Jump to content

DOWNLOAD MODS

Are you looking for something shiny for your load order? We have many exclusive mods and resources you won't find anywhere else. Start your search now...

LEARN MODDING

Ready to try your hand at making your own mod creations? Visit the Enclave, the original ES/FO modding school, and learn the tricks of the trade from veteran modders...

JOIN THE ALLIANCE

Membership is free and registering unlocks image galleries, project hosting, live chat, unlimited downloads, & more...

A good thread explaining texture optimization


ThomasKaira
 Share

Recommended Posts

Martigen created this thread when he noticed a notable number of texture mods being released for Skyrim are lacking basic optimization.

Thread is here

The most common mistakes are using the wrong compression algorithms for specific textures (like saving a fully opaque texture in DXT5, such a texture has no use for an alpha layer) and forgetting to generate mip-maps (increases VRAM usage and causes a shimmer effect on distant normals)

It does not point any fingers, instead Martigen explained in brief what the various missing optimization tricks do and why it is necessary to use them. Also, the discussion is quite enlightening on certain things (like providing links that explain how the DDS DXTC algorithms work).

If you are interested in making texture mods for Skyrim, I would recommend giving this thread a glance. There is a lot of good information in there.

Edited by ThomasKaira
Link to comment
Share on other sites

Last time I saw Martigen about was when I did some textures for his Mutant mod.

Its definately a useful thread for newcomers but there are some minor points I'd like to point out.

DXT1: While this is the format you want the vast majority of your colour/diffuse maps saved as you have to remember that this has a compression ratio of 8:1. It can result in lossy image quality, certain colours can also bleed and have unwanted discolouration but these artifacts can't be avoided and are a result of the very high compression.

One of the biggest mistakes I see is people painting over the original DDS texture and saving it again. This has the result of multiplying the compression further. 8 times compression saved again becomes 16 times and so on. This is why you start to get horrible artifacts. The best way is to increase the original texture to the size you want and use it as a template/guide to paint an entirely new texture, obviously you delete the original guide before saving your nice new texture.

DXT5 is superior to DXT1 simply because DXT5 uses 4:1 compression as opposed to DXT1 8:1 compression. Of course DXT5 needs to be saved with an alpha channel so its not something you'd typically save as for colour maps.

Martigen suggests DXT3 or DXT5 for normal maps. Use DXT5 in all cases its an improved format over the older DXT3, the quality is better and the file size is the same. I used to use DXT3 in some cases but it was pointed out to me by other artists on Nexus that I should always save as DXT5.

Not saying Martigen is wrong at all, I just thought I clarify a little. :D

Link to comment
Share on other sites

DXT5 is superior to DXT1 simply because DXT5 uses 4:1 compression as opposed to DXT1 8:1 compression. Of course DXT5 needs to be saved with an alpha channel so its not something you'd typically save as for colour maps.

The reason DXT3 has a 4:1 compression ratio is because an additional 64 bit are being allocated per 16-pixel grouping for alpha channel information. So the compression for 16 pixels is 32 bit for color, 32 bit for the indices, and 64 bit for the alpha channel. DXT 1 uses 32 bit for the color and 32 bit for indices while leaving out the 64 bit for the alpha, so the compression ratio increases. Exactly the same compression for colors as DXT3 and 5. The compression ratio is higher simply because alpha channel information is not being saved for the texture, which results in less space being needed to store those 16 pixels.

Here's an explanation.

Simply put, that DXT1 results in a larger quality loss over DXT3 and 5 is a myth. Each compression algorithm uses exactly the same color interpolation to compress the diffuse.

Also, it was suggested later on in the thread to store normal-maps at 1/4 resolution using 8.8.8.8 uncompressed, because DXT compression leaves undesirable artifacts on the normal maps. Reducing the resolution but not compressing the color data in a normal is the better option because it retains the original precision that the normal was generated with (which is fairly important if the normal was baked from a high-poly model). DXT compression mucks with the normal's coloration through the interpolation, causing artifacts.

Edited by ThomasKaira
Link to comment
Share on other sites

ThomasKaira: Well all I can say is that I'm only going by what I've read in 3d books relating to DDS Textures as well as what other people have told me. Still working over original DDS textures and resaving is not a good idea, so I stand by that comment. DXT5 is also bettter than DXT3 and saves typically result in exactly the same file size.

Regarding Martigen mentioning the ATI Compressonator. I have to say I think the results are far better than the DDS Plugin from Nvidia.

Heres a link to an example to download. You can see the difference better if you zoom in. There are colour discolouration artifacts, noticeably green and purple with the Nvidia version. Please download the image its easier to see the difference in quality.

Link in spoiler tab

http://www.mediafire.com/?044vftnnphcc17s

Link to comment
Share on other sites

ThomasKaira: Well all I can say is that I'm only going by what I've read in 3d books relating to DDS Textures as well as what other people have told me. Still working over original DDS textures and resaving is not a good idea, so I stand by that comment. DXT5 is also bettter than DXT3 and saves typically result in exactly the same file size.

This I agree with. Particularly with Normals, you can definitely tell when a DDS texture has been double-compressed. Not pretty.

DXT5 is better than DXT3 for textures with gradual transparency, but for a flat alpha, it doesn't really matter which you choose. The only difference between the two is how the alpha channels get saved.

Link to comment
Share on other sites

I have found in personal experience, that saving in the 8.8.8 or 8.8.8.8 format with a "smaller" image looks better in-game than an image saved at DXT5 that is much larger. Especially with highly detailed textures where you want the normal map to really pop out of the surface.

So, you could save a 2048x2048 image in DXT5, or 512x512 in 8.8.8.8 and the smaller one looks better. Not always, but most times. Just my :2cents:

Link to comment
Share on other sites

  • 2 weeks later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

×
×
  • Create New...