Could anyone with a GFFX please...

[demalion]

As I said above, I can't reproduce your results, Bambers. Can anyone else with an 8500 do so to see if I'm doing something wrong?

I can get results that are somewhat similar to your first shot when using point sampling for the mip map filter, though.

Do you have any registry settings tweaked, Bambers? Is your control panel properly installed? What driver version? (sorry if you said).

 

[Bambers]

I've swapped the pictures with ones that show the effect up a little better, still quite hard to show in still pictures though. (and yes I had noticed my taskbar is showing throw, its because I run cs in a window (1024x768 at desktop of 1152x864) and the taskbar overlaps a little, opengl screenshot doesn't pick it up, d3d does)

3.3 and I have no tweak programs installed, control panel does seem to be working fine, I've seen this on previous drivers however.

I was using the defualt settings for both programs anisotropy set to 1x, minification: anisotropic, magnification: anisotropic, mipmap: linear for the d3d app and minification: linear_mipmap_linear, magnification: linear for the opengl one.

You're correct that there is a difference, but your conclusion is incorrect.

Bah, you could at least correct me then. :|

 

[Arun]

You're correct that there is a difference, but your conclusion is incorrect.

It's a LOD difference, then?
No matter what, this seems awfully similar to what nVidia is doing with their balanced mode. Maybe you aren't doing the same thing, but the effect is still rather equivalent.

The question, though, is knowing whether the drivers are not doing the same thing when doing colored mipmaps. If they aren't, then I'd definitively call that cheating.

And please nobody say "Yes, but then look at nVidia" - I *am* of the opinion they're cheating by using FX12 in 3DMark 2003. I *am* of the opinion their Aggressive AF is crap. So please, let's try to be objective


Uttar

 

[Dio]

One point to note: there are trilinear filtering techniques that work correctly if the mipmaps are 'closely related' that do not work if they are not.

 

[Dave Baumann]

Well, easy way of telling if they are changing things with mip colouring on is to bench a Quake title or SS:SE with coloured mips on and off. I don't have an R200 based board in a machine at the moment though.

 

[demalion]

Can anyone else reproduce this? I can only get results close to that (same effect, less "stages") by specifying point sampling for mip map selection. To me, this makes it seem likely to be a control panel conflict or bug.

So, my new question is, where did you get the 3.3 drivers?

However, if you're sure you have the right control panel version installed, and you say this happened with previous drivers (which seems to rule out that it is something new to your 3.3 installation), perhaps someone else with your OS can double check (I'm using ME) as maybe the problem is OS specific?

In the absence of confirmation from someone else, another thing to check is whether you have a 0000/0001 type conflict from a bad install.

 

[Bambers]

3.2, sorry that was a typo.

 

[snk]

Well, I am able to reproduce bambers' results. This is with Radeon 8500 and Cat 3.2's:

Direct3D with texture preference at quality:

And this time with high quality:

And for OpenGL:

 

[Colourless]

Um, I have an explination for the Radeon results here....

You are using a large Positive LOD bias correct? If so, then what you are seeing 'might' be Bilinear Magnification.

I am about to run the tests myself to find out for sure.

 

[Colourless]

Yes, just checked, and it is because of the Bilinear Magnification. However, it would appear there is a 'bug' in the program. The Magnification and Minification setting are swapped. Go figure.

Set the Minification setting to 'Point' and the image will no longer be blurred with a Large Positive LOD Bias.

 

[Bambers]

I would say it isn't as its (if you actually look for it) noticable in games at normal lods, I used the high lod just to make it show more. The railyard area of RTCW shows it up fairly well, I don't have that installed currently though.

Well, I am able to reproduce bambers' results. This is with Radeon 8500 and Cat 3.2's:

Direct3D with texture preference at quality:

Glad someone else can, I was starting to wonder if I'd installed the drivers wrong or something.

 

[Bambers]

I did some benchmarks with q3, four.dm_68 timedemo, vsync off, max details, 1280x1024x32 and in drivers, mipmap pref: high quality, AA and AF off.

Trilinear, texture pref: quality

coloured mipmaps off: 135.7fps
coloured mipmaps on: 126.5fps

Trilinear, texture pref: high quality

coloured mipmaps off 96.5fps
coloured mipmaps on: 96.1fps

Bilinear, texture pref: quality

coloured mipmaps off 138.2fps
coloured mipmaps on 137.9fps

Bilinear, texture pref: high quality

coloured mipmaps off 128.7fps
coloured mipmaps on 128.3fps

There is a bigger difference between coloured mipmaps on and off with the quality setting and trilinear.

 

[Dave Baumann]

Mmmmm.

Would have been more conclusive if they were in line with high quality though.

OGL Guy?

 

[Colourless]

Mmmmm.

Would have been more conclusive if they were in line with high quality though.

OGL Guy?

Dave, read my previous 2 posts. There is nothing fancy going on here. You are seeing the result of bilinear magnification.

If the 'plane' mode is used, it's pretty obvious what is going on.

Or if you want to use the tunnel mode, set the texture to -6.

Then change the Minification Filter (which in this app really changes the magnification filter)

 

[Xmas]

Then change the Minification Filter (which in this app really changes the magnification filter)

Are you talking about the OpenGL or D3D app now?

 

[Colourless]

Direct3D