Silent_Buddha
Legend
Recently I had a chance to try out the 8.38 release 7 drivers on my HD 2900 XT. And I've noticed that in games where traditional MSAA isn't supported the wide and narrow tent is still applied if AA is forced through the drivers.
Thus in a game like, for example, Vanguard, I still get a very minimal amount of AA, although in this partcular title Wide Tent with no MSAA isn't very useful other than for some of the alpha textures. IE - Wide tent by itself without MSAA appears to my eye to be somewhat less than 2xMSAA for edges.
The question I have is.
Since the shader/ALU part of this AA algorhythm is being used, is it possible to come up with an efficient way for a completely shader based AA algorhythm to provide AA in titles that don't support tradition MSAA?
I certainly don't mind the performance hit if it allows me to use AA in games that I'm currently avoiding due to no AA.
To get away from the general question for a moment and to something a little more specific.
Considering the R600 has an excess of theoretical ALU power and a "fast path" from the RBEs to the ALU arrays, is this something that could reduce the performance hit of doing AA completely in shaders through the drivers?
Regards,
SB
Thus in a game like, for example, Vanguard, I still get a very minimal amount of AA, although in this partcular title Wide Tent with no MSAA isn't very useful other than for some of the alpha textures. IE - Wide tent by itself without MSAA appears to my eye to be somewhat less than 2xMSAA for edges.
The question I have is.
Since the shader/ALU part of this AA algorhythm is being used, is it possible to come up with an efficient way for a completely shader based AA algorhythm to provide AA in titles that don't support tradition MSAA?
I certainly don't mind the performance hit if it allows me to use AA in games that I'm currently avoiding due to no AA.
To get away from the general question for a moment and to something a little more specific.
Considering the R600 has an excess of theoretical ALU power and a "fast path" from the RBEs to the ALU arrays, is this something that could reduce the performance hit of doing AA completely in shaders through the drivers?
Regards,
SB