David Kirk on HDR+AA
- Thread starter _xxx_
- Start date
- Status
Jawed said:
Kirk's answer doesn't really care either way, does it? In fact, as someone alluded to above, it could be read as wanting to give up on MSAA in hardware entirely for some intermediate number of years while developers get their s*it together, then seeing what they can do to accelerate the resulting developer solution(s).
JF_Aidan_Pryde
Regular
Xmas said:
Would you care to give a more accurate explaination?
Jawed said:
On the nitty-gritty tech end. . .heckifino. But I look back at the 24/32 argument, and I see an ATI that is 1). Very practical, while also 2). Spending a lot of time figuring out ahead of time if the limitations are going to end up smacking us in the eyeballs within the lifetime of the arch.
For the most part, those types of discussions are most relevant pre-implementation (of either new feature or enhancement to existing feature). I'd be very surprised if whatever tack they take doesn't look a lot better to the average Joe in existing games and games available over the next year, than current implementations, whatever its theoretical limitations (which of course we shall all have great fun dissecting here).
Jawed
Legend
Well, in a sense his comments could have some merit in the distant future.
There's a tendency with consumer graphics, to make all the old "fixed function, implemented in hardware" capabilities "programmable". That's why we have shaders now.
Also, if the shader aliasing ("sparkly effects on shaded surfaces" is the common complaint) can only be solved by putting anti-aliasing into shader code, then you can see right there an entry-point into a programmable AA future with no fixed function hardware.
But I can't see it happening soon.
Jawed
There's a tendency with consumer graphics, to make all the old "fixed function, implemented in hardware" capabilities "programmable". That's why we have shaders now.
Also, if the shader aliasing ("sparkly effects on shaded surfaces" is the common complaint) can only be solved by putting anti-aliasing into shader code, then you can see right there an entry-point into a programmable AA future with no fixed function hardware.
But I can't see it happening soon.
Jawed
That's right, because as we all know, it's impossible for a GPU with a 128-bit memory bus to beat the crap out of one with a 256-bit memory bus. Notice that this is Half-Life 2 we're looking at, and not Doom 3.
geo said:
No, I think you're right (assuming that all of Wavey's hints are based on accurate info and have been collectively interpreted correctly).
Jawed
Legend
Not too bad. On a benchmark that is known to favor ATI GPUs, the 6600GT (with a 128-bit memory bus) is within 15% of the 9800 Pro. That's not too bad.
Of course if you look at all the other benchmarks, the 6600 GT even edges out over the 9800 XT.
Now, where are those 9700 Pro numbers, so we can look at that David Kirk quote in context?
Silly comments like "oh man, DK doesn't know what he's talking about lolol!!1~ 128-bit s00x" have no place here. Maybe, just maybe, David Kirk is a smarter guy than you give him credit for.
Still sticking up for the 128-bit crowd pretty well - even with AA
http://www.xbitlabs.com/articles/video/display/2004-27gpu2_8.html
And what Bob said too....
http://www.xbitlabs.com/articles/video/display/2004-27gpu2_8.html
And what Bob said too....
Kirk is right
If you want HDR with AA, then just render a bigger buffer and downsample it to the screens resolution - sould work fine. Or you can apply some edge smoothing filters to the scene. The problem is - as he said - that there are no displayable floating-point buffers in curent hardware. That means, one must tonemap it manually and write the output to the visible screen. That's why the hardware antialiasing on the floating-pint buffer doesn't make much sence(commonly spoken).
If you want HDR with AA, then just render a bigger buffer and downsample it to the screens resolution - sould work fine. Or you can apply some edge smoothing filters to the scene. The problem is - as he said - that there are no displayable floating-point buffers in curent hardware. That means, one must tonemap it manually and write the output to the visible screen. That's why the hardware antialiasing on the floating-pint buffer doesn't make much sence(commonly spoken).
trinibwoy said:Still sticking up for the 128-bit crowd pretty well - even with AA
http://www.xbitlabs.com/articles/video/display/2004-27gpu2_8.html
And what Bob said too....
And what was the primary improvement of the rushed-forward 5900 over the cancelled-as-quickly-as-possible 5800? I can't believe we're even talking about this like it isn't a settled issue.
- Status
