LOD Factor; My Guilt. (Part 3)

Now that we’ve discussed the consumer frustrations and what LODs are, we’re ready to talk about the biggest bandaid enabling them – the debug setting of RenderVolumeLODFactor. The wrong solution all along – allowing us to ignore the quality issue on the viewing end, instead of following good practices on upload.

I’d like to preface this with the fact that many of us had crazy RenderVolumeLODFactor long long before mesh even arrived, as sculpties were notorious for this same issue. Instead of allowing us to make special LODs, they generated their own rather terrible ones that largely consisted of devolving into scribbles. I think upping this setting back then was totally justified. This is a widespread, and longstanding issue. We have a long road to fix it.

So what is LOD Factor, and why do I care?

RenderVolumeLODFactor as seen in the Debug Settings

LOD Factor is short for a SL Client debug setting called RenderVolumeLODFactor, a direct portion of the algorithm that handles the way mesh transitions to lower LODs on the client end. It is a multiplier in how far away from our camera they step down. Or: In the case of many hacked up product LODs, dissolves rapidly into triangles.

Many LOD Factor Evangelists, guides, creators, and notecards misguidedly inform others to jack up to 4, 8, 16, 32, or even higher to work around meshes which have not been optimized. It was discovered that anything over 8 basically plateaus, and that 3-4 were ideal for day to day use.

This setting affects every single item in every sim you visit – mesh, sculptie, and even worn items that you see on other avatars. To include things whole sims away, regardless of if they were optimized. Such high multipliers put an unreasonable burden on your PC by forcing you to draw and download super high detailed meshes at distances that you could never possibly notice them, much less appreciate.

The Firestorm Quick Prefs Panel

Linden Labs is aware of an issue of poorly optimized creations, yet it is hard to control a user-generated world except to put systems into play to encourage them to operate by good game dev guidelines – the land impact (LI) algorithm and the avatar rendering cost (ARC) algorithm. Even our Uploader tries its best to generate sensible levels for us.

LL, third party client devs, and community members met and came to a resolution of capping the LOD Factor on client ends to 4. This was twice the default Firestorm value (2), and nearly 4x the default LL Client value (1.125). LL also upped the land impact allotments on most mainland and private parcels. All which was to try and help in fixing the artificially lower LI or ARC numbers caused by using LOD Factor as a crutch to ignore our poorly made items.

The sad truth is we have ignored these helpful nudges. Our higher LOD Factor lets us turn a blind eye to these Low and Lowest LOD settings for many products. The only reason that so many blogs, notecards, profiles, and SL resources recommend turning the LOD Factor up so high is because: the consumer has no choice if they wish to enjoy the products they paid for from many high end SL brands that were just following along in tricks passed to them by friends.

We are the same, and like our consumers, we set our LOD Factor to 4 a long time ago and forgot about it. Now, when we create and upload an item, we only check to make sure it looks good at that Factor 4 setting without any regard for those on the default settings.

It is nearly impossible for Linden Labs to make us fix an issue that we ourselves aren’t all aware exists – no extra land impact allowances, forced debug settings, or encouraging posts can fix a deep-rooted lack of understanding.

A quick detour down Land Impact Lane.

Take a look at the Advanced tab and look at your Weights!

Land impact is part of the reason for how we all got here. It is what all Home and Garden designers strive to keep low, but also the root cause of why we have messed up LODs – to cheat the LI algorithm. For wearables, just replace this with ARC.

As explained in this knowledgebase article, land impact is a calculation of all aspects of our mesh, including inworld size, textures, polygon count, physics, and scripts.

A simpler way to think of it is:

  • Download Weight – anything we visually see client side
  • Physics Weight – anything that physically interacts with others
  • Server Weight – anything the server has to run or calculate

Through a combination of these, and either optimizing or cheating, is how we get our standard Land Impact or Avatar Rendering Cost numbers.

One of the singlemost heavy impacts besides physics or textures (a whole set of articles in and of itself), is the LODs impacting the Download Weight. This is directly tied to size of the in-game assets due to how LODs work.

LODs are also the easiest way to cheese out the algorithm to get a lower land impact or avatar rendering cost without actually evaluating your work for ways to trim down. Just set your Low or Lowest to 0 and be on your way. Cheating them, sadly, is only cheating ourselves and our consumers.

