Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

k4bn

Member
  • Content Count

    95
  • Joined

  • Last visited

Awards


This user doesn't have any awards

1 Follower

About k4bn

  • Title
    Member

Recent Profile Visitors

1,092 profile views
  1. @Magix I can't really see the issues that you are describing on your in-game screenshots, it looks like standard Far Cry 5 to me. I would love to help more, but that might be why it's hard to convince other people, assuming that there is actually something abnormal going on with your system.
  2. Yeah, it looks like there's some aliasing left on the helmet, primarily the white parts where the light gets reflected. However, MSAA doesn't work on this particular type of aliasing, it can't solve shader aliasing from the shader on the helmet because it only works once per pixel. This can only get solved by some kind of supersampling. The white lines on the car just look low res. It's a really old game, so that's understandable. And I'm not sure about the jagged windows icons and other stuff, can you show some examples of that? The in-game screenshots are still incomprehensible, there's nothing to even comment on there.
  3. Not sure what you mean about the colors, reflections etc. I haven't played rFactor, but those look fine in the screenshots. The only difference is that the fence uses a high negative LOD/mip bias, probably like -5 or something. Since clamp doesn't work, the sharpness from the high negative mip bias causes the aliasing/shimmering. In the other screenshot clamp works and the mip bias of the fence gets set to 0. Compared to the fence and the road lines(and some other stuff you can see in the screenshot), the other objects seem to have a default mip bias of 0 or 1, so there is no difference when clamping. And yes, AA is not the "issue" here, both examples use 4x MSAA. The issue is that clamp doesn't work in both examples. As for the pics you posted, I'm sorry, but it's incredibly hard to tell what is going on here because I have no reference point to anything. I advise you to at least take full screen screenshots in-game and explain what you think is wrong or, like I did, provide comparison screenshots.
  4. Never said you had any misunderstanding about AA, only that some users in the thread do. Yeah, I read your other posts and I get the general idea, but I don't think that you have the same issues that Kelven is talking about. It's mostly this difference:
  5. You can only fix the first issue completely in some games that have DX9 or DX10 by using AMD driver supersampling(games like Dishonored, Half-Life 2, L4D2...) or doing manual LOD bias adjustment in regedit. For NVIDIA, you can do it with SGSSAA(DX9/10 only) or by changing LOD bias in inspector. Both AMD supersampling and NVIDIA SGSSAA do an auto-LOD adjustment so you get rid of the negative LOD shimmering/aliasing and the temporal aliasing almost completely on 8x(depends on the game though, some are better, some are worse). https://docs.google.com/spreadsheets/d/1ekUZsK2YXgd5XjjH1M7QkHIQgKO_i4bHCUdPeAd6OCo/pub?output=html You can also try to do manual LOD bias adjustment in the NVIDIA inspector. https://forums.geforce.com/default/topic/529407/geforce-drivers/clamp-negative-lod-bias/14/ All of this only works in DX9/10, for DX11 games it's impossible AFAIK. The only thing you can do is regular downsampling/VSR/DSR, but this doesn't do auto-LOD adjustment so you will never get results as good as you would with SGSSAA, thin objects like fences will still flicker in the distance. The only new AA type that works nicely with negative LOD is TAA and the developers need to implement it correctly so it's works similarly to SGSSAA. This is still one of the best implementations that I know of, and it's been modded into the game: A decent TAA + at least 2x DSR/OGSSAA is IMO the best solution for newer games for image stability if you can afford it and if the game supports it.
  6. The problem is that there's a variety of issues mentioned in the thread. The one that is most prominent that seems to be the case for most of the users is the universally broken LOD bias. For example: https://www.reddit.com/r/hardware/comments/3rwy5m/nvidia_will_fix_excessive_gpu_idle_power_use_at/cws4xl4/ (#6) Note that this isn't nvidia exclusive, it's tied into the rise of DX11 and the lack of DX9 and DX10 in games where the developer has full control. The next one is the misunderstanding of how anti-aliasing works and how a raw aliased image looks like in motion. For example, a lot of people still think that this is an "issue", while it is most certainly not and I've proved it countless times: https://www.youtube.com/watch?v=2XXS5UyNjjU&t=164s (at 2:44 you can obviously see what is going on) I'm not sure what issues you have, if you want further help you should show us exactly what you mean.
  7. I saw the video you sent me and I saw the explanation. The person is not talking about any sort of a problem. He is simply telling you how to make and add your own mipmaps to an SDK for a texture that has no mipmaps to begin with. This is only useful if you intend to do modding. I am glad that you are finding this interesting, but I advise you to first get into the fundamentals of graphics, then the details of how aliasing and anti-aliasing works before going into the more complicated bits of graphics. Trust me, these sort of ideas come here almost monthly and everyone gives up after a few days because they are trying to solve something that they don't understand. https://alfonse.bitbucket.io/oldtut/ http://naturalviolence.webs.com/generalaa.htm Also, try doing this in OpenGL to get a general idea of what mipmapping is and to test out if it works. https://www.youtube.com/watch?v=H1ogRfMsc64
