In this thread: lots of placebo.
RG/SGSSAA
|
09-26-2013, 11:16 AM
@delroth
Screenshots don't lie. And he posted some. Oh boy haven't had one of these in awhile! Spoiler: BONKERS Wrote:There is no "official" definition for many of these items. Yes there is. Ask a hundred different scientists or engineers what temporal means and you'll get the exact same answer from every single one of them. The same goes for asking anyone in the visual effects industry (CGI) about temporal aliasing. BONKERS Wrote:Temporal may mean silly things like Motion Blur to reduce aliasing to you. Not just me, the entire film and game development industries. Just do a google search on it if you don't believe me. Every result on the first page links to documentation using the term this way. If you mention temporal aliasing that way in front of anyone who does CGI work they will shoot you some funny looks. BONKERS Wrote:But to those of us who've spent hundreds to thousands of hours working on reducing any and all aliasing in every game available they have a different meaning. If there are a few people that don't work in the industry that incorrectly assign a completely different meaning to a term that is none of my concern and does not make their definition any more correct. BONKERS Wrote:To be temporally stable is to have a believable coherent image that doesn't suffer the artifacts inherent to the medium and method which they are created. But now you're talking about temporal coherence in image processing. Not temporal aliasing. Those are two completely different terms. BONKERS Wrote:One look at definitions for the word "Temporal" Produce many meanings From that very link: Wikipedia Wrote:Temporal can refer to time, or to material existence and secularity, or to the temple in anatomy. Below is a list of possible uses. In science, engineering, math, or anything related to those three (including computer science) the word temporal ALWAYS refers to time. Ask anyone in the industry. It's also used as a word in some religions and in medicine to refer to temporal lobes. But that's about it. Temporal here clearly means time. There should be absolutely no ambiguity about that. BONKERS Wrote:http://www.cg.tuwien.ac.at/research/publ...ements.mov Here's a video from The type of algorithm and implementation of concepts are not even remotely similar. BONKERS Wrote:I've seen so called "Developers" refer to the same and similar issues by a million different terms. I'm not sure what issues you're referring to but I've never seen one use the terms temporal or temporal aliasing incorrectly. BONKERS Wrote:I'll even leave this here Definitions are not opinions. And that image is referring to a temporally stable algorithm. Not temporal aliasing. BONKERS Wrote:SGSSAA, DOES infact reduce pixel crawl. I said this. BONKERS Wrote:SGSSAA DOES Infact reduce shimmering(Which can result from MORE than just geometry edges.). And this. BONKERS Wrote:SGSSAA DOES infact reduce MOIRE. And this. BONKERS Wrote:SGSSAA DOES infact reduce specular aliasing (And resulting shimmering/pixel crawl that so often HAPPENS because of said specular aliasing). And this. BONKERS Wrote:SGSSAA DOES Infact reduce basically ALL Geometric Aliasing. And this. BONKERS Wrote:SGSSAA does infact reduce just about ALL aliasing. Presents a temporally and believably coherent image that has next to 0 artifacts in motion and in stills. BONKERS Wrote:And if you actually watched the video I posted, you can plainly see with your eyeballs that SGSSAA destroys almost all aliasing of every kind. And that's using a game as an example that is one of the few which SGSSAA doesn't get all Aliasing visible to the eye. I never said it was ineffective at removing aliasing. Again you're putting words in my mouth. BONKERS Wrote:You'd do well to study and figure out what makes Nvidia SGSSAA work. Because it WOULD benefit just about ANY 3D rendered game. I did. Extensively. See below. BONKERS Wrote:Much more so than simple SSAA patterns like OGSSAA. There are a number of games where OGSSAA produces the same or better image quality. This is entirely dependent on which postprocessing effects are implemented by the game and how. BONKERS Wrote:Again, in motion SGSSAA blows any other current method out of the water. Especially bog standard SSAA Ever seen HSAA? It's a real time method and I would certainly argue its superiority over nvidia's SGSSAA implementation. I have been using SGSSAA since almost the beginning when the first German guides were being written (before any English guides were available) while we began experimenting with more modern solutions and cracking the hex compatibility codes. I was using SGSSAA literally from the very first beta driver that supported it. Back then I started writing an AA guide. And while I never finished it several of the articles are still referred to by a number of communities involved in this movement to this day. Trust me I am well aware of the benefits of SGSSAA and SSAA in general. Anyone here will tell you that I was the main proponent in pushing our graphics devs to develop and maintain a functional SSAA option in the d3d9 backend. I like SGSSAA too. So stop trying to fight me on an argument that I haven't made. Let me let me break down the exact points my original post made one by one for the sake of debating them: 1. Your numerical values (90% and 99%) make no sense. I don't know how you're measuring them and I assume that you are pulling them out of thin air. 2. Specular aliasing is a form of shader aliasing so listing them as two separate types of aliasing makes no sense. 