First of all I should point out that this isn't a criticism of ISI's track making techniques - my knowledge of how you do all this stuff ranges from very little to absolute zero, so this is simply a query. When I drive circuits made by feels3 I have very little (if any) aliasing on the white lines - I can use a basic 'Level 4' AA from the rFactor 2 Config program and they look great. Here are two example screenshots, Yes you can see very slight aliasing on the second screenshot, but it's nothing major and nothing that would make me want to add extreme AA. However when I drive ISI circuits with the same Level 4 (or even Level 8) AA I get aliasing which I consider to be pretty bad - an attack of the dreaded 'jagged white lines', as they're infamously known. The only way I can improve them to what I consider an acceptable level (yes I am very picky) is to use Nvidia Inspector to add 2x or 4x SparseGrid SuperSampling. Even normal SuperSampling doesn't seem to improve things - it has to be the special 'SparseGrid' setting. I have some example screenshots, Basic Level 4 AA Level 4 AA + 4x SparseGrid Basic Level 4 AA Level 4 AA + 4x SparseGrid So my basic question is this - do you know why feels3's circuits don't have the same level of aliasing as your own, and why the white lines on those tracks look smooth even with just basic levels of AA? I'm lucky that I can actually use 2x SparseGrid on tracks like Silverstone and still have a decent framerate, but there's no doubt that using this extreme AA hurts fps a lot. And I know some people might think "who cares if the white lines are a bit jagged?", but I like my image quality to be as high as possible - those painted white lines are all over the track surface right where your eyes are focusing, so you can't really get away from them. Thanks for any info.
Yes I do. I sometimes turn it off on feels3 tracks though as it causes problems with his trees for some reason.
Because those Poznam stripes are 2 times wider than SS stripes and as they're both real mesh (decals above the road) they're affected by how many pixels are used to describe the object along the screen pixel grid. If you compare two similar stripes (even the Poznam is still wider) you can see the AA is pretty the same. View attachment 11728
I see. Well actually I don't really understand what that means exactly, but if that is the reason maybe Luc or Tuttle could comment on this.
No no, they're both decals. BTW a decal is a separated mesh above the road and aliasing is affecting the mesh edges; When a stripe is painted into the main diffuse it is rendered as part of the raster image and it's affected by texture filtering per texture resolution. We're using a separated mesh to give stripes dedicated physics material proprieties (.TDF file) for wet and dry conditions.
So it's down to the width of the white lines. That makes sense, as a really thin line will be much harder to smooth out when viewed from a low angle. I never thought of that. Thank you for the reply - it's something that i've wondered about for a while but just didn't get around to asking.
Yeah. In that specific case (Poznam Vs Silverstone) Feels3 is using a very dense mesh with far more tris to describe the road line lofting. This also helps to reduce aliasing where the line is describing a curve in front of you, as it has more triangles to fill the screen pixel grid. As Silverstone is a HUGE track with plenty of lines, intersections and track variations, we used a less dense mesh to model all white stripes to reduce the tricount impact which is very high due the circuit complexity.
That is interesting. I assume the tricount is GPU intensive, so maybe in the future (when graphics cards will obviously be more powerful) you could consider using the higher tricounts for any new tracks as hopefully it won't have such a big impact on framerates. Or perhaps you could have two versions (using a detail setting in the graphic options), so people with top end systems can use the higher tricount and people with lower end systems can turn the setting down. I get the impression making two versions of all the white lines on a circuit might be a bit of a time consuming nightmare though. It's amazing to think so much thought and technical detail goes into something as simple as a white line on a track - thanks for the insight.
That is just ridiculously awesome. It's these little things which are slowly but surely raising this sim to the point where I don't drive anything else. I had heard you were doing things like that for the curbs (which I already found impressive), but for the painted white lines on the tarmac? Crazy! I love this level of detail.
Eh eh...in rF2 everything you see along the road it is NOT just visual. White stripes for example, as you know now, are drivable surfaces so they're going to add informations inside the collision database (as for curbs). This is why we can't think drivable decals, side terrains and curbs just like visual objects and freely use a custom design/resolution, just for those details. We're using the same resolution for both roads and stripes to help a clean drivable surface continuity and to reduce at minimum unwanted physics behaviors when driving above different drivable surfaces using different longitudinal resolutions. Poznam, to stay on the OP example, is using 2x resolution on turns, this means 2 polys for the decal and 1 poly for the road, while Silverstone (as Mores or other ISI contents) is using a 1:1 sync between the two overlapped surfaces. Not a big deal with a 2:1 but it's better keeping the vertices to vertices approach, even when surfaces are overlapped and using a vertical distance offset. Mores looks a bit more dense on decals as the road mesh is using a linear quad poly design versus the "optimized" one (which reduce polys along straights). As the quad poly density helps a lot on bump/depression and morphology design I guess we'll use that way in the future for all new tracks so the visual issue we're talking in the OP should disappear or, at least, substantially reduced.
quick question for someone who doesn't track mod: as the ISI white lines are their own entity does that mean they react differently to tire inputs? I.e. in the rain they are more slippery than the track surface.
I been digging in rf2 tracks and discover that, white lines, curbs, outside alphalt and the very road has distinctive real road atributes, graphical and physical, people when talk this sim and that sim, should know that, rf2 for me, is the best.