Textures should have been a heavier weight in land impact, but sadly were not considered. However, in our upcoming texture series I’ll discuss how these are the true killers of otherwise beautiful works of art.

My own guilt.

I know all about this trickery, many of the older creations in my store were created with such laziness, and I’m sorry. As someone who hungrily and blindly pursued that 1 land impact number, I am sorry. I can’t make them go away, but in the future I can do better.

In my store, I’ve always wanted to make low “prim” or as we have now, “land impact” items because I catered heavily to roleplay sims from the offset. Many of these sims had stringent, often brutal allowances forcing you to furnish whole homes on something absurd like 25 land impact. Their reasons aren’t unjustified – on these often crowded sims, it was the sane way to allow for building a beautiful environment without overloading the sim or viewer lag issues. In prim days, this meant doing wild contortions, abusing textures, learning tricks to cut and twist torii.

An early creation of a brazier, bottom level set to 0. using only generated LODs.

When mesh came out in late 2011, we had new and shiny land impact algorithms, and an unprecedented opportunity to squeeze crazy detail out of miniscule numbers as easily as ticking that Lowest down to 0 (as shown above). So I pursued low land impacts with a zeal, learning fast and early my tricks, what would yield a very low land impact item. My LOD Factor was set to 4 and long forgotten, so I and others released item after item for years without recognizing that for the average user, our items turned into nothingness sometimes at very short distances.

The creators I worked with and myself came up with workarounds for everything on the sole basis of Land Impact optimisations. Tips such as falsely inflating the size of items to trick the LOD, even getting rid of multiple LODs by setting to 0, without any regard for what we were actually doing. The hacks we used worked, our items were low land impact, and almost no one ever complained – so why change? Consumers didn’t speak up, they showed their approval by buying the items en masse. They spoke with their wallets.

Reality & when things are too good to be true…

I reinstalled SL one day on the default client as Firestorm was giving me issues uploading, had a revelation that a good half of my store looked like angular black-pocked messes. I was shocked – I never knew this is how things really looked. I’d been in my LOD Factor 4 bubble for so long, and that actual default yielded a jagged world unless I zoomed in.

Example of illustrating a triangulated medium LOD from a collaborative designer.

My friend’s stores and creations were the same – triangular disasters, some of them worse than myself – and all along I’d been a creator who openly endorsed dropping these standards. I had earned a reputation as someone with ‘magic’ for making things low land impact. There was no magic, just dirty tricks fooling myself into contentedness. Now that turned into guilt.

So I changed my workflow, and began examining what makes that land impact algorithm tick from the ground up. Weeks uploading and reuploading the same mesh, trying to make my conscience happy while still sating that need for 1 land impact. I read, researched, asked. How could I maintain that unrealistic standard without using bad creation methods?

Finding my happy medium.

It took me a lot of time to research and develop what I wanted to provide as representations of my brand and creations. I went to bed mulling over it, scribbling lists on my phone of ideas and goals.

Ultimately, I determined I wanted:

  • Items to be as low land impact as possible, ideally 1
  • LODs to reflect the intended use and scale of the item
  • Acknowledge 0 in Low or Lowest was acceptable sometimes
  • Small items needed all 4 levels in some form, even if 2 were shared
  • All items should be tested and satisfactory on 2.0 LOD Factor
  • I didn’t want to spend more than 20% extra time per project
  • Indoor items should be recognizable for 40-50m
  • Outdoor items should be recognizable for 100-150m
  • Recognizable does not mean detailed, it means “passes as the object it is”
  • To maintain a personal amount of integrity in releases
  • You are not a wizard, Anke. 1 LI isn’t always possible.

Thanks to everyone who is taking the time to read my rambles about the issues of LODs, and especially to those who are having an introverted look into themselves to change their own methods as I have.

In my next post, I’m hoping to give creators a quick but not-so-dirty method for reducing their LI and ARC numbers within a sensible period of time, without resorting to lazily dropping Low or Lowest to 0 most of the time.

This next post will take a long time to refine and edit, so it may not be released as hastily as these have been. It cannot be all-encompassing, but should provide a good at a glance resource for getting to a better place.

Leave a Reply