  8. I see where the confusion is, but there is a difference between using alpha tested objects + negative LOD(these trees for example at 5:42) https://www.youtube.com/watch?v=KTyhI0BCP4U vs no mipmapping and nearest neighbour. Read this: https://medium.com/@bgolus/anti-aliased-alpha-test-the-esoteric-alpha-to-coverage-8b177335ae4f The first video is basically a person modding in his own mipmaps, so if this is what you want, my advice is to start modding every single video game that has no mipmaps, I'm fairly certain that this has nothing to do with your GPU not being able to mipmap or something along those lines. As for LOD/mip bias, read this: http://naturalviolence.webs.com/lodbias.htm Here's some reasons why developers use TAA with negative LOD, the results are pretty close to normal supersampling. https://twitter.com/SebAaltonen/status/943078228071198720 Maybe tweet some of them if you are more interested in this stuff.
  9. Hello and welcome to the thread! Thanks for the information, a lot of it has already been addressed and you are definitely on the right track regarding your issue. The Skyrim S.E. textures in the video most definitely use mipmapping, however, as you've probably figured out, a negative LOD bias is in use, hence the texture shimmering. The problem with your video is that you do not have anti-aliasing on(correct me if I'm wrong). The reason why the game uses negative LOD is because it uses TAA, I explained this in one of my last posts. I've booted up Skyrim S.E and the game looks pretty clean with TAA on, so make sure to check if that was the issue. Similar thing happened to Witcher 3 when the devs changed the TextureMipBias to -1 in a patch. Users quickly started complaining, so the devs put the bias in the config so you could change it back to 0. Compared to Skyrim S.E, Witcher 3 has a weaker temporal AA implementation, so it doesn't work as well as Skyrim's negative LOD does with TAA.
  10. Props to xquizt for making a PUBG comparison with his friend! https://www.techpowerup.com/forums/threads/anti-aliasing-not-working-for-any-game-on-my-pc.232367/page-3#post-3829357 @xquizt Even though the aliasing in this example is technically "normal", it doesn't mean that it should remain that way. The 1080 has a lot of extra GPU power on 1080p and you should be able to force a higher level of supersampling to reduce aliasing to a minimum. MrBonk's post on temporal aliasing summarizes this very well. https://forums.guru3d.com/threads/which-anti-aliasing-reduces-temporal-aliasing.394462/ Note that I don't play PUBG, so I'm not sure how temporally stable the game is on higher levels of OGSSAA/DSR.
  11. The "flickering on edge" is due to negative LOD/mip bias. Negative bias with TAA emulates supersampling for textures and gives you 2x AF for free, it essentially converges to temporal supersampling. Like when you do 8x SGSSAA, the driver usually does auto-LOD adjustment to -1.5 LOD bias(you had to set these values manually in the good old days), this compensates for the blur of the SGSSAA. Similar thing happens here, this is basically why TAA is used with negative LOD/mip bias and gives such a nice result both in Fortnite and in the BF1 example with the alpha tested negative bias shimmering foliage. And yes, TAA is implemented differently in different video games. It doesn't 100% remove aliasing itself and sometimes it doesn't even work on some types of aliasing(above example with moire pattern). The point is, xquizt says that changing AA settings doesn't make a difference in any game, so apparently he doesn't have the same "problem" that you have, because clearly TAA works for you in this particular instance.
  12. Can you record gameplay footage of a game that actually has good anti-aliasing where it should be very easy to notice the difference? Like Fortnite for example, I remember that lordante was able to get the game approx. 90% clean with "Ultra" anti-aliasing, compared to "Low", which had a lot of aliasing. If you decide to do that, make sure to record changing the different AA settings and recording the differences, or lack thereof.
  13. The aliasing in the last few videos is mostly due to negative LOD bias and alpha testing, this has already been discussed. I don't play PUBG, but this video seems to explain it all: Check out 1:43 for example, "Ultra AA" also has a lot of aliasing, so it's probably just the game. You can't change LOD bias in dx10+ if the dev chooses to prevent driver overrides, so best thing you can do is try out what I already mentioned in one of my posts. However, most dx10+ games don't allow you to do that, so the only thing left is probably just supersampling the game. Your 1080 should have no problem with that. http://naturalviolence.webs.com/lodbias.htm
  14. Hello and welcome back to the thread. Please re-read my previous posts, you seem to be be confused on what AA is and how it works. As for your friend, please post some gameplay footage showing that. Although, honestly that wouldn't really prove anything because you can still set SGSSAA x4+ and AA off in-game and "rig" your proof just to prove your point. Also, please consider that you may not have the same problem as some people in this thread. You mention that AA doesn't work for any game, but a lot of people are claiming that temporal anti-aliasing solves the "problem". That's partially correct, but even the "best" TAA methods don't work on all types of aliasing, for example: left side is TAA + triangle-geometry, right side is TAA + raytrace and jitter, but it's basically a custom shader for this particular surface
  15. It seems that you are mostly concerned with temporal aliasing(shown in 0:27 here https://www.youtube.com/watch?v=_V8bBLcbRIs for example) Use this to reduce it: https://www.youtube.com/watch?v=8scOOiAvyj8 if driver supersampling isn't working. You can read my older posts about this sort of stuff and there's useful stuff in here http://naturalviolence.webs.com/generalaa.htm
×