E3 Doom3 screenies -- what AA and/or AF applied?

[3dcgi]

I thought that there was a problem with anti-aliasing the shadows with Doom3? I guess I am mistaken....

You're thinking of Parhelia's FAA.

 

[Archaeolept]

R300
6XAA
8 or 16XAF

at least from the jpeg screenshots, the Anti-aliasing seems close to perfect.

 

[Oratius]

My guess is:
R3X0, 6XAA, 8XAF.

 

[cho]

NV35 "hiden" AA and Ultrashadow HCT AA

 

[Squidlor]

Is there some type of error with the shadow being cast by the rear left leg of the upfront Trite? The shadow isn't connected to the tip of the leg like the other shadows...

I don't know why, but I'm going to say these are Radeon 3XX with 6xAA and 16xAF, even though I would expect most screen shots from id to originate from Nvidia cards...

 

[Joe DeFuria]

Is there some type of error with the shadow being cast by the rear left leg of the upfront Trite? The shadow isn't connected to the tip of the leg like the other shadows...

Not sure which leg you are referring to, but if a shadow isn't "connected" to it, that's just because the leg isn't touching the ground.

 

[Tagrineth]

Is there some type of error with the shadow being cast by the rear left leg of the upfront Trite? The shadow isn't connected to the tip of the leg like the other shadows...

I don't know why, but I'm going to say these are Radeon 3XX with 6xAA and 16xAF, even though I would expect most screen shots from id to originate from Nvidia cards...

Do you mean its left or our left?

Because if you're talking about ITS left...

The leg tip isn't connected to the shadow tip because, the leg curls back inward, and the light source in question is above the Trite. So the leg passes back into its own shadow.

 

[Laa-Yosh]

There's neither any AA or AF on those screenshots

They were simply rendered in some ungodly high resolution, 6000*4000 or something, then downscaled. I can try to dig up a link to the originals - the zipped JPEGs were some 11 megabytes....

 

[Hyp-X]

There's neither any AA or AF on those screenshots

They were simply rendered in some ungodly high resolution, 6000*4000 or something, then downscaled. I can try to dig up a link to the originals - the zipped JPEGs were some 11 megabytes....

Actually this qualifies as ordered grid SSAA.
For example 5120x3840 is 16x AA, and it provides 4x aniso for "free".

The thing was evident from the screenshots that it doesn't look like any of the R300s AA modes. (It has more than 4 samples per pixel, but it looks worse than 6x AA on the R300.)

 

[mczak]

There's neither any AA or AF on those screenshots

They were simply rendered in some ungodly high resolution, 6000*4000 or something, then downscaled. I can try to dig up a link to the originals - the zipped JPEGs were some 11 megabytes....

Actually this qualifies as ordered grid SSAA.
For example 5120x3840 is 16x AA, and it provides 4x aniso for "free".

Does downscaling always use just an ordered grid sampling pattern, averaging the pixels? I'd think there are better algorithms available.

 

[Hyp-X]

Does downscaling always use just an ordered grid sampling pattern, averaging the pixels? I'd think there are better algorithms available.

It always uses an ordered grid. (Do to the lack of choice, since the grid is already there.)

However there are better algorithm than just averaging (which is usually referred as block filter).
What you wan't is a high-pass filter with a carefully chosen filter kernel. Image manipulator programs like Photoshop tend to do quality downsampling (it's one of their job afterall.)

 

[micron]

"I just want that dam Cramack to finish Dom3 I wanna play now!" - young rage3d forum member

 

[Laa-Yosh]

Actually this qualifies as ordered grid SSAA.
For example 5120x3840 is 16x AA, and it provides 4x aniso for "free".

Of course you're right, but it's not FSAA in the sense as a card applying it... you get the point

 

[Tahir2]

Can you do the Doom III shots again in .png format? Perhaps in a .zip file to download.

The other shots don't seem to have any AA applied to them.

Regards.

 

[mczak]

Does downscaling always use just an ordered grid sampling pattern, averaging the pixels? I'd think there are better algorithms available.

It always uses an ordered grid. (Do to the lack of choice, since the grid is already there.)

Seems pretty obvious. Not sure why I got that strange idea you could use a different grid, stupid me.

 

[Calavaro]

One of those "guess the 'x'" where the author went AWOL, eh? :?