Separate names with a comma.
Issues with the game?
Check the Known Issues list before reporting!
Discussion in 'General Discussion' started by Mythbuster, Sep 26, 2014.
the comments on the video make me want to cry
I'm a serious driving game kind of guy and I think BeamNG Drive is a blast. The attention to detail on the roads (ie bumps, cambers, etc) make this supremely enjoyable so maybe this reviewer isn't quite the car/driving guy he claims to be.
Well no, he thought tin wasn't a metal, so I'm wondering if its more a case of him being a fucking idiot (sadly it is a condition for which we have no cure)
Many of the bumps hutch himself has said are too big (usually on the threads complaining that cars are too bouncy when driving down the road). But still, thats something that will only improve as he makes more maps and as BeamNG grows.
I fucking love it.
So I can't voice my opinion and say its a revolutionary technology just because it has imperfections?
-It's my opinion and I am entitled to that-
I absolutely love the game too, no question. And yes, you're right, this guy isn't anywhere near as smart as he thinks he is. Problem is, he has a huge big valid point about no FFB, no steering wheel support, and wheels that are way, way worse than any other game out there that claims to be a car or a driving simulator(okay, maybe apart from the actual "driving simulator 2014" or whatever...)... What Hati says is true. I too hope at some point the devs will find a way to create wheels that do not use nodes and beams at all, and just fake everything. As it stands, I'd honestly rather have GTA 4 unrealistic tires than the current BeamNG ones, so we can at least have cars that grip, and don't feel like they grip but actually slide around for 10 cm before they suddenly snap :/
I've suggested to Estama to create wheels with way more width-wise segments(just for those who have a computer fast enough to run it), but I think he's too busy, sadly :/ I'll do some experiments with the mid-engine car I'm working on, see if I can come up with some hacky way to abuse the current wheels...
That isn't what I implied, I'll bullet point it
Yes, the node/beam approach to physics is unique and has potential, the way things can flex, bend and crumple is its main selling point
However, objectively speaking the handling of vehicles doesn't hold up under scrutiny
There's missing support for what are fairly basic things to expect of a simulator these days
Things like wheels use too many resources for what they are. The T75 is this really realistic truck body I can't enjoy properly because there's 200+ physics objects used to simulate its wheels
Its a bit early to scream revolution just yet, its a revolution in realtime softbody physics but not in terms of driving simulators. Yet.
The tire model is very inaccurate in general, think of it this way. A tire in BeamNG has an average of 12 numrays (going by official vehicles). Numrays are basically the number of length wise segments that go along the surface of the wheel, for anyone who doesn't know. This model is constructed the same as all the wheels are and is using the 12 segments like the official vehicles.
The obvious first problem with this is the number of length wise segments, while it is helped by the fact that tires are inherently not completely solid (it would destroy the suspension otherwise) and have significant amount of flex this would be like driving on cobble stones in reality. This problem is very noticeable at low speeds and at around a steady 5-10mph is practically a death wobble. Try stopping the car on the edge of one of the tire numrays and it will be almost impossible to do and if you do the car will roll back and forth like crazy.
A good video demonstration.
There is also the major issue that there are no width wise segments what so ever. This means that there is no flex at all along the tread surface meaning there is no curvature at all on the tire but there is again a similar issue with the side walls. A real tire wall will also bend outwards along in the centre when under force, this is currently impossible with the current tire model as there is again no width wise segments on the side wall. Currently the side walls behave like shock absorbers as they only compress in one direction, this is just very inaccurate.
I'm still unsure on how we could fake a tire model though, something like what LFS uses would be amazing but I'm not sure how that could be created.
Isnt the reason gabester doesnt make higher (or less if it should be, no idea what any of this means) numrays is performance concerns?
Sent from the 3rd galaxy from the talks of tapping
Yes, thats the reason we don't have a much more highly detailed tire model, its all performance issues. Nobody would suggesting faking it otherwise.
It's the reason we don't have something like this from LFS
Also physics problems. That many small beams with those properties along with superlight nodes will cause problems.
Its annoying that the current method is worse and far more resource intensive. If we could appeal to rigid body kinematics by finding a way to stick a rotating rigid body object to a couple of nodes like a bearing and code an LFS style simulation around it then that would be nice. It might even be a non-intensive way of getting wheels to self collide in vehicle wheel wells. You never know. It should really be worth discussing. We can't exactly go on like this and pretend that nodes and beams can do everything. WIth enough computing power we can, but at some point you have to realize the limitations of hardware realistically. As gabester put it, cylinders are resource hogs. The devs probably should hunt down an alternate solution until computers can handle 40 T75's with trailers on a node/beam replica of the golden gate bridge at 60fps.
That may be too much
I sort of agree about the Node and Wheel 'problem'.
Even tough this method is fully based on the Nodes behaviour, it just doesn't bring the wanted results yet. Gabester kept fine tuning wheels like a mad.
If he improves the e.g. handling/grip, the top speed worsen, and etc.
While having those fake wheels, would bring much easier tuning probably, better handling probably (since it will be a proper circle, not a 12-nodes-sided wheel), and etc. I'm not really sure how they would implement things like rim deformation with that, but it should be possible somehow (I don't like this comparison, but look at NCG method).
Oh, most important would be FPS improvements: On the T75 there are 10 wheels (2 front, and 4 duallies on the back if I'm not wrong). I think each wheel is like 64 nodes (32 tire nodes and 32 rims nodes), so there is a total of 640 nodes (along with Beams and Coltris) being simulated for the wheels only (The D15 has 196 nodes for the wheels only). That's maybe the cause of the hard hit on some "not-so-old hardware too", and explain why removing the wheels would pump my (and not only) FPS from 10 to 130.
I think the devs have to choice between keeping a pure Node&Beam simulation, or going for a more accessible (In terms of playability and etc) game/sim.
IMHO, I'd like to see the fake wheels (and see if they work out better than the actual), and I'm not saying that just so I could play with T75 with proper FPS, but because it would open the game to a wider-range of people. I don't think that there are lot of people here really interested that wheels are made of Nodes and Beams, but they want cars to have realistic handling (feel free to take that as bs if I'm wrong)
There are as many nodes in the T75 as there are in the wheels, which means half of the load of the T75 is it's wheels. 660 nodes for the T75, and 640 for it's wheels. Its beyond silly when you think about it.
A little test.
Wheels made of 3 nodes for each sides brought my FPS back to 60FPS stable.
(imported from here)
He could have used the "free tech demo". He didn't have to buy it.....
Remember your CPU is underperforming badly. I can have 2 t75 with the same cpu and still keep 60+ fps.
I can barely handle one.
Your sig says you OCed to 4ghz, that's why the difference. OCing changes a lot in BeamNG. IIRC each vehicle uses it's own core (dunno how it splits after you go past the avalaible number of cores), so increasing the speed of the CPU would drastically change the amount of nodes and beams a single core could simulate in realtime. Again, stay alert for bs warning on this post too.
Sadly non-k version here, so can't really test it.
Hmm well currently there are unusual optimization issues with the game in general causing slow downs on identical setups were one will run fine where as another could have pretty big issues. This happens with all of the vehicles but is very noticable with the T75, its being worked on though.
Gabe's post from the facepunch forum:
We can't just inject magic "good tires" into the physics engine. That's not how anything works. If we could, we would've done it 2 years ago. LFS simulates nothing beyond tires and suspension arms and their tire model has taken what... a decade?