3. The term temporal aliasing refers to motion judder. 4. OGSSAA/RGSSAA effect the same types of aliasing as SGSSAA. While you didn't directly state that they don't you did imply it. 5. The effectiveness of SGSSAA depends more on the game engine than the algorithm itself. It has no significant inherent advantage. Other forms of SSAA can be implemented in a virtually indistinguishable manner. 6. The popularity of SGSSAA over OGSSAA/RGSSAA among enthusiasts is mainly due to its widespread compatibility. Not necessarily its image quality. 7. Pixel crawl is a form of geometric aliasing, not temporal aliasing. 8. Texture shimmering and moire are caused by texture aliasing, not temporal aliasing. 9. The quality advantages of SSAA can be observed with screenshots and do not require video. Please note that nowhere did I claim SGSSAA was bad or ineffective at removing aliasing so you're preaching to the choir so to speak. Now points 1, 2, 4, 6, 7, 8, and 9 you did not address so I assume that you agree with me. In fact you posted proof of point number 9. 3 you have debated but I still strongly disagree. The use of this term with its correct definition is extremely widespread. There should be no ambiguity about it and randomly using a completely different definition for that term is both wrong and confusing. I don't hate you and in fact I'm on your side in this "battle". But I saw you make an error. You used a term incorrectly. So I corrected you. It happens. There is no shame in it but you should have verified your definition with google before trying to argue that I was wrong. Point 5 we have kind of sort of touched on. This can be verified both by looking at the specific implementation and via screenshots. SGSSAA attempts to remove static aliasing only. Since aliasing in a sequence of frames is made up of the aliasing in individual frames "temporal coherency" (consistency with respect to time of something, in this case aliasing) is directly effected by the use of static AA algorithms. An AA algorithm doesn't even have to have a temporal component to reduce it.
"Normally if given a choice between doing something and nothing, I’d choose to do nothing. But I would do something if it helps someone else do nothing. I’d work all night if it meant nothing got done."
-Ron Swanson "I shall be a good politician, even if it kills me. Or if it kills anyone else for that matter. " -Mark Antony (09-25-2013, 03:19 PM)NaturalViolence Wrote: In addition to that OGSSAA and RGSSAA effect the exact same forms of aliasing as SGSSAA. The only difference is the sampling pattern. They have no inherent disadvantage against SGSSAA in this area. The main reason SGSSAA is so popular is because you can easily get it working on so many games. OGSSAA (nvidia) and RGSSAA (AMD) by comparison work on a much smaller number of games. The quality difference between them though is negligible and depends more on the game engines design than the AA algorithm itself. I would say: SGSSAA needs less supixels to achieve the same EER as OGSSAA (hence the name sparse). 8xSGSSAA has a better EER then 9xOGSSAA, for example. If you have the same number of samples, SGSSAA will work better on the most important angles (the nearly horizontal/vertical ones - 45 degree angles are not that important, obviously), and it will do a better job fighting moiree effects due to the irregular sample pattern and the fact that the subpixel grid does not equal the pixel grid alignment. OGSSAA has a slightly (!) better effect on textures when you have the same number of samples in the border-regions where textures are just about to shimmer, as the distance between the subpixels is optimal. Also, if you consider Downsampling (by custom driver resolutions), which is basically the same as OGSSAA, you can make OGSSAA work in almost ANY game, . It even works in Dolphin. The max resolution depends on your monitor, tho, and only 1.5x or 2x factors are adviced (for a good reason). SGSSAA does not work in many games (DX10/11 titles without MSAA support) and can cause artifacts/errors/blur when forced. The latter is also true for OGSSAA set in the drivers; Downsampling does NOT cause this. 10-02-2013, 01:44 PM
Y'know, once high-DPI screens are more commonplace you guys are gonna stop giving a shit about any variety of AA more specific than "9x SSAA".
