Performances regarding texture compression.

Started by
3 comments, last by Raduprv 19 years, 11 months ago
I was wondering, did anyone notice negative performances while using gl_arb_texture_compression ? I mean, OK, it saves some bandwidth, but it also adds some penalty due to the decompression, right? I am thinking about implementing it in our engine, but I am not sure if it''s worth the effort. So please let me know about your experiences with it.
Advertisement
It''s well worth it. Very nice performance increase, atleast on larger textures. Please note that some compressions can make for instance bump/normalmaps go bad.

My suggestion would be that you implement it and do some tests on the hardware you''re targetting, and then check which textures, if any that you want to have compressed.
Thanks
I will implement it, and see for myself.
We don''t need to update the textures on the fly, and so far we don''t use bump maps, or other ''fancy'' things.
Oh, one more thing. Is the image quality affected? If so, how much?
If you use DXT5 compression, the image will look as good as the original one whereas if you use a DXT1 compression, the output will be ugly...
- Artist on the web -- Lwjgl Ressources & blog -
I am using DXT5, and most of the images look the very same. However, our console image (it''s a dark blue image, with some small variations) looks like shit...
I have a GF4 Ti 4400 with 128 MB, and we do not fill the entire video memory with textures. So I didn''t see any improvement, or any penalty. Hopefully, it will work better on 16-32 MB cards...

This topic is closed to new replies.

Advertisement