As we've been discussing, Second Life mesh, especially the most popular avatar bodies, tend to be extremely resource heavy, causing you (and everyone around you) low FPS, unrezzed graphics, or what's colloquially called LAG. Unless and until Linden Lab starts curbing this excess, watch this great tutorial by NWN's Cassie Middles, who explains some tips on improving your SL performance. Top tips include:
Be careful what you wear in certain areas. They might add to the lag of the sim based off of the Poly count, VRAM, scripts, and more.
Learn how to spot VRAM, poly counts, and scripts. There's an easy way to help you decorate a sim without being too lag heavy.
Derender people that have super high complexity and items or click 'show friends only.
Designers, look into trying to put more of the textures onto a single texture face, and also at a lower size. 512 x 512 will perform much better lag wise than 1024 x 1024. Weigh out which is okay to use with lower detail vs the moments of higher detail.
More tips to add? By all means do so below!
As previously explained, complexity is meaningless. I honestly wouldn't give two cents about it in any other Viewer other than my own its far too forgiving, random and punishes tiny things a thousand times more than real framerate killers. I suggest hand-picking people to jellydoll/derender based on their VRAM usage and polycount as well as perceived complexity.
Wether to use one bigger texture or multiple smaller ones is a complex topic.
Smaller but more textures allow you finer control over how detailed and resource heavy every single face is but they will incur texture allocation impacts, which are arguable how much they really impact the performance. Textures are loaded and decoded once unless they are released either because the Viewer has to or because you left the place (or looked away long enough). So technically the impact is a one time thing which would mean having more textures but smaller ones sound like a great deal but having one bigger texture with everything included isn't exactly bad either, it saves you a lot of texture allocations but might waste a lot of space depending on how much of the texture area you fill in for actual use. In general i'd rather say use smaller but more textures because it also allows the Viewer to mipmap them easier and separately making texture memory optimizations algorithms more efficient in doing what they are supposed to do.
Posted by: NiranV | Friday, December 14, 2018 at 02:45 PM