GFFX Anti-Aliasing with FSAA Tester

Randell said:
Of course all of our comparisons have been based on edge quality only and not the other benefits SS AA can bring when looking at 4xS/6xS/8xS.

Brent have you played any games with excessive LOD, alpha's etc where SSAA can help aliaising better than MSAA, e.g. F1 2001 or Counterstrike. Operation Flashpoint, even Black & White?

i haven't looked at that yet, i will though, its on the list :D
 
Dave H said:
Not to be pedantic, but it's Javascript, not Java. Completely and utterly totally different 100% unrelated things.

(Ok, was perhaps a bit pedantic...) :)

noted ;)
 
Between 80 - 100 and 135 - 155 degree angles even ATI x2AA does better than NV30 8xs mode. ATI x4AA is better than any NV mode.
That explains the HQ UT2003 image from HardOCP where the near horizontal pole in the right side of the image looked so jagged.
 
Speaking of which, it doesn't seem that 8xS should ever look like 2x FSAA, at any angle. I think that mode is also not being properly-captured. The minimum level of gradients at near horizontal/near vertical should be 4. It should look the worst near 135 degrees (0 degrees pointing directly to the right), with a minimum of approximately 3 gradients.
 
I suppose the only way to confirm that it's being captured properly is ask Brent to confirm if the jaggies on the near vertical pole from the UT2003 image are still there when ingame.
 
One thing you have to keep in mind is that Fablemark probably won't use the two-sided stencil acceleration available in the Radeon 9700 and GeForce FX.
 
I've not noticed texture aliasing in CS myself :?

The 8x and 6x modes do seem a bit odd. It was mentioned about in the ut2003 and nfshp2 pictures too.

I think its capturing properly though and I agree with colourless in that its using 4xOGMS and a relevant horizontal SS instead.

6xSx6.png


This is from the 6xS shot and it clearly shows all the shades and the arrangement of them is just as would be expected of an ordered grid.
 
Thanks Brent and Thomas... great stuff here!

Since I think that 4xFSAA should be the sweet spot between image quality and performance I have to say I'm sorely disappointed with the NV30. It's simply not good enough at the horizontal lines. 3dfx tech my a**...
 
To my old, aged eyes, the ATI 4X looks as good or better than any of the GFFX's AA solutions. This is really disappointing, to say the least. The more I see & read about the GFFX, the less I even want to try one.....
 
Nagorak said:
Why is Nvidia so afraid to go with RGMS?

they probably wanted to concentrate on other "important stuff" thus they just left AA they way it's been since GF3 and just added some new modes through their drivers instead of rethinking the hardware implementation
 
I agree with demalion. Those should be actually called 6x and 8x modes, the added "S" is misleading if it's there in the driver panel. The 6xS/8xS modes I can unlock here on the NV25 with the latest Rivatuner have little in common neither in terms of edge or texture quality compared to what I see on the FX.

They probably haven't exposed those S modes yet and the driver is just picking 6x and 8x when anything with an S is selected.

One thing you have to keep in mind is that Fablemark probably won't use the two-sided stencil acceleration available in the Radeon 9700 and GeForce FX.

Would it even make a difference if the application would?

Kristof himself has stated that FableMark is fillrate limited rather than vertex throughput limited.
 
ahh. :)

Maybe the FX has driver (or worse hardware) bugs stopping it from using the 2xRGMS in combination. With the problems in the 4xS modes stopping them from using the 6xS and 8xS modes of the gf4, maybe nvidia thought that they had to have the 8x and 6x mode working in some form for reviews so used the 4xOSMS versions instead. :?
 
Maybe it's just me, but I would prefer the 9700's 2X to any of the GFX modes. There is that one radial line that just disappears on the GFX.
 
Back
Top