Home | Wiki | Discord | Dev Stream | YouTube | Archived Forums | Contact

Undriveable car at high speed on BeamNG


Some testing…

Exhibit 1 - Open diff in automation:

		["shaft",            "rearDriveshaft",      "gearbox", 1 ],
	["differential", "rearDiff", "rearDriveshaft", 1, {"diