Just curious... I know in this day and age, storage is dirt cheap, even SSD space, so filesize for mods is not really a concern. But for example as I recently prepared a "Best of" DVD-ROM with rFactor2 content which is after all only 4.4GB or such, I realized I could get three Virtua_LM tracks on it for dropping one small ISI track (Mores). Sebring and Mid Ohio are still to me some of the better looking tracks we have available today, sometimes better than ISI and rather comparable with Feels3 stuff. Yet they are under 100MB each. How does this work? Doesn't this - in a reverse kind of way - suggest that many tracks are using up a lot of space for little visual difference? And since we know that RF2 has memory (RAM) limits as it's still a 32bit application, maybe we could squeeze a few more cars per race into our sessions if more tracks were made smaller. I'm running full details btw. on a 1920x1200 resolution, so it's not like I wouldn't be noticing a difference in texture sizes.
This is only a guess but conversions from rFactor seem to me on average a bit larger. VLM in rF was 50-100 mb Longford in rF is 240 mb, rF2 is 270mb But take Silverstone 2012 in rF it is barely 100mb Scratch ISI rF2 is 354mb All ISI tracks apart from Kart are 200mb+ Wild guess Mores I thought may have been higher mb for appearances because of the outfield road runs around it. lopl Scratch or extensively reworked tracks are also high mb Ponzan 300mb + Croft 224mb+ I would assume with reworked terrain to " suit rF2 properly " and graphical enhancements VLM circuits would be larger. On the beta versions ( v0.90-v1.91) they were nice additions but as yet do not convey same visual immersion and stimulation as rF , even GTR or GTL. Longford v0.92 is best "conversion" for mine to date.
The reason why tracks are larger comes down to better shaders, better model details and better textures. Conversions in most cases don't even change the textures on buildings etc so you end up with a bunch of small texture files for one model. Graphically these conversions are crap for many reasons from a model casting a shadow on an object that also has a fake shadow in the texture to the fact that also the models use one (1) texture and not the additional Bump and Spec map textures. So straight away, a proper track will have in most case 3 x textures to it's 1 x texture on a model. Another thing that will increase file size on an object is the fact that we can now export our models using 3Ds Max smoothing groups rather than using GMTs smoothing groups. The differences (visually/graphically) if done correctly is huge, and in file size is larger (not by a massive amount, but is larger in compassion). So you add these little increases up and you will also end up with a larger track. Another thing to add, are that we are now seeing more and more textures at 2048 in size with a couple 1024 or less, but in rF1 you would hardly see many 2048 texture sizes so this too adds to the file size. The advantages to larger texture sizes is less fps using one texture on one model (refreshing one image) to a model using several textures to one model (refreshing multiple images). Hope that helps..
Just want to point out that although it's nice to be able to export using Max normals, it should only be used as a last resort option. Based on my test experience with them, I have noticed a massive increase in trilists upon export, which is nice if you like bottlenecks. It takes a lot more power and time to process these objects. I can understand that vehicle artists would make good use of them for complex high poly shapes, but since this thread is about tracks, let me tell you that we have yet to come across a situation where we would have to take that last resort option. I just can't think of any track object that would require Max normals to make a huge difference. I would definitely recommend against using Max normals 'just because you can'. As for the secret behind small filesizes ... just remove most of the unique 3D details (and the dedicated 2D maps that go with them); replace large texture maps with small tileable textures that ... tile visibly; UV map all of the variation textures for curbs, grass, tyre walls, fences, walls, ... to one and the same small texture for each type, and you'd end up with a 30-50% reduction in filesize on the spot. You'd also reduce the atmosphere factor though ... Some people care about that, some don't .
Thanks for the info about the normals Luc That bit of info would be nice to be included in track technology.pdf
Thanks for the info guys. I was thinking it might have to do with shader maps, but to be honest I don't see "ugly" buildings in Mid Ohio or Sebring... it's not like they look out of place next to ISI tracks.
Great to see you appreciate our tracks that much Luc. There's no secret or magic about it, it's just how they're built. MonSum, who is the main author of Sebring and Mid-Ohio, is just very very good at doing a lot with very little. He has done some terrific work on doing some manual ambient occlusion, produced wonderfully effective textures with gradients. They're very efficient on poly use as well, which is the main thing keeping Also, Mid-Ohio is fairly short and Sebring in general is a fairly open space so even with the same modeling style and efficiency they'd probably at the lower end of the filesize spectrum. As you'll notice Le Mans is a bit bigger (although still below 200MB), mainly because I'm not quite as good as MonSum at making wonderful things from very little. Major things that drive up filesize I observed are mostly just complex 3d models. The guardrails alone are 68.5MB now, uncompressed. That is unless you use completely over the top texture sizes that some third party tracks use... but then you're just wasting space, bandwidth, and worst of all, performance.
