I think we can expect both with WGF2.0 HW as compared to today and even perhaps R520 or NV's next part given some factors.
If R520 is still in the lineage of the R300 then I feel safe in what I will safe although I admit I could be wrong.
I think unified shaders will be quite significant if the concept of, "load balancing," is truly valuable in real word terms along with unified shaders allowing for new features to be implemented.
I digress for a moment...(digress = back up I hope
)
With the advent of HW that is capable of outputting resolutions higher than 2048 x 1536 ,as I think would be the case in 2006 when the R600 and NV's competing part arrives, I think the value of some things will diminish and/or come to a level of IQ where no further improvement is needed or justifiable. What comes to mind is AA and AF.
At such resolutions I cannot imagine much more emphasis would be placed on AA when let's say temporal AA of 8x would be perceivably equivalent to 16x or 24x. I don't see it as unfair to think an R600 could keep the frame rate high enough at 8x MSAA as to turn TAA into the, "end of the line," so to speak and TAA would be a great saving to performance. I know some would still prefer SSAA and no doubt it would be an option, but ATI did say they are working on method to filter alpha-tested stuff without using super sampling. (that's in my memory of course, but perhaps time and drugs have...well got a few wires frayed here and there).
I would also argue that at that point in time 16x AF no doubt "forced" so to speak or 32x AF still using optimizations and selective filtering would be at or approaching all that is really needed in this area.
Keep in mind I'm talking about for games not for Dreamworks or anything nor am I implying a halt to things in these areas. I'm saying the emphasis should be greatly diminished IMO.
I won't bore people here smarter and more in the know than me with talking about "load balancing" in relation to having unified shading at the HW level. (How ironic is the spill about AA and AF then? Yeah I know ) To me this presents the potential for a very nice overall performance boost.
I feel that the emphasis then would be on finding new features to grab the attention of enthusiasts or would be bringing old promises into the realm of reality. Global illumination and things of this sort so to speak. (if possible of course) I am at a loss to even propose what these things could be really. It could be to just use old things in new ways as raw power begs for usage....perhaps a per-pixel, real time environment/x,y,x/(new kind of map) mapped water beast boasting sub-surface light scattering, reflection, refraction, differing opacities, insane particle effects as it sloshes along after emerging from a seemingly harmless pool of water via true displacement mapping and who knows what else! I don't know what I just said but it sure sounded kewl and looked likewise the same in my mind!
Rad man....dare I say tOtaLlAy RaDiCaL Dude! ....ahhh it's been years hasn't it...
Ahem.
I think that aside from new features devs would want greater ease of use with them over more raw performance from IHVs. There going to be allot of new possibilities to explore and nothing would seem to suck worse than not being able to do so not because things are out reach but it's just too darn annoying hard to justify making the effort. (...but then coding anything is too hard right? )
Anyway I think it'll be a combination of getting new things to play with and talk about and getting better performance....err because some of the new features look like like they should really improve performance and allow for other features to consume the surplus they provide.
Did that make sense...did I even make a point?
Well...I tried anyway at least.
If R520 is still in the lineage of the R300 then I feel safe in what I will safe although I admit I could be wrong.
I think unified shaders will be quite significant if the concept of, "load balancing," is truly valuable in real word terms along with unified shaders allowing for new features to be implemented.
I digress for a moment...(digress = back up I hope
)With the advent of HW that is capable of outputting resolutions higher than 2048 x 1536 ,as I think would be the case in 2006 when the R600 and NV's competing part arrives, I think the value of some things will diminish and/or come to a level of IQ where no further improvement is needed or justifiable. What comes to mind is AA and AF.
At such resolutions I cannot imagine much more emphasis would be placed on AA when let's say temporal AA of 8x would be perceivably equivalent to 16x or 24x. I don't see it as unfair to think an R600 could keep the frame rate high enough at 8x MSAA as to turn TAA into the, "end of the line," so to speak and TAA would be a great saving to performance. I know some would still prefer SSAA and no doubt it would be an option, but ATI did say they are working on method to filter alpha-tested stuff without using super sampling. (that's in my memory of course, but perhaps time and drugs have...well got a few wires frayed here and there).
I would also argue that at that point in time 16x AF no doubt "forced" so to speak or 32x AF still using optimizations and selective filtering would be at or approaching all that is really needed in this area.
Keep in mind I'm talking about for games not for Dreamworks or anything nor am I implying a halt to things in these areas. I'm saying the emphasis should be greatly diminished IMO.
I won't bore people here smarter and more in the know than me with talking about "load balancing" in relation to having unified shading at the HW level. (How ironic is the spill about AA and AF then? Yeah I know
) To me this presents the potential for a very nice overall performance boost.I feel that the emphasis then would be on finding new features to grab the attention of enthusiasts or would be bringing old promises into the realm of reality. Global illumination and things of this sort so to speak. (if possible of course) I am at a loss to even propose what these things could be really. It could be to just use old things in new ways as raw power begs for usage....perhaps a per-pixel, real time environment/x,y,x/(new kind of map) mapped water beast boasting sub-surface light scattering, reflection, refraction, differing opacities, insane particle effects as it sloshes along after emerging from a seemingly harmless pool of water via true displacement mapping and who knows what else! I don't know what I just said but it sure sounded kewl and looked likewise the same in my mind!
Rad man....dare I say tOtaLlAy RaDiCaL Dude! ....ahhh it's been years hasn't it...Ahem.
I think that aside from new features devs would want greater ease of use with them over more raw performance from IHVs. There going to be allot of new possibilities to explore and nothing would seem to suck worse than not being able to do so not because things are out reach but it's just too darn annoying hard to justify making the effort. (...but then coding anything is too hard right? )
Anyway I think it'll be a combination of getting new things to play with and talk about and getting better performance....err because some of the new features look like like they should really improve performance and allow for other features to consume the surplus they provide.
Did that make sense...did I even make a point?
Well...I tried anyway at least.
