Page 1 of 2 12 LastLast
Results 1 to 10 of 20

Thread: DXT1 vs ZLIB Rendering

  1. #1
    Join Date
    Jun 2004
    Location
    Vancouver, Canada
    Posts
    351

    Default DXT1 vs ZLIB Rendering

    Been intending to initiate a discussion about this for some time now and has to do with how the sim renders DXT1 and ZLIB compressed ACE files. Some of you may have noticed a "banding" effect (improperly rendered gradations in color, hue and intensity). When I first noticed this with highly detailed ACE files, I was astonished and so began a little testing. I was especially upset because I had just reskinned a bunch of released hoppers back to ZLIB compressed files from DXT1 compressed files which I believed at the time were better quality. However, I discovered that DXT1 compressed files produce a far superior image in the sim, even though DXT1 is much more efficient being 1/4 the size as well as a little blurry. All of my releases now use DXT1 compression wherever possible (e.g. won't support window transparencies).

    Just curious what other painters/repainters think about this. I find myself converting all downloads to DXT1 for a better quality image in the sim. I always keep 2 copies of the image, one in ZLIB compression format for editing and never re-edit an already DXT1 compressed one.

    I have found the same problem with all types of rolling stock. Here are some examples of ZLIB vs DXT1 rendering, the top image of each pair being ZLIB and the bottom DXT1.
















    Hugh

  2. #2
    Join Date
    Jun 2004
    Location
    Vancouver, Canada
    Posts
    351

    Default DXT1 vs ZLIB Rendering

    Been intending to initiate a discussion about this for some time now and has to do with how the sim renders DXT1 and ZLIB compressed ACE files. Some of you may have noticed a "banding" effect (improperly rendered gradations in color, hue and intensity). When I first noticed this with highly detailed ACE files, I was astonished and so began a little testing. I was especially upset because I had just reskinned a bunch of released hoppers back to ZLIB compressed files from DXT1 compressed files which I believed at the time were better quality. However, I discovered that DXT1 compressed files produce a far superior image in the sim, even though DXT1 is much more efficient being 1/4 the size as well as a little blurry. All of my releases now use DXT1 compression wherever possible (e.g. won't support window transparencies).

    Just curious what other painters/repainters think about this. I find myself converting all downloads to DXT1 for a better quality image in the sim. I always keep 2 copies of the image, one in ZLIB compression format for editing and never re-edit an already DXT1 compressed one.

    I have found the same problem with all types of rolling stock. Here are some examples of ZLIB vs DXT1 rendering, the top image of each pair being ZLIB and the bottom DXT1.
















    Hugh

  3. #3
    Join Date
    Jul 2002
    Location
    oklahoma city, ok, usa.
    Posts
    1,473

    Default RE: DXT1 vs ZLIB Rendering

    Concur. The only way I've been able to eliminate banding is by using DXT1 in AceIt. Some painters refer to DXT1 as "lossy", though, and it seems to me that running the same .ace through TGATool2A and AceIt repeatedly while editing does significantly degrade a texture. For this reason I edit to a separate .bmp and save as a fresh .tga every time.

  4. #4
    Join Date
    Jul 2002
    Location
    oklahoma city, ok, usa.
    Posts
    1,473

    Default RE: DXT1 vs ZLIB Rendering

    Concur. The only way I've been able to eliminate banding is by using DXT1 in AceIt. Some painters refer to DXT1 as "lossy", though, and it seems to me that running the same .ace through TGATool2A and AceIt repeatedly while editing does significantly degrade a texture. For this reason I edit to a separate .bmp and save as a fresh .tga every time.

  5. #5
    Join Date
    Jun 2004
    Location
    Vancouver, Canada
    Posts
    351

    Default RE: DXT1 vs ZLIB Rendering

    Yes.. DXT1 degrades rapidly with repeated editing which makes it only useable as a final output and thus requires a 2nd set of working files. So, seeing no other way around it, I will just keep doing what I'm doing.. :) ..curious MSTS prefers a "lossy" image over a more detailed one?
    Hugh

  6. #6
    Join Date
    Jun 2004
    Location
    Vancouver, Canada
    Posts
    351

    Default RE: DXT1 vs ZLIB Rendering

    Yes.. DXT1 degrades rapidly with repeated editing which makes it only useable as a final output and thus requires a 2nd set of working files. So, seeing no other way around it, I will just keep doing what I'm doing.. :) ..curious MSTS prefers a "lossy" image over a more detailed one?
    Hugh

  7. #7
    Join Date
    Jul 2006
    Location
    Nevada, USA.
    Posts
    298

    Default RE: DXT1 vs ZLIB Rendering

    Do your textures have a "full" (8 bit) alpha channel? MSTS chops this down to 4096 colours + 4 bit alpha (which definitely looks blotchy). It is better to use either 1 bit transparency or no transparency when ever possible. See http://www.mnwright.btinternet.co.uk/train/train1.htm
    ...rich

  8. #8
    Join Date
    Jul 2006
    Location
    Nevada, USA.
    Posts
    298

    Default RE: DXT1 vs ZLIB Rendering

    Do your textures have a "full" (8 bit) alpha channel? MSTS chops this down to 4096 colours + 4 bit alpha (which definitely looks blotchy). It is better to use either 1 bit transparency or no transparency when ever possible. See http://www.mnwright.btinternet.co.uk/train/train1.htm
    ...rich

  9. #9
    Join Date
    Nov 1999
    Location
    London, UK.
    Posts
    570

    Default RE: DXT1 vs ZLIB Rendering

    DXT1 supports total transparency, but not translucency.

    DXT1 is "lossy", but not too bad, it makes up for it using interpolated colors in groups of 4 pixels which is why the banding is less (and why they are about 1/4 of the size)
    You can also apply zlib compression on top of DXT1.

    DXT1 is a 16 bit format so it appears the same in shape viewer as it does in the sim.

    It is also much more memory efficient for video cards.
    Most have native support internally so less memory is used. zlib compressed textures are uncompressed in video memory.

    Overall, unless you're using translucency, I would recommend using DXT1.

    MSTS-2/X will almost certainly support DXT3 and DXT5 formats as DDS files, but these are 32bit formats and do support translucency.
    * Paul Gausden *
    * The Highworth project http://www.highworth.freeuk.com
    * Blog - http://decapod-3d.spaces.live.com/

  10. #10
    Join Date
    Nov 1999
    Location
    London, UK.
    Posts
    570

    Default RE: DXT1 vs ZLIB Rendering

    DXT1 supports total transparency, but not translucency.

    DXT1 is "lossy", but not too bad, it makes up for it using interpolated colors in groups of 4 pixels which is why the banding is less (and why they are about 1/4 of the size)
    You can also apply zlib compression on top of DXT1.

    DXT1 is a 16 bit format so it appears the same in shape viewer as it does in the sim.

    It is also much more memory efficient for video cards.
    Most have native support internally so less memory is used. zlib compressed textures are uncompressed in video memory.

    Overall, unless you're using translucency, I would recommend using DXT1.

    MSTS-2/X will almost certainly support DXT3 and DXT5 formats as DDS files, but these are 32bit formats and do support translucency.
    * Paul Gausden *
    * The Highworth project http://www.highworth.freeuk.com
    * Blog - http://decapod-3d.spaces.live.com/

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
-->