R300 pics?!!{56k warning}
- Thread starter Ascended Saiyan
- Start date
Doomtrooper
Veteran
Smoothvision II from my contacts was a radically different way of doing FSAA, like Matrox's FAA yet these screens show otherwise. Possibly they have fixed the flaws in the 8500 Smoothvision and went that route.
Reverend said:
No argument there!
In fact, supersampling layers nicely on top of multi-sampling, too, if you've got the fillrate to spare anyways. I'd like to see future anti-aliasing implementations focus less on a single restrictive approach, and instead allow robust inter-mixing of multiple techniques. The ideal compromise between performance and quality is in there somewhere, and it's likely to be different for each application....
Nvidia's support for "4xS" mode isn't a bad start.... but why stop there?
Bambers said:
That's not entirely true.
4xRGSS as in V5 works fine on edges, and gives a virtual 4x resolution increase along each axis. The same would be true for a color-edge in a texture too, meaning that you could change mip lod -2 and get equivalent to 4xAF.
But then you'd get aliasing in textures that are more "random".
The reason is that OGSS relies heavily on the color changing in one direction, but being constant in the other. Great for edges, but not elsewhere.
[Edit]
Sign on mip lod change.
If you are indeed translating what it says under the 2x-4x-6x SVII setting, are you sure that's not just a driver quirk that's merely reporting the res the card is at? The previous screen shows the R300 is at 10x7 desktop res. And 3DM is at 10x7, tho I suppose it doesn't require the desktop to be at the same res as the test.cho said:
Exciting. So, Mufu, are we going to see it debut before QuakeCon?
*sprains eyelid* Doomtrooper
Veteran
The driver panel looks like a Catalyst release and currently on my setup that reference has to do with maxium resolution that the FSAA will run at.
This could also be a very early driver that doesn't even support the 'NEW' FSAA that supposed to come with the R300.
This could also be a very early driver that doesn't even support the 'NEW' FSAA that supposed to come with the R300.
Ascended Saiyan
Newcomer
So is it safe to say that this is .13 or not ?
Doom,If that is even a possibility it would be very bad for ATI because the rumour has already been circulated that this card has been sampling for months now & some if not most might or will be looking for a fully functional product but I'll be looking for the latter.I hope we just don't see another R8500 launch.
It should be interesting to see if they choose to go with a stochastic sampling methods in the ant-aliasing implementations & to be able to use Trilinear with Anistropic plus if it can do that at a reasoneable rate in todays games like Morrorwind without any problems then I'm sold. 8)
Doom,If that is even a possibility it would be very bad for ATI because the rumour has already been circulated that this card has been sampling for months now & some if not most might or will be looking for a fully functional product but I'll be looking for the latter.I hope we just don't see another R8500 launch.
It should be interesting to see if they choose to go with a stochastic sampling methods in the ant-aliasing implementations & to be able to use Trilinear with Anistropic plus if it can do that at a reasoneable rate in todays games like Morrorwind without any problems then I'm sold. 8)
Bambers said:
It's seems to me unlikely... with super-sampling, the intent is generally to attempt to evenly distribute the samples across the surface area of the pixel. The anisotropy of the texture would favor sampling with points distributed across a specific axis within the pixel. The two sampling distributions won't normally coincide optimally... or, at least, won't do so consistently.
Hmm... I might need to revise my statement.
I was thinking in terms of 4xanisotropy => -2 mip lod relative no AA and no AF => aliasing in some cases.
But I didn't think about if it was possible to avoid those cases. So here's take two.
Do 4xOGSS and use mip lod -1 relative no AA when looking straight at a surface. But as the needed anisotropy increases, increase the mip lod to -2 relative no AA (at needed anisotropy = 4x).
The reson it should work without getting aliasing is that when the needed anisotropy goes up, color changes will indeed become "1D" as in changing fast in one direction and slow in the other.
As Dan G says, it won't be a perfect sampling pattern for AF, so the mip lod might need to be adjusted slightly. Experiments need to be done to determine how to do it, but the idea should work.
Sorry for being OT.
I was thinking in terms of 4xanisotropy => -2 mip lod relative no AA and no AF => aliasing in some cases.
But I didn't think about if it was possible to avoid those cases.
So here's take two.Do 4xOGSS and use mip lod -1 relative no AA when looking straight at a surface. But as the needed anisotropy increases, increase the mip lod to -2 relative no AA (at needed anisotropy = 4x).
The reson it should work without getting aliasing is that when the needed anisotropy goes up, color changes will indeed become "1D" as in changing fast in one direction and slow in the other.
As Dan G says, it won't be a perfect sampling pattern for AF, so the mip lod might need to be adjusted slightly. Experiments need to be done to determine how to do it, but the idea should work.
Sorry for being OT.
Ascended Saiyan:
I don't think that any card is really going to play morrowind reasonably for some time. It seems to be doing back to front rendering, and basically all cards right now are sucking it up with that game.
