How Do’s Physics Compare To Reality?

I like careful, reflective journeys in as much as Alec does, but my other great vehicular passion is tossing cars down mountains and slamming them into brick walls. There’s no greater game for that than [official site], which although still in alpha has the best physics and damage model of any driving game I’ve ever played. To prove the point, YouTube channel DragCarTV recreated common crash testing scenarios in the game and then put the footage side by side with real crashes to see how they matched up.

As impressive as I find the modelling in BeamNG, I’m surprised by how similar some of these accidents are. Impressive!

The description under the video explains how the video was made. The clips of the actual crashes come from the IIHS (Insurance Institute for Highway Safety), DTC (dynamictestcenter) and the TV shows Mythbusters, Fifth Gear and Top Gear. The game footage was made by taking advantage of BeamNG’s thriving mod community, who share cars and tracks on a regular basis through the game’s forums. There are links to the mods used, including vehicles, tracks and crash test dummy models through on the YouTube video above.


  1. GameCat says:

    Meanwhile in my Unity car game you can drive up almost 90 degree slope and it will launch you a few meters in the air.
    What am I doing wrong?

    This Beam is really impressive.

    • Harlander says:

      You could work hard on making your car game more physically realistic.

      Or you could parlay cars getting wicked sick air into a gameplay mechanic.

      The choice is yours.

    • TheMightyEthan says:

      Isn’t a 90 degree slope more commonly known as a “wall”? ;-)

  2. bit.bat says:

    It would be great if their technology could be licensed to other developers. A GTA game with this type of damage modelling would be great!

    • Chiron says:

      Spintyres+BeamNG is basically my dream car game at this point

    • jabbywocky says:

      The dev team has said after the standalone game is complete their overall goal is to be able to license the tech to other game developers. I know early on Star Citizen had BeamNG demonstrate the tech as applied to their ships but the processing power demands meant it wasn’t feasible especially for multiplayer, which is a gap yet to be crossed by

      • WaytoomanyUIDs says:

        IIRC they only started making the game because everyone ooohed and aaahed over their soft body physics but never got around to licencing it.

    • king0zymandias says:

      I don’t think licencing the technology will really help very much, because first of all, what they are doing isn’t technically all that complicated, and secondly it’s not the sort of plug-and-play solution that can just be applied to any game that hasn’t be designed around it. Which means you actually have to model each and every car while keeping in mind that you want them to shatter into individual components later. So the process actually starts very early in production, right when modelling starts.

      Additionally just like any destruction models used in the world of non-realtime VFX all they are doing is modeling different parts of the car as separate components with hinges at the right places. And then when collision is triggered a rigid body or soft body simulation is initiated based on which component is colliding. In VFX in addition to this you usually also have things like dynamic voronoi shattering, fluid simulation, and many more additional pieces inside that shatter all over the place to make it even more convincing.

      So my point is that it’s not some great secret technique that they have stumbled upon, it’s just that developers haven’t had the luxury of implementing proper soft body simulations in real time interactive media on account of performance issues. However, now that hardware is getting more and more powerful you’ll see more of this. Although all the algorithm and fundamentals of the technology required has been developed many many years ago.

      • cafeoh says:

        I entirely agree that this isn’t usable in most games, and can’t just be plug-and-played like Rockstar did with Euphoria for example. If only because the driving mechanics would need to cohabit with the soft-body simulation, and driving in Beam.NG is an absolute mess, as is to be expected.

        Although saying that “we already have the algorithm” is extremely reductive. The amount of work they put in soft body simulation is clearly amazing. Yes, maybe we already had the algorithms, just like a lot of the most clever algorithmic and data structure designs we use today were created before they even made sense outside of discrete mathematics.

        I don’t care if Pixar can render amazing fluid, hair and soft-body simulation on their rendering farms, if a small developer makes even a small breakthrough in real-time simulation/rendering, he deserves all the credit.

        If Crytek devs only wanted to implement existing techs we’d never see them at SIGGRAPH, and even if everybody knew what ambient occlusion was, we still might not have SSAO in every single game and 3D game engine to this day.

        • king0zymandias says:

          Fair enough. I agree with what you are saying, they definitely deserve due credit. Although I still stand by my sentiment that compared to the huge amount of work that goes into creating and maintaining a 3D game engine, and creating all the fairly complex gameplay mechanics in any modern high end game, implementing a real-time soft body simulation that performs well on high end machines isn’t all that difficult nowadays.

