Home Unreal Engine

[UE4] ~ Packed png VS unpacked jpeg maps, number of samples or video memory?

triangle
Offline / Send Message
Steamy_Steve triangle
Here's another "shower thought" of mines: jpeg maps take *very* little space, but at the same time they dont have an alpha channel; furthermore, they nearly always present some minor compression artefacts, no matter how much you crank up the compression quality.
On the other side, the same maps as png can easily take up several megabytes, but their better quality is undisputable and they do include an alpha channel to store extra information.

Now, there can be objects that are so minimal or small, in a scene (eg: smaller foliage elements), that eventual compression artefacts would pass unnoticed or even be completely invisible.
I'd say using jpeg maps could help save a *LOT* of video memory, in such cases, and possibly even speed up renderings (I'm not techy enough to assert it), but at the same time you might need some extra files that would otherwise be packed together into a lesser number of input samples.

Thus the question is: given the compression quality is not a factor, is it better to pack maps into the lowest number of files possible....or to use a more compressed file format to take the lowest amount of memory possible?
And why one or the other? Maybe both?

Replies

  • Alex_J
    Offline / Send Message
    Alex_J grand marshal polycounter
    I dunno too much about this type of stuff, but what I've been told by pro's is that targa is what is mostly used in games.

    I also read in some older thread here that it's usually better to use a grayscale image for alpha's, rather than a packed alpha channel. I dunno enough to justify these answers, but the main point is, you got more than two choices.
  • poopipe
    Offline / Send Message
    poopipe grand marshal polycounter

    Tga doesn't support high bit depths,
    Jpeg uses lossy compression and doesn't support high bit depths

    They are both shit, don't use either. 

    The textures are compressed by the engine so their size on disk has no influence on how much gpu memory is used. 

    The higher the precision of the input data, the better job the compression algorithms will do of maintaining information

    Packing is done to reduce the number of texture reads the gpu has to do, as you note it is not done to save memory - in fact it can cost more. 

    You should use 16bit png in 99% of cases and stuck to the conventional packing setup everyone advises for ue4 unless you have a good reason not to 

Sign In or Register to comment.