<@skid_au> fishing resort is still broken: http://i.imgur.com/dvPiQKg.png
<@neobrain> dafuq <+JMC47> no dude, you're just holding the postcard upside down ---------------------------------------- <@Lioncash> pauldachz in charge of shitposting :^) 10-02-2013, 02:39 PM
(10-02-2013, 02:29 PM)Gabbyjay Wrote: 9xSSAA is not the best available option in Dolphin today. That probably wasn't his point. He's pointing out that high DPI monitors and display may make most aliasing imperceptible to many people, thus you could just use 9xSSAA and call it a day in terms of image quality. I'm out of my depth here, so feel free to ignore me. 10-02-2013, 02:52 PM
Are you sure this is also true if you're projecting the image on a 20 feet screen?
Not everyone is playing on a 27 inch monitor. Also, there's more to it then just edge (anti) aliasing, it's also about texture sharpness, fine details and the accuracy of post processing effects. Also, you might want you future graphic cards you buy in three years to sweat a little. 10-02-2013, 03:23 PM
Gabbyjay Wrote:Are you sure this is also true if you're projecting the image on a 20 feet screen? The operative words were "high DPI screen", which a projector is not (at least not high enough, see following sentence). I get the feeling paul was hinting towards "Retina" like displays, think 220 PPI and above. Even if you used projectors, you could probably get away with it if/once they reach "Retina" levels in regards to DPI and resolution. Gabbyjay Wrote:Also, there's more to it then just edge (anti) aliasing, it's also about texture sharpness, fine details and the accuracy of post processing effects. I was under the impression that SSAA takes care of a range of aliasing scenarios (edge, texture, light/shadows, and post processing effects). (10-02-2013, 03:23 PM)Shonumi Wrote: The operative words were "high DPI screen", which a projector is not (at least not high enough, see following sentence). I get the feeling paul was hinting towards "Retina" like displays, think 220 PPI and above. Even if you used projectors, you could probably get away with it if/once they reach "Retina" levels in regards to DPI and resolution. There are already some UHD projectors. Problem is, it will take a long time till prices drop enough, and even if they can be bought for a reasonable price: Aliasing on a large projection screen is far worse then on a standard PC monitor, even if both use 3840x2160. We won't see more than this for several years (there goes the "high dpi retina quality" ...), as this market is also tied to the resolution of the content delivered to it: Movies. Quote:I was under the impression that SSAA takes care of a range of aliasing scenarios (edge, texture, light/shadows, and post processing effects). It does, what i was trying to say was: While 9xSSAA might be enough for some cases (edges, for example), other scenarios (like textures or post processing filters) still might take advantage of further refined anti aliasing methods. 10-02-2013, 04:06 PM
Gabbyjay Wrote:While 9xSSAA might be enough for some cases (edges, for example), other scenarios (like textures or post processing filters) still might take advantage of further refined anti aliasing methods. Again, paul's point was that any further AA beyond SSAA might be moot in terms of IQ once you reach such high DPIs and resolutions. Aliasing is a byproduct of the display not having sufficient picture elements to reproduce a given image accurately enough. The two obvious solutions are AA techniques, and/or just increase the amount of available pixels. If you have a sufficient amount of picture elements to begin with, the need for AA, or even refined levels of AA, begins to drop. |
« Next Oldest | Next Newest »
|
Users browsing this thread: 1 Guest(s)