Non AF IQ Comparison

[Dave Baumann]

R3x0 Default Trilinear:

NV3x 'Application' Trilinear:

NV3x 'Balanced' Trilinear:

R3x0 Default Trilinear:

NV3x 'Application' Trilinear:

NV3x 'Balanced' Trilinear:

 

[Brent]

looking at it closely now i think GFFX's Application looks closer to ATI's default Trilinear in those shots

there are differences in 'balanced' compared to ati's default trilinear, but application does look like a closer match to me

 

[Luminescent]

Crap , I voted for balanced trilinear by mistake.

 

[RussSchultz]

In the ingame shots, just going by the color bands, the output is pretty equivalent in all the shots.

In the testor application, its obvious that the balanced one is slightly different as the bands between the mip maps are more abrubt.

 

[Hyp-X]

If you are asking the texture lod, it's an easy call: Application.

What I found strange that the FX shots have more texture detail, while having the same texture LOD.
Hmm.
I never thought trilinear could look that different between two cards.

 

[Nite_Hawk]

As far as I can tell, ATI's implementation and the nvidia application level trilinear look almost exactly the same. There seems to be very little if any difference. The balanced mode shows some pretty obvious differences though.

How much of a performance difference is there between application and balanced when anisotropic isn't enabled?

Nite_Hawk

 

[kyleb]

r300 default to nv3x look virtualy identical application in both game and test shots. balanced on the nv3x is definatly less, but still looks respectable.

 

[Dave Baumann]

SS:SE Citadel, 1600x1200 -

Balanced: 89.6FPS
Application 82.0 FPS

9% Diff (in that case).

 

[Marc]

Hi,

For HardWare.fr i use application setting for trilinear (balanced for aniso) for the 9800 Pro review and for the future FX 5600 / 5200 one.

For example on 5800 Ultra (balanced/application)

Q3 1.32 demo four 1600*1200 236,7 / 214,9 (10%)
SSE2 Valley Jaguar 1600*1200 75,4 / 68,5 (10%)
UT2003 Flyby 1600*1200 141,9 / 110,8 (28% !?!)
AquaMark 1.3 1600*1200 62,8 / 59 (6%)

Application setting for trilinear is also better to compare GF4 4x00 with GFX 5x00 in trilinear ....

 

[kyleb]

Marc, shouldn't your main concern on a high end product be the high end quality of it though? i sure don't buy a top-self videocard to play at anything less than the best settings, at least not untell it has some grey hairs on it.

 

[Marc]

kyleb> Yes and that's why i use the best (application) setting on NVIDIA card for trilinear, it give a similar result that ATI one or GF4 one

 

[kyleb]

but i am refering to aniso here were i would perfer to see aplication and not balanced, i haven't cared about plain triliniar sence i got my gefoce3 nearly two years ago.

 

[John Reynolds]

In the ingame shots, just going by the color bands, the output is pretty equivalent in all the shots.

Really? I think the Balanced shots bands shows very little color blending between each one. That definitely stood right out to me.

I voted Application.

 

[Bambers]

Application.

Balanced appears to be doing ~ half the screen (in total) with trilinear and half with bilinear.

 

[Brent]

but i am refering to aniso here were i would perfer to see aplication and not balanced, i haven't cared about plain triliniar sence i got my gefoce3 nearly two years ago.

the way i understand it right now balanced is better in AF

but i could be wrong, i need to look at it myself this weekend...

 

[Marc]

but i am refering to aniso here were i would perfer to see aplication and not balanced, i haven't cared about plain triliniar sence i got my gefoce3 nearly two years ago.

The problem is that ATI anisotropic is not perfect even with quality setting. It will be unfair to compare DX9 ATI & NVIDIA card with quality setting for ATI and application setting for NVIDIA since NVIDIA anisotropic is better with this setting ...

I think ATI quality vs NVIDIA balanced is much better for a fair comparison. But not perfect

 

[BRiT]

To play the flip side of the coin:

The problem is that Nvidia Balanced anisotropic far from perfect. It will be unfair to compare DX9 ATI & NVIDIA card with quality setting for ATI and balanced setting for NVIDIA since ATI anisotropic is better with this setting .

I think ATI quality vs NVIDIA application is much better for a fair comparison.

 

[kyleb]

but i am refering to aniso here were i would perfer to see aplication and not balanced, i haven't cared about plain triliniar sence i got my gefoce3 nearly two years ago.

The problem is that ATI anisotropic is not perfect even with quality setting. It will be unfair to compare DX9 ATI & NVIDIA card with quality setting for ATI and application setting for NVIDIA since NVIDIA anisotropic is better with this setting ...

I think ATI quality vs NVIDIA balanced is much better for a fair comparison. But not perfect

how is that a problem, i mean what do you base your idea of "perfect" on anyway? besides, i wasn't asking to compare anything, just asking to see how well the card performs when it is provideing the best image quality it can.

 

[John Reynolds]

To play the flip side of the coin:

The problem is that Nvidia Balanced anisotropic far from perfect. It will be unfair to compare DX9 ATI & NVIDIA card with quality setting for ATI and balanced setting for NVIDIA since ATI anisotropic is better with this setting .

I think ATI quality vs NVIDIA application is much better for a fair comparison.

Completely agree, particularly since Balanced isn't real Trilinear.

 

[Nite_Hawk]

The problem really stems from the fact that ATI and nvidia use different approaches to make aniso fast. Personally, I like ATI's approach because I don't notice the downsides of the implementation most of the time. (atleast with the R300 approach). I also like the fact that I know pretty much exactly what it's doing. I can have a certain level of aniso along with bilinear or trilinear filtering. The aniso implementation works well at 45 and 90 degree angles, and less so at 22.5.

So what are we left with? ATI's trilinear aniso implementation should look very similar to Nvidia's application level filtering most of the time. (so long as there are no angle close to 22.5 degrees). Still, Nvidia's application level filtering is probably going to be slightly better at the same aniso level(though it doesn't go up to 16x as does the ATI cards). Nvidia's balanced mode is going to look somewhat worse than ATI's trilinear most of the time, and I think the screenshots at anandtech and some of the other reviews show that. Certainly, the agressive modes look much worse than ATI's with either bilinear or trilinear. Over all, I'd say I'm happier with ATI's aniso implementation in the R300 than the implementation on the FX.

Nite_Hawk