Talk:Choose the right DXTC compression algorithm

This is a very interesting discussion, but it seems like this article is quite a bit quicker and easier to use... Perhaps a link to it would be appropriate?
Dragoon Wraith TALK 10:11, 6 July 2007 (EDT)

8.8.8.8 ARGB

From this topic on the Construction Set forum, a post by Ghogiel about normal maps and compression

If the DXT compression is destroying your normal maps use 8.8.8.8.

However 1024x1024 is a bit too big for a sword- especially 8888. Without using DXT compression, a 512x512 will likely look just as good and will be a 1/4 the file size, seriously. Often its virtually impossible to see a difference unless you are really inspecting it. Even zoomed right in on the largest mipmap. Always do a little test to see which resolution/mipmap profile/compression yields the best result for the least cost in file size. Please don't send out a 5 meg normal map.

I made a topic on my experiments for normal map compression. Results were this. http://i25.photobucket.com/albums/c85/lego...malmapping3.jpg

In most cases where I was really setting in large height changes in the normal map, DXT compression didn't deal with it at all. Even on less organic normals. I found on organic deep flowing normals with smooth transitions in height, saving as a reduced resolution 8.8.8.8 won out over DXT every time. for my eye at least. the pixelation artifacts DXT puts into deep normal maps is horrible. Because the normal map bends how light reflects off the mesh surface makes it more important to sort out the ugly compression then the diffuse map.

more information in the thread about this subject here.