'Balanced', 'Application' and GeForce 3/4

[Dark]

Hi,

Since almost all reviews concentrate on the Geforce FX's IQ and Nvidia's introduction of 'Balanced' and 'Performance' modes for their new line of cards I'm wondering how Nvidia's new 'settings' affect their older GeForce 3/4 line of cards. Owning a Geforce 3 myself I noticed quite a large performance drop when choosing 'Application' over 'Balanced' yet I am not sure if image quality is actually improved.

Also using Xmas' AF Tester I am not able to tell a difference between the three modes (maybe I'm just doing something wrong though - don't know).

So back to the question, What do the different settings do? Do they actually change anything (I assume they do - see performance hit)? What do they change?

Thanks for any responses,

Dark

 

[demalion]

Here is where I learned about it.

And here is some further discussion.

And here is some discussion of the impact when Aniso is off but the slider is selected to the different positions.

Here is a thread I started to find out what changed with 43.51 drivers. I'm not sure of the answer yet.

 

[Dark]

While I thank you for the links they do not really answer my questions. Like most recent reviews they concentrate solely on the GeForce FX (unless I am stupid and can't read. In this case would you care to direct me to the relevant posts - or better just tell me to look again).

As I mentioned in my post above I do notice a rather large performance hit with my Geforce 3 running with 'Application' mode yet I am unable to get a visual difference with Xmas' AF Tester. That's my reason for asking.

Maybe someone else with a Geforce 3/4 would care to share his/her experience?

Dark

 

[Xmas]

As I mentioned in my post above I do notice a rather large performance hit with my Geforce 3 running with 'Application' mode yet I am unable to get a visual difference with Xmas' AF Tester. That's my reason for asking.

My app only uses the first texture stage which is usually applying the base texture. Here AF makes the biggest difference.
Detail maps also look better with AF, but you rarely see a difference with lightmaps, and using AF for normal (bump) maps isn't a terribly good idea.
So my guess is that NVidia uses reduced AF for all but the first texture, or they use reduced AF depending on the texture format.

 

[Arun]

So my guess is that NVidia uses reduced AF for all but the first texture, or they use reduced AF depending on the texture format.

Hmm, interesting info... I'd bet on the "using reduced AF for all but the first texture" - it's been in nV drivers for ages.

Nobody trusted me when I said that was what nVidia was gonna try to do that to make the NV28 competitive against the Radeon 9500... Eh, so they did... Too bad it ain't sufficent Ah well, still nice to know they gave such a performance hit for GF4 users so long after its release!


Uttar

 

[demalion]

My apologies, discussion for the non FX were completely absent in some of the links, and the phrasing of the last part of your question led me to forget your mention of the GF 3 before that.
Until more answers show up in this thread (if you haven't gotten your answer already):

There was some discussion of your question as far as non FX cards, but it was pretty cursory. For example, Wavey did reference this for some (unsubstantiated) mentions that addresses that, and the 43.51 dsicussion seems to indicate investigation results for both the GF FX and earlier cards might be forthcoming.

I think the initial discussion link referenced GF 4/3 behavior, and some guesse that the hidden settings for RivaTuner that exposed similar behavior for the GF 4 and GF 3 before the nv30 launch, so perhaps a search with RivaTuner here and some nvidia centric sites ( www.nvnews.net might have had discussions, though I forget the date before which their forum data was lost) might help you get information from that angle.

 

[Arun]

Found the nV News post you were talking about, demalion...

http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=7245


Uttar

 

[Dark]

Thanks. That thread from nvnews seems to cover exactly what I was wondering about.

Dark

 

[Pete]

Ante P's recent 5800U review had a GF4 AF screenshot, IIRC. The GF4's AF was much better than the 5800U's.

 

[Ailuros]

Hmm, interesting info... I'd bet on the "using reduced AF for all but the first texture" - it's been in nV drivers for ages.

No they're not. At least not up until 43.45 I've tried.

The 40xx series of drivers introduced the application/balanced/aggressive slider, with the only other difference compared to former driver sets that if NOT set to "application" in openGL lightmaps wouldn't get filtered (comparable to Rivatuner Quality=Application and Performance=Balanced/Aggressive for openGL).

However to get per texturing stage optimised AF in D3D, I needed to use Unwinder's AnisoBooster D3D patch within Rivatuner.

I also haven't noticed any bi-/trilinear flipflopping nor TC compression artifacts when using aggressive AF, they all have the same quality on NV25 and performance (up until 43.45 as I noted above).

***edit: I just reran a couple of botmatches in UT2003 and application is ~12% slower than aggressive or balanced, yet still no differences visible.

 

[Althornin]

one odd thing for me:
If i play some games (Serious Sam mainly) on my GF3 with it set to "Balanced" or "agressive" i get crashes. I have to set it to application for stability.

 

[chavvdarrr]

very detailed view on GF4's aniso in Det 4X.xx can be found on Unwinder's site - www.nvworld.ru

 

[Ailuros]

one odd thing for me:
If i play some games (Serious Sam mainly) on my GF3 with it set to "Balanced" or "agressive" i get crashes. I have to set it to application for stability.

Not that it's related, but have you forced SS or SS:SE to use a 24bit Z buffer?

If I don't use the following, I get horrendous Z-aliasing in SS:SE:

gap_iDepthBits=(INDEX)24;
gap_iStencilBits=(INDEX)8;

Those crashes with anything other than balanced or aggressive are strange.

 

[Ailuros]

very detailed view on GF4's aniso in Det 4X.xx can be found on Unwinder's site - www.nvworld.ru

I couldn't find anything in english. Do you have a direct link to an english article, if it exists in that form?

 

[Althornin]

Not that it's related, but have you forced SS or SS:SE to use a 24bit Z buffer?

If I don't use the following, I get horrendous Z-aliasing in SS:SE:

gap_iDepthBits=(INDEX)24;
gap_iStencilBits=(INDEX)8;

Those crashes with anything other than balanced or aggressive are strange.

yes, i too had the zbuffer issues.
And it crashes with anything other than application settings, sorry if that wasnt clear. It is strange. But very replicateable.