Great, great track Feels.. Well done and I second the comments others have said about the dirt on the track in some corners, the grass etc.. Tire bundles I loved, personally I would make them a little harder (if possible), they seemed to move too easily if hit, but then this could also be an ISI damage issue not working properly as yet.. I did notice one thing with Croft.. I noticed on the very last corner coming onto the front straight (before pit lane entry), you couldn't see any walls in the rearview mirror but you could see the opponents cars coming into the last corner. The brick work wall (along the grid) was fine and you could see this fine in the rear view mirror, but the railing / wall before this was not visable in the rear mirror. Other than that, I have nothing but phrases for your work.. Well done and hats of to you..
I think it might be as the bales and other objects on the likes of historical Spa etc hover sometimes when hit too.
Yes. Game engines can't afford to do physics for all moveable objects on tracks. Only those that are in motion at the moment will be simulated. When object stops, it's simulation is disabled again to save some CPU power. The trick is to recognize when object has come to a complete stop. It may happen that object will come to a stop in mid air for a moment before it starts to fall down again. Game may consider it to be in a state of rest and disable it's physics until another collision will set the object into motion again. This is the kind of stuff that needs to be sorted out by ISI before gold release. It's possible that they're going to let this one slip until some update after gold relase.
For example Rigs of rod has used method where there is certain range from player vehicle where physics are run, but when object is outside of that range physics of object are not run, which could result object to stop at mid air and then continue falling after player came near again. Don't know if such is used in rF2, but it is hard to get work right, imo. Method that K Szczech explains is much better, imo. Of course every method requires beta testing and fine tuning. Thx from this track, after an hour my download might be ready and can test if my computer is worthy
BTW and a bit OT, but I can see a space for using CUDA calculation here. I know that for example Arma3 will be/is using cuda for calculations. I know that compatibility problem must be resolved, and for some reasons it cannot be used for calculate car physics. But in case of minor objects it would be used imo with success. In worse case cuda may be calculated on CPU. But for most of us, cuda units 'are bored'. So there is unused power which may be utilized.
Latest version of Blender has cycles rendering engine, my GPU is not supported for GPU calculation, I use CPU and compared to Blender rendering engine computing time is something like 100 times larger, for some with GPU scene renders in 2-3 seconds for me with cpu 10-20 minutes. That might become of issue with CPU CUDA, or then not, but this is rather OT for here.
You mean PhysX? In most cases PhysX runs on GPU. Not many users have separate PhysX card. So it's not like there are unused resources because GPU is constantly working on rendering graphics. In the end it depends where you have reserves. If game engine is using tons of CPU then you may be interested in moving workload towards GPU, but it may be the opposite case just as well. Considering that some people will want to run on 3 monitors, then offloading GPU may be a good idea. And there's always a matter of kernel size. GPU is good for big kernels, so if you have thousands of objects then GPU will do it a lot faster. In case of racing games we have some objects around the track but they spend most of the time in rest. Only some of them will be in motion at a time, so CPU can handle it and you don't have to deal with CPU-GPU communication for physics. It's not like GPU is, let's say 1000x faster than CPU. It's only 1000x faster if you give it 1000 identical things to do simultaneously.
Vehicle's overall physics calculated on CUDA? I don't think so (won't be that effective), but aerodynamics (so, CFD basically) and maybe tyre physics in the future? Oh, and not CUDA and its legacy (which is backwards compatibility). I think DX11 and OpenCL, as it can utilize both nV and Ati gpus and is more flexible.
Damn, when I see "physics" I think "vehicles" Still though as long as such collision won't incorporate tens of thousands objects interacting each other, I don't think hardware CUDA/OpenCL will be better than CPU.
I remember Croft being my favorite track from Codemasters Toca 2 (1998). After the first lap with a Clio, I remembered why! What I really like are the 'imperfections', no straight line is straight: http://feels3.strefa.pl/sim/srpl/croft/pblc/CROFT_14.jpg (check the line on the right). The pit exit is quite dangerous. I wonder if it is a problem in real life.
Yes, I noticed that too, a lot of effort must have gone into that and the result is well worth it I'd say. Not just on the x/y axes, there's also a great amount of variation vertically.
I'd venture a guess and say even if you had 1080p onboard footage you couldn't tell either way with sufficient accuracy. Doesn't matter to me either to be perfectly frank. We'd be talking about the last 0.00001% of accuracy. I just don't care if it's perfectly spot on or 2cm off. The psychological effect of creating a more bumpy/alive/varied environment is what I enjoy about it.