      • estama says:

        If something can be done offline, this is no indication that it is easy to be done real-time (notice how much we are not drowning in realtime raytraced games).

        BeamNG’s physics core, contrary to your explanation, always runs in soft body mode. There is no rigid body subsystem in there. To be able to achieve it, the physics core always runs at 2Khz. When you go into slow motion, you just see more of the physics frames (which are not shown in real-time).

        There is a huge difference between VFX engines and simulation engines. VFX physics engines have to look good. BeamNG’s physics engine has to simulate correctly. Think of the difference between SSAO and radiocity methods. One is for looks and the other calculates the real thing. SSAO is light to compute, radiocity is extremely heavy.

        Concerning the preexisting algorithms argument of yours. If the existing algorithms were enough, then i would not have had to create a faster selection algorithm:

        link to

        (above is just one example of what is going inside BeamNG’s physics core)

        Concerning your comparison with 3d game engines. Game engines have to run at 60-100fps. In the time that a game engine that runs at 100fps has finished one single frame, BeamNG’s physics core has to finish 20 physics frames (this is what 2Khz means). Each one of these scheduling all the different CPU threads which calculate all the parts (nodes/beams), friction, aerodynamics, hydrodynamics, buoyancy, torque, collisions with other vehicles, collisions with environment, vehicle’s engine, ABS, ESC…

        Finally, to understand the difference between offline and online at 2khz. Imagine some real life action that everyone has been doing slowly. Like walking up the stairs. This is the offline case. Imagine now that someone invents the elevator. This is the online case. Just because we knew how to walk up the stairs does not mean that the elevator concept was simple to invent…

        • TacticalNuclearPenguin says:

          Surfing is easy and it has been done before, you just have to stand on a board.

        • Rumpelstiltskin says:

          “BeamNG’s physics core, contrary to your explanation, always runs in soft body mode. There is no rigid body subsystem in there.”

          Do you think it’s a good thing though? Softbody sim is really only useful during crashes, and running it always, at a whopping 2kHz, doesn’t achieve anything apart from making the driving wonkier.

        • MOONRGR says:

          Impressive explanation

  3. Pizzzahut says:

    I was content with the physics from Destruction Derby in 1996. Everything else has just been a bonus.

  4. liquidsoap89 says:

    It’s odd that they had to increase the speed of the cars in the simulation to match the destruction of the cars in the real world.

    • Cyroch says:

      Is it possible you just misread? The speeds in the bottom are not a comparison. Left is mph, right kph.

  5. Unsheep says:

    You can even get the game DRM-free from their website, which is very nice of them.

    The game itself looks to be more of physics puzzle game than a racing game though, like Trials and Spintires. I’d like to see more of the actual content, the different game modes. So far the game looks quite superficial, like Goat Simulator or I Am Bread, but the game is still in development so we’ll see.

    In general I think mainstream gamer’s obsession with the crashing spectacle in racing games is quite ridiculous; if you crash in a sim game, or even a sim-arcade like Dirt or F1, you are playing very badly. Even in arcade games like GTA crashing is something to be avoided unless you want to restart the mission.

    I think mainstream gamer’s obsession with crashing effects is an excuse for being bad at racing games or simply too lazy to master the physics.

    • TacticalNuclearPenguin says:

      That is a perfectly possible explanation, and indeed i wouldn’t care in the slightest if Assetto Corsa didn’t even bother to implement visual damage effects.

      This is a different thing though, the point of this game is to observe a realistic behaviour and enjoying that experience.

    • Capt. Bumchum McMerryweather says:

      Or maybe people want the immersion to extend further into the game, so when you DO crash (a la Gran Truismo) you don’t just bounce off the wall like a fucking tonka toy.

    • LennyLeonardo says:

      Yeah, but it sure is fun watching your competitors crash. In games.

    • SVW says:

      Can we please stop using laziness as an argument about gaming?

      • Harlander says:

        It’s weird, isn’t it?

        Of course I’m a lazy gamer. If I wasn’t lazy, I’d be doing something useful.

  6. EvaUnit02 says:

    They should now do a “how does the glorified tech demo compare to actual games?” comparison.