priority by the creators. If everything was set 'high' it woudn have any effect, but if some parts were set higher/lower aka telling the system to prefer adding detail to one part of a structure by taking some away from another..

Someone runing maxxed would never even notice the difference, and on average settings the person making the object would have some level of control over the granularity over which 'culls' within the object happen, aka it would stop random corners from protruding out from inside of another object oddly, as detail is raised/lowered as say you walk past something.

Basically just some form of 'hint' to the LOD renderer as to what parts the creator would like to see 'culled out' first, or held onto last, as the various LOD phases are run through on someone's client.

Its one of those basic ideas though where every year that goes by, the percent of people running 'average' hardware as opposed to 'really old crud' will remain the same, but the actual *POWER* of that average hardware will go up just as much in a year, as the average power of the highest end cards does... Just 2-3 years behind the 'curve'

So while the average user today might have an intel chipset, about equivalent to a geforce 5200, next year the average user could have something more closer to a much more capable 7300.

Arbitrarily limiting the system today, to work with 2004's 'average' hardware, will just serve to long term 'lock' second life into looking old/antequated/ugly.

LOD 'hints' would allow all spectrums of detail settings to work better with all spectrums of possible user hardware, allow the system to scale higher and higher as people do get better systems over time, and will ensure that a content creator can at least get *SOME* form of acceptable shape/style across a broad spectrum of the possible detail slider levels.