New ATI/NVidia-Cards and dithering...

Pyro · Post by **Pyro** » Wed Mar 18, 2009 5:35 pm

Yeah, I know you aren't trying to insult, just saying. I just thought of something that could explain what you are seeing. It is possible the maps were not ported exactly. So when the map from TFL were saved and made into Myth 2 maps its possible some quality was lost in between. So it is possible it had nothing to do with the 2 game engines just some quality loss when being ported.

Balor's Nephew · Post by **Balor's Nephew** » Wed Mar 18, 2009 5:45 pm

You are probably right, for when I tried to use 3dfx in Myth II (you can use it with the wrapper, but, as Melekor mentioned, it's bugged), it looks very similar as in OpenGL.

It's the "Barrier"-Mission in the ported version:

http://i640.photobucket.com/albums/uu12 ... cus/10.jpg

btw you see it too, don't you?

Pyro · Post by **Pyro** » Wed Mar 18, 2009 5:57 pm

Yes I do, which is what made me think about the the possible change while it was being ported.

Balor's Nephew · Post by **Balor's Nephew** » Wed Mar 18, 2009 6:14 pm

I downloaded the content from tain.net, v2. I have same results on my second PC.

But now the next riddle: When I load the same savegame in Myth TFL 1.5 (the one in the Barrier mission) with OpenGL, the result is the same like in the ported version in Myth II (The resolution is 1152x864 because patch 1.5 forces game to use the desktop's resolution - is there a way to prevent that?):

Myth TFL 1.2 + wrapper (again):

http://i640.photobucket.com/albums/uu12 ... icus/5.jpg

Myth TFL 1.5 OpenGL:

http://i640.photobucket.com/albums/uu12 ... cus/11.jpg

So it's no effect of the portation, right?

Pyro · Post by **Pyro** » Wed Mar 18, 2009 6:20 pm

Well porting from TFL to M2 wouldn't affect TFL at all.

Post by **haravikk** » Wed Mar 18, 2009 6:33 pm

There does appear to be some definite "banding" in the Open GL picture, I think that's the correct term anyway
Does OpenGL have any kind of sampling radius when it performs dithering? If this is too small then it could account for that, while a larger one may result in smoother blending, otherwise it's likely that OpenGL and Direct3D simply handles dithering differently, possibly for performance reasons, and it ends up creating these visible "bands".

Balor's Nephew · Post by **Balor's Nephew** » Wed Mar 18, 2009 6:36 pm

So there are the same graphical losses in Myth TFL 1.5 and Myth II 1.6 by using OpenGL. But there are different results by using emulated 3dfx without patch 1.5 for Myth TFL.

If there were such a .exe, that prevents the annoying memory bug, it should be able to solve the riddle (it was called TFL_patch.exe two years ago, but it's not usable for configurations above 1 Gig of ram)...

Are these "bands" (thanks haravikk for the term

) normal or am I the only one who has them in the game?

Post by **Melekor** » Thu Mar 19, 2009 6:03 am

I made a loader that lets you bypass the memory bug.
download here: http://www.projectmagma.net/~melekor/tfllauncher/

Post by **Melekor** » Thu Mar 19, 2009 6:03 am

Earlier in this thread we already confirmed that emulated glide is currently superior to OpenGL (16 bit vs 15 bit).

Now I can confirm that emulated glide on Myth TFL (1.3; doesn't work on 1.5) is superior to emulated glide on Myth II.

Here is the comparison.

TFL 1.3 emu. glide:
http://www.projectmagma.net/~melekor/up ... ide_m1.jpg
Myth II emu. glide:
http://www.projectmagma.net/~melekor/up ... ide_m2.jpg

and for reference, the same scene in Oak (full 24 bit precision):
http://www.projectmagma.net/~melekor/up ... _24bit.jpg

I haven't yet figured out why this is. It's also interesting that m1 under glide looks practically identical to the full 24 bit result.

Post by **Myrd** » Thu Mar 19, 2009 10:44 am

That's a pretty significant difference.

Any reason not to make Myth II use all 24-bits by default?

Eddaweaver · Post by **Eddaweaver** » Thu Mar 19, 2009 12:37 pm

Myrd wrote:That's a pretty significant difference.

Any reason not to make Myth II use all 24-bits by default?

System requirements?

Balor's Nephew · Post by **Balor's Nephew** » Thu Mar 19, 2009 1:36 pm

@Melekor: If you were a woman, I would ask you if you would marry me

Since two years I'm waiting for such a patch! Thanks a lot!
I tried v1.2 and it worked. Strangely, in v1.3 or even 1.1 it could not recognize the disc (original) and without any patch, the game crashed when I tried to start a game with 3dfx.
But finally I'm able to play it under glide. Thanks again!
Probably you can use these improvements in your next patches.

Post by **Melekor** » Thu Mar 19, 2009 3:50 pm

After looking at the myth 1 source, I understand what is happening.

Glide under TFL renders the mesh using two passes: colormap and lightmap. The colormap pass uses hardware paletted textures, allowing the 8-bit colormap to be displayed exactly as it is. The lightmap is then multiplied on top. This was changed in Myth II to be done in a single pass, drawing color*light with a 16 bit texture, improving performance but sacrificing quality.

It isn't possible to do the exact same thing in OpenGL since modern hardware does not support paletted textures. Any way I can think of for getting full 24 bit in OpenGL is going to have a performance cost. It might be well worth it though, and I would like to try it some time.

@Balor's nephew:
Glad you found the loader useful. It was an interesting experiment, I've never played around with DLL injection before. TFL 1.3 worked for me, I wonder why it did not for you. I have the TFL iso mounted in a virtual drive, perhaps that makes a difference.

The Tain

New ATI/NVidia-Cards and dithering...