Tom's GF4 anisotropic scores in his Parhelia review

Noko,

Of course mipmaps change with higher aniso degrees. I just can't reproduce the difference in LOD adjustments for 8 aniso between 28.32 and 29.42 drivers as Dave showed on absolute default game and driver settings.

By the way I always liked those disco coloured screenshots :LOL:

MikeC,

Rev stated he used default driver settings/no 3rd party utility. Assuming 29xx drivers default to the "performance optimisations" what would the difference then be between quality and performance mode? If between those two the performance drop would rather small I wouldn't even care. As said it looks to me like quality mode is filtering lightmaps too in q3a. If that should be the case then there's really not much use to use quality mode.

I simply can't get rid of the suspicion that high degree aniso in quality mode filters textures that are hard to distinct in normal gameplay (I mean the difference) and while skipping those (using performance mode) there's really not much to loose.

Mipmaps can be adjusted in both D3D and openGL and they become handy for me only with Quincunx and upwards and then even if rarely over -0.8 to -1.0.
 
Of course mipmaps change with higher aniso degrees. I just can't reproduce the difference in LOD adjustments for 8 aniso between 28.32 and 29.42 drivers as Dave showed on absolute default game and driver settings.
Click to expand...

Oh, between drivers was the point :oops:. I didn't know that was the case with the GF3/GF4 or I never really noticed. Which makes alot of sense for that to happen since texture filtering is increasing. I knew on the Radeon that happens but never really tested it out on my GF3. I guess I learn something everyday, especially here. Thanks

Oh yea playing the game in those pyscho colors does definitely give you more then you bargain for - - a migraine :-? :eek:
 
Of course I'm not doubting Dave's results at all. If I use the extreme addon in SS:SE on 29:42 the texture quality slider gets pushed to sharper from normal. Bugger is I don't recall if the same happens with 28:32. Hence my comment that SS controls apparently exclusively aniso settings.

I had a closer look on extreme settings at SS, when I saw Reverend's shots.
 
One thing that I have to say is that we all know that SS will often automatically determine quality settings without informing the user, meaning that it could very well have detected something different between the driver sets, and switched it itself. I think other games with more reliable settings should be used for comparisons...I think Rev's shots show such a case.
 
One thing that I have to say is that we all know that SS will often automatically determine quality settings without informing the user, meaning that it could very well have detected something different between the driver sets, and switched it itself.
Click to expand...

Not if you override it or use one of the Addon settings.

What you are seeing from the SS shots are what NVIDIA have exposed the settings as in through the OpenGL extensions - probably safer to use and know what going on than driver settings (if you have tweakers and god knows what else) as these are unambiguous as to whats occuring.

If (and I don't know if they have) they have changed the way the texture slider interacts with the LOD settings between drivers is because nVIDIA have enabled/changed this functionality in how OpenGL interacts with the hardware.
 
***SUBJECTIVENESS ALERT***

my thoughts on the screen shots in the article.

2X Ansio
http://www.beyond3d.com/reverend/gf4aniso/2751/moh2xaniso.jpg
to
http://www.beyond3d.com/reverend/gf4aniso/2942/moh2xaniso.jpg

look at the hall just beyond the soldiers bodies, there is more "bluriness" (for lack of better word) on the 2942 drivers. The edge of the hall on the right side near the far door is also less crisp than on the 2751 drivers. Its like it is only 1.8 ansio versus 2X, not noticable unless you switch back and forth continuously.
Also, look at the right side of the door, you can see a black stripe in the 2751 drivers, not there in the 2942. And the black bars on the bottom of the door are a good reference, along with the handles, you can see the difference.

4X Ansio
look at the top right bar in the far end of the hallway, it looks like the 2942 drivers aren't smoothing it out quite as much as the 2751 do. The door isnt as crisp either, like it is 3.8X ansio (like previous example)

8X Ansio
Same as with 4X Ansio.

What it looks like is happening, is that they have chosen more of a fine line as far as distance goes with Ansio versus the previous driver set. It is barely, barely discernable from a gaming point of view, but there is a difference.

Beofre anyone rants on this post, it is a subjective post, image interpretation is always subjective, (so dont get your panites all bunched up and soiled)
 
Now that I looked at the 2x shots (I didn't before...I saw no reason to...I only compared the 8x shots), I do see very slight differences. What I did see was a few black 'specks' in the 27.51 shot that weren't there in the 29.42 shot. Whether this is more or less correct rendering, I do not know. I couldn't see any difference in texture clarity.
 
Wavey (or anyone for the matter),

WRT SS:SE. I have never been able to either adjust the in-game texture aniso quality (it always stays at 8 ) nor does forcing various aniso level via the OpenGL display properties work on a GF4 Ti. This is regardless of the driver sets I use.

On Wavey's page, he stated using 8xAniso, which I'll assume he definitely sees a difference by either forcing it through the OpenGL properties page or through the in-game texture anisotropy adjuster.

Can anyone confirm this or is something wrong with my system?
 
