Texture map question

I got some items that are part of an open source build. The textures are BEAUTIFUL, but, (my guess is,) in order to save upload costs, the artist mapped multiple items to 1024 x 1024 sheets and by using offsets and repeats, used one texture to map multiple items. With all these huge textures, lag was killer.

I downloaded the textures, reduced size to 512 x 512, re-uploaded and re-mapped. This had the the desired effect of eliminating the lag, but, in some cases, there was severe degradation of the perceived images.

SO, I was wondering. Should I cut these textures up? I know that I could get better resolution, and the upload cost, while significant is acceptable. But, the re-texturing would be a lot more work, and I wonder if the additional individual textures would offset the smaller sizes of the textures.

Also, I'm aware that I've taken someones hard work, and redid it in crayon. The textures were full perm, and freely given, and described as "open source," but this probably isn't what the artist had in mind.

Any comments or suggestions?

Technically using one 1024 x 1024 texture map should not cause any more lag than four 512 x 512 texture maps. In fact, using a "texture sheet" with multiple textures on it usually loads faster, since there is only one file load and all the textures on it will blink into rezzing at once.

Still... 1024 x 1024 is pretty big. So try cutting it into 512 x 512 panels, if that works.

I generally use 256 x 256 or 512 x 512 for my building textures. I only use 1024 x 512 for complex walls with alpha-mapped windows that need a lot of detail to look good.

Just to add: You may have (or not) redid them and saved as jpeg's. This will lower quality too. Stick to tga's.

I'm just wondering, what is the purpose of reconstructing the work? Is it just to reduce the lag? If that is the case. you might find that it is worth investing in a better graphics card so you can appreciate better things overall, instead of spending the effort so that you can view 1 object on some lower end hardware. They don't cost that much these days.

The actual reason for doing this is to stop lag. Sure 1024x1024 *is* big, but how many textures are on that one image? If they are using this technique properly, then when you split them up, you will have not only more total texture memory used, but also more images as well.

Assuming the worst for this technique, and they are only using a 2x2 grid of 512x512 (Note: There could be *many* more textures in the one image if done well, 10, 20 or more

Both will use the same texture memory.
1024x1024: one download, one file to cache, and just one to load, smaller total file size.
512x512: 4 downloads, 4 to cache, 4 to load and manage, larger total file size.
So as you can see, it will reduce lag, compared to *same* person doing the other way, even for bad cases like this (1024x1024, only 4 textures per image). Better cases will have *many* more textures per file.


If someone knows what they are doing with the textures, then yes they should redo them more efficiently, if possible, and offer them back to the creator (as you said it was open source).
Of course this is only for those who really know how to get the best out of SL (and note, you would want the original, non-compressed versions to start from, so you would need to contact the creator before you did it), and know all of the tradeoffs between image size, file size, number of textures per image, resolution, and the rest.