the texture seems to be 1024x1024x32 -> 32 bit one. Its also a non-transparent one.
Btw, in this article i read that there seem to be different settings for .dds that are particulary good/bad in handling for example transparency:
DDS DXT1:
Low levels of loss in RGB channel, only supports 1-bit alpha. Great compression, and uses 1/4 of the texture RAM of lossless formats. Ideal for items that aren't using detailed alpha channels.
Hard to use, if you want professional results. All of the DDS formats require quite a bit of practice to get good results out of the filtering available in the DDS plugins / conversion tools available. If you don't have Photoshop, I recommend using the plugin for GIMP or ATi's Compressionator application, to get good control over the settings. However, I'm just going to be real and say that, having tried all of that out... the Photoshop plugin is markedly superior, in terms of detailed control. DDS DXT3: Low levels of loss in the RGB channel, alpha is lossy 8-bit but tends to come out "crisp", due to compression algorithm used. Uses half the texture RAM of lossless formats. Needs sharpening applied to mips for good results, and tends to work best if you accept some high contrast where you might want subtlety ideally. Good for "glow" style teamcolor, where the alpha compression is acceptable, but works fairly poorly for "paint" style, where you need more subtle alphas. Can sometimes cause major artifacts in RGB channel (blocks of area where the compression method doesn't handle the area well), which are very hard to get rid of. DDS DXT5: RGB channel is usually a little more lossy than DXT3. Alpha channel is more "soft" than DXT3. Uses half the texture RAM of lossless formats. DXT5 is tricky, and you may have to adjust sharpness / contrast between mips per piece, due to the way that the compression works- it's not as consistent as DXT3, in terms of artifacts. Alpha is better for "paint" style teamcolor. Works far worse for "glow". Artifacts may shift color values enough that it's a problem for texture2, where tiny changes in RG values make a big difference in terms of what you see in the engine. Can sometimes cause major artifacts in RGB channel (blocks of area where the compression method doesn't handle the area well), which are very hard to get rid of.
Anyway, i still dont know why its not working?