No Reverend I've been able to set any level of aniso from 27.50 to the current 29.42 inside the game.

But you're correct though that driver openGL aniso settings get overridden by the game. You may set in the drivers 8 aniso and pick 2 in the game or any other settings the game will apply the level of aniso it's been set to from it's own settings.

By the way I saw you and Wavey stating that you don't use 3rd party tweaking utilities. Are you guys enabling aniso for D3D via the registry?
 
No, it is the same if using D3D in SS:SE - neither options (using RivaTuner this time since there is no aniso option in NV's D3D Properties, nor in-game texture anisotropy) work.

For OpenGL, I remember that SSam (the original) will allow its in-game texture anisotropy adjuster, as well as forcing it via NV's OpenGL properties, to work as long as the ICD reveals the GL_EXT_texture_filter_anisotropic extension. I know that the latest 29.42 reveals this extension but I have no idea why neither works in SS:SE. Forcing various aniso levels works in the other two OpenGL games I have, MOHAA and JK2, both Q3-engines.

While it may appear to me that this could be specific to only SS:SE, I'm curious as to Wavey being able to "force" (whether it is thru the in-game related option or thru NV's OpenGL properties) 8xAniso in SS:SE as evidenced by that page of his. As it is, I see absolutely no difference in quality in SS:SE either thru the game's in-game texture aniso adjuster nor thru NV's OpenGL properties aniso filtering setting option using the 29.42s - neither appears to work at all... in my system at least. So, I'm curious if my system is screwed or some folks have made a mistake.
 
Ailuros said:
No Reverend I've been able to set any level of aniso from 27.50 to the current 29.42 inside the game.
Click to expand...
Are you sure about this? Because I can actually set Texture Anisotropy all the way to 128 but as soon as I leave the Advanced Rendering menu and then get back to it, it reverts to 8. Try that and let me know.

By the way I saw you and Wavey stating that you don't use 3rd party tweaking utilities. Are you guys enabling aniso for D3D via the registry?
Click to expand...
I do use RivaTuner, not sure about Wavey.
 
Just tried it again and 128, 64, 32, 16 flip back to 8 as they should, while all other settings lower than 8 stay as set, wether I exit/enter the advanced menu, the whole game or a simple map.

Why don't you try killing your persistent symbols.ini?
 
You see what I mean, Ailuros? It's easy for folks to get things wrong if they don't really investigate things!

My guess right now is that SS:SE, being an updated engine compared to the original game, now determines/sets the Texture Anisotropy according to degree of anisotropy a hardware supports, not number of available point samples a hardware can perform. This wasn't the case with the original game. SS:SE, IMO, will only ever allow the maximum aniso degree. I'll check with Croteam for verification later.

As for PersistentSymbols.ini, if you apply the Extreme GFX add-on everytime, this file is made redundant.
 
Both Direct3D and OpenGL only perform anisotropic filtering on a basis of degree. The actual hardware implementation of how to render that degree of anisotropic is entirely up to the vendor.
 
Chalnoth, I'm talking about a game's implementation and use of descriptive words of the specific implementation. Folks get deceived if they don't know what Croteam, for example, mean when their game use the words "Texture Anisotropy" in their in-game menu.
 
I did notice the higher settings in the past but I never really bothered with them, since it was obvious to me that 8x sets the maximum aniso level on the card.

I'm a bit lost of course with your last post: so you mean the numbers could be that they were actually meant to read as 16 tap 32 tap up to 128tap, yet turn out to work only as aniso degree levels? Forgive my slowliness today but I'm a bit overworked.

I assume 128tap was then meant to address ATI's highest aniso level (according to their implementation).
 
Given the performance (And image quality), I highly doubt that the Radeon 8500 actually uses 128-tap anisotropic in any situation.

I would be willing to bet that ATI uses a maximum of 2x16 (32-tap) for 16-degree anisotropic, or possibly as low as 1x16 (16-tap). This would explain some of the screenshots where textures show some problems off-center. But, I don't really know for certain. It just makes sense given the performance and image quality.

By contrast, it makes sense to me that nVidia's anisotropic increases the number of taps at a given degree depending on the angle and orientation of a surface (an absolutely optimal method would, instead of increasing the number of samples, just adjust the sample pattern...I don't think nVidia's hardware is capable of non-grid sample patterns). I believe the performance hit would be significantly higher if, for example, 8-degree aniso always used 64 taps.
 
Reverend said:
WRT SS:SE. I have never been able to either adjust the in-game texture aniso quality (it always stays at 8 ) nor does forcing various aniso level via the OpenGL display properties work on a GF4 Ti. This is regardless of the driver sets I use.

On Wavey's page, he stated using 8xAniso, which I'll assume he definitely sees a difference by either forcing it through the OpenGL properties page or through the in-game texture anisotropy adjuster.
Click to expand...

Well, 8x is what gets set by the extreme addon for GF3/4 as the extreme addon will set th max level available to the card. However, there does appear to be something wrong there as you should just be able to alter this in the advanced rendering options.
 
You must log in or register to reply here.
Back
Top