New D3D FSAA Viewer

[iodsnips]

So is 3xt the highest you can go, and 6xaa is the highest AA the R3xx can go, so when done right do we get an effective 18xaa?

If the rumor that the R420 will have 8xAA, would that mean when used with 3xt, we'd get an effective 24xaa?

My main question is not focusing on wither you think anybody needs more then 6xaa, but if the highest aa combined with 3xt would provide better aa?

 

[MuFu]

R420 doesn't support higher FSAA modes than R3x0.

 

[tEd]

R420 doesn't support higher FSAA modes than R3x0.

i would liked to have 8x native mode though

 

[SsP45]

R420 doesn't support higher FSAA modes than R3x0.

:? I was looking forward to 8x AA.
________
Ford Explorer Sport Trac

 

[bloodbob]

R420 doesn't support higher FSAA modes than R3x0.

OMG HOW SAD.



To all those people who are trying temporal AA set to 1

IT MEANS THEIR IS ONLY 1 PATTERN IE NORMAL AA think about it :/

Looks like temporal AA will be great for people who have very good crts they can effectively double their AA level but for all the people who have cheaper monitors or who have LCD screens its not gonna be much of a benefit.

Thanks for putting up those pics Pual.

 

[digitalwanderer]

What would a temporal value of 13 do? :|

 

[DemoCoder]

Not for us poor LCD users.

Well, it's looking good for this LCD user!

Haven't tried games yet, but FSAA Viewer, 1600x1200@60 Hz, 4X+2T looking solid! Haven't tried any other modes yet.

Well, I actually tried it with games, and it looked pretty crappy to me. For moving objects, it didn't do squat, and for stationary objects, it just made edge ghosting flicker for me that normally wouldn't. I'm already sensitized to looking for ghosting artifacts on LCDs.

For LCDs, a 3D version of "ClearType" would be much better.

 

[bloodbob]

What would a temporal value of 13 do? :|

Well if the driver had 13 sets of patterns for each aa mode then you would just a flickering mess :/

 

[dr3amz]

What would a temporal value of 13 do? :|

Well if the driver had 13 sets of patterns for each aa mode then you would just a flickering mess :/

lol yer

and to who said `omg how sad` at the standard ati aa modes... have you even looked at 6xAA? there's NOTHING to touch it in the home market for quality.

let alone Temperal 6xAA (12xAA effective? try it - the quality is stunning)

 

[Hanners]

What would a temporal value of 13 do? :|

Give you a seizure most likely.

 

[bloodbob]

and to who said `omg how sad` at the standard ati aa modes... have you even looked at 6xAA? there's NOTHING to touch it in the home market for quality.

I was hoping for

have you even looked at 6xAA? there's NOTHING to touch it in the home market for quality YET wait till the 5th of may.

 

[dr3amz]

and to who said `omg how sad` at the standard ati aa modes... have you even looked at 6xAA? there's NOTHING to touch it in the home market for quality.

I was hoping for

have you even looked at 6xAA? there's NOTHING to touch it in the home market for quality YET wait till the 5th of may.

lol - hopefully that will be true, i just got some mem i bought off a guy on futuremark forums (cheers Brad!) so now my mem is running at 400mhz not 333 (yes i will o'c it later but i gotta go to work now 8) )

new bench - http://service.futuremark.com/compare?2k1=7790000

and i added the extra TEMPAA commands - i set the fps point to 0 so that its always working - as tempAA gives NO fps drop - and this morning, the 2xAA that you can normally see on the roof of the hut in nature test (as its drawing the AA in) is basically gone... now im seeing at least 4xAA at the performance of 2x - with no flickering etc.

if it wasnt for the vsync limiting - i'm sure i'd easily break 13k - with effective 4xAA

STUNNING!

 

[Bambers]

Ah...

I think you actually answered your (and my) own question.

The patterns they selected are probably based on a comprimise between getting the best quality on moving edges, and the best quality on non-moving edges....

Currently all the patterns do result in the final one being sparse (eg the 2 patterns in 2x2t make a standard 4xRGAA pattern) with a couple of exceptions.

4x2t is sparseish )). The grid on the pixels is 12x12 and so the 8 samples aren't spread out completely evenly.


6x3t is not spase again due to the 12x12 grid. It also means it's really got no benifit over 6x2t or 4x3t since the extra 6 subsamples are just sharing positions, vertically and horizontally with other samples.

 

[Kombatant]

Anyone thought that this new AA technique is not finalized yet in the build you're testing? Hence it's not enabled yet?

 

[Colourless]

Note that the sample positions in that ppt presentation slide don't match exactly with what you actually get. Things are a little different (slightly better).

 

[dr3amz]

Anyone thought that this new AA technique is not finalized yet in the build you're testing? Hence it's not enabled yet?

enabled in what? you can clearly see its working (visually) within farcry/halflife/anything d3d/fsaaviewer etc.

i dont understand what you mean? :?

 

[Kombatant]

Anyone thought that this new AA technique is not finalized yet in the build you're testing? Hence it's not enabled yet?

enabled in what? you can clearly see its working (visually) within farcry/halflife/anything d3d/fsaaviewer etc.

i dont understand what you mean? :?

I mean exposed in the drivers. Just because something is in working state within the drivers doesn't mean it's finalized. I don't think this AA technique was advertised by ATI as "complete and ready for public use". So there is a chance that it's not finalised yet. That's all I am saying.

 

[Ante P]

When I try it it seems to use a different pattern every frame. ie not just switching between two patterns rather just producing random patterns all the time???

 

[dr3amz]

Anyone thought that this new AA technique is not finalized yet in the build you're testing? Hence it's not enabled yet?

enabled in what? you can clearly see its working (visually) within farcry/halflife/anything d3d/fsaaviewer etc.

i dont understand what you mean? :?

I mean exposed in the drivers. Just because something is in working state within the drivers doesn't mean it's finalized. I don't think this AA technique was advertised by ATI as "complete and ready for public use". So there is a chance that it's not finalised yet. That's all I am saying.

ah got ya - well from what i can tell it doesnt work on anything less than cat4.4s (maybe 4.2s but definately not before)

so it could be something they are tweaking still, and it may just be in there for test purposes. i'd imagine the final version would have sliders etc in the settings for it and possibly also work for opengl

When I try it it seems to use a different pattern every frame. ie not just switching between two patterns rather just producing random patterns all the time???

if you run the FSAAViewer you will see that it clearly switches between 2xFSAA samples/patterns - usually with 2x aa you would just get the one, but with tempaa you get both (effective 4x) through fast switching of the sample/patterns.

But I think this is probably the `surprise` ATI had for us existing users - what else could there be?

free antialiasing (well to a degree) sure is sweet!

 

[MistaPi]

Sorry to have to ask, but whats the big benefit if it only work on nonmoving images? Have I missed something?