I've got a athlon 1700+ with a gf3 ti200, and in balmora I get a min fps of about 8fps, with maybe average hanging around 10. My friend who just built a 2.4GHz P4 system with a ti4600 pulled about 15fps min and maybe about 17-20 on average.
If you consider 15-25fps average in the more complex areas a "reasonable" rate, then I guess it's all good. Just don't expect any new cards to perform miracles on that game.
I don't think that any card is really going to play morrowind reasonably for some time. It seems to be doing back to front rendering, and basically all cards right now are sucking it up with that game.
I've got a athlon 1700+ with a gf3 ti200, and in balmora I get a min fps of about 8fps, with maybe average hanging around 10. My friend who just built a 2.4GHz P4 system with a ti4600 pulled about 15fps min and maybe about 17-20 on average.
If you consider 15-25fps average in the more complex areas a "reasonable" rate, then I guess it's all good. Just don't expect any new cards to perform miracles on that game.
Nite_Hawk said:
Wouldn't a defered renderer be a perfect fit for this situation? If so, perhaps when/if the next PVR card comes out, it would do wonders for this game.
Though it sounds like the game needs to be patched to do front-to-back rendering, which is friendlier towards non-defered cards.
--|BRiT|
Why?
Well, the game chugs badly with many houses on-screen, which could well indicate massive overdraw.
But I don't think it's doing simple back-to-front rendering. For starters, I can't believe the coders would ACTUALLY be THAT incompetent,
and besides... When I O/C my GF3 to 255/558MHz I get those well-known squares of white pixels. Funny thing is, they don't appear more or less random in this game, they definitely favor certain areas on the screen, areas that cover another poly such as a rock in front of a hill or such. That makes me believe there is indeed occlusion detection going on in the engine, and that the overclocking exposes some kind of critical speed path in that area of the chip that causes these white quads to appear.
Also, the Unreal2 engine stresses the GF3 massively more than ANYTHING ELSE that I've seen so far. Older engines run just fine on my card with nary a white pixel quad to be seen, but there's tons and tons of 'em in America's Army. And there's discolored polys too that appear fairly frequently, not sure if that's due to a bug in the engine, game or drivers, or my O/C:ing, I haven't tried to reduce speed yet. But it definitely pushes the card harder than anything else before it. And it runs S.L.O.W. in 800*600 with 4xS AA activated. Way slow!
...And there's not even any heavy action going on!
Well, the game chugs badly with many houses on-screen, which could well indicate massive overdraw.
But I don't think it's doing simple back-to-front rendering. For starters, I can't believe the coders would ACTUALLY be THAT incompetent,
and besides... When I O/C my GF3 to 255/558MHz I get those well-known squares of white pixels. Funny thing is, they don't appear more or less random in this game, they definitely favor certain areas on the screen, areas that cover another poly such as a rock in front of a hill or such. That makes me believe there is indeed occlusion detection going on in the engine, and that the overclocking exposes some kind of critical speed path in that area of the chip that causes these white quads to appear.
Also, the Unreal2 engine stresses the GF3 massively more than ANYTHING ELSE that I've seen so far. Older engines run just fine on my card with nary a white pixel quad to be seen, but there's tons and tons of 'em in America's Army. And there's discolored polys too that appear fairly frequently, not sure if that's due to a bug in the engine, game or drivers, or my O/C:ing, I haven't tried to reduce speed yet. But it definitely pushes the card harder than anything else before it. And it runs S.L.O.W. in 800*600 with 4xS AA activated. Way slow!
...And there's not even any heavy action going on!
Morrowind and back-to-front sorting
Hi there,
regarding Morrowind and sorting, here's some info from the NetImmerse web site (http://www.ndl.com/netimmerse/technical.html):
so, yeah--it's possible that Morrowind uses back-to-front sorting for the largest part of the rendering process. Whyever they would do that, though, is another question entirely . . .
ta,
-Sascha.rb
Hi there,
regarding Morrowind and sorting, here's some info from the NetImmerse web site (http://www.ndl.com/netimmerse/technical.html):
so, yeah--it's possible that Morrowind uses back-to-front sorting for the largest part of the rendering process. Whyever they would do that, though, is another question entirely . . .
ta,
-Sascha.rb
As some other people mentioned, there is a *lot* of overdraw going on in morrowind. Unfortunately the pictures have been removed, but the following thread details some of what people have come up with as to what the engine has been doing:
http://www.elderscrolls.com/ubbthre...2&page=&view=&sb=&o=&vc=1
Nite_Hawk
http://www.elderscrolls.com/ubbthre...2&page=&view=&sb=&o=&vc=1
Nite_Hawk
Nite_Hawk said:
8fps?! Thats sounds very low. I have an 8500 and 1Ghz tbird and get 14fps minimum in Balmora with max details except AI distance is set to minimum.
I think morrowind is very cpu limited as indoors as everytime another NPC comes into veiw I loose a large amount of fps.
Similar threads
- Locked