It's not about "ugly" VS "good" so what I'm saying here is not a judgement for other tracks or track makers nor I'm taking any track out there as a comparison. Tricky to talk about this stuff without appearing critic over something else but the OP is asking "what's the secret" and I'll try to explain my point. It is true you can make nice tracks using less resources but there are differences, of course. It's mostly a matter of a choice (you can call it company goal, artist choice, standard or whatever); you can do something using more bits, you can do the same using less but no, they're not going to looks the same for obvious reasons. High fidelity tracks demands more than averaging, when you're describing something very complex like reality - with triangles, pixels and colors. This is also where games are going in this century, thanks to the hardware available in the consumer market. Putting more information on a track it's like introducing more life into it; something we think it's very important for a hopefully long life product (that's part of the goal). About Mores, used as example in your OP, it has been surveyed from A to Z, as for Silverstone, Lime Rock Park and Indianapolis. All the collected material has been used to build the track for both 3D and textures. This means you've unique features, like dedicated textures per curbs/turn, armco, gates...dedicated texture per different asphalts, dedicated textures for grass, verges and gravel, oil patches, wall paintings, rock embankments, different garage doors etc.... Almost 100% you see in the track it's based on IRL material, offering variations per object and this needs more memory than basics (and sometimes small) non-dedicated maps. Same for 3D; when you survey a track you have tons of images available per object/building/environment. This give you the opportunity to go into details and focus on model fidelity to match, as much as possible and without kill performances, the IRL counterpart. As for textures, this means more triangles and more memory. Another aspect is the pixel/meter ratio used for maxed mips, which affects the minimum distance available before starting seeing a texture looking bad. This is where you can "feel" a track similar to another until you start looking from different POVs than the driving seat. Same for 3D object models all around. This is also a choice, with different results. Sure you can do a Mores version using 1 single curb tile, 1 single grass tile, 1 single garage door, 1 single square tile rock for every rock surface, a single 1x1 paint color to cover both grandstand, pitwalls, track walls and also you could avoid CCTVs, detailed gates, heroes side vehicles and then use a 2D ring for the 360° hills panorama instead a series of 3D mesh.. etc... but the final atmosphere will be less similar to the real thing and for a such small and easy place, it's very important for us to focus on details or the cheap feeling will stand at the door and knock soon, with the risk to make the track boring to drive... due the lack of immersion and diversity. It's a continuous fight between the real life entropy and the discrete digital world and let me say this is the challenging (and fun) part of the art involved in the process. It's also important to say there is a difference when you build a track using web resources and when you grab lot of reference materials in location; I think it is pretty normal to put more information to recreate a place you surveyed to collect more information for this specific purpose (apart collecting track surface data).
Look closely Variety costs. Details cost. Finishing touches (edges, transitions) cost. Drive around Poznan by Feels 3 and look around. See how grass looks differently here and there. See how transitions between grass and road look like. Then try looking for such things at Mid Ohio. Virtua_LM tracks are great piece of work, but they don't seem to have a good technique for terrains. You could also compare texture size on buildings etc. Mid Ohio was originally made for rF1, so it had to deal with graphics cards having less memory. So time played it's part as well. Yes. You will not notice differences in texture resolution while driving. Not even during replays. Look at this curb: Would you notice it while driving, if the texture had 2x less resolution (meaning 4x less memory footprint) ? Not a chance. But it makes this screenshot look better and will make impression. It will also play part when people start comparing graphics between different titles. People like to have more of everything. Even if it's something they don't use In the end it's like Tuttle said - all about choices
These realism details are what separates the "immersive" tracks from the ones that look "digital" and artificial. But boy, does it take a lot of time and talent to create something realistic looking.
Interesting discussion that envolved here. I see the points made in favor of "more information tracks", and for sure Mores looks really good (for such a short bastard). OTOH until we get a 64bit Exe we will see people run out of memory as they add more AI. Somewhere in the middle is a nice place. Very good point made though that a lot of stuff is for "screenshots versus driving" - i.e. pCARs produces excellent screenshots at high cost, how much do we see of that while in cockpit?
I think part of the whole immersed thing is to be able to stop any point on a track and admire the fine details, whether you know they're there or not. Sure you may not see something from the cockpit while racing around the track, but knowing it's there and ready to be admired are a big factor for realism to me. I personally only have a select handful of tracks installed. All of which are either very high [ISI] quality along with a couple of my "older" favorites. When I see new content released I usually download and test. If it's not up to rF2 standards, I press delete. With the abundance of 3rd party content at this stage of development, it's only a matter of time till we have top quality versions of almost everything out there.
Another vote here for immersion. For me the operative word is "simulation". I want to be immersed in the experience. For me, a sim should have excellent physics AND sound AND visuals. I want to forget I am playing a game, and get as close to reality as possible. This is why I am keenly waiting for the Oculus Rift (and hoping that ISI implement it well). The game provides the framework, the tracks and cars provide the content, and the peripherals provide the actual sensory input (display, steering wheel, speakers/headphones) - all elements have to be as good as possible to provide an immersive simulation of reality. Therefore, in the interest of getting photorealistic visuals I much prefer highly detailed tracks with unique textures. I hate seeing tiling of textures, when using high altitude trackside cameras, or just looking down from an elevated portion of the track seeing obvious tiling and repetition of trackside textures is a major immersion-breaker. Photorealism is a product of both the graphics engine (technology) and 3D assets, textures, and shaders (content). I personally don't want to "settle" for less than what modern hardware is capable of. If the content is super-duper high resolution with fancy shaders - Crysis level quality - then it's ok to expect Crysis level graphics card torture I believe that ISI (and modders) should release the highest possible quality content. Those who have the hardware to support it (SLI/Crossfire GTX 780 for example) can enjoy it, and those who don't have the horsepower can easily crank down the ingame graphics settings. This way there is a choice. Thus I am glad to see that ISI's tracks that are coming out these days do indeed have high detail sizes. The way I see it, the larger the filesize of a track, the better, since it means higher res textures and more textures ISI, please keep making the highly detailed stuff (tracks and cars). In fact, I request, nay, I DEMAND that you make future stuff even MORE detailed than what you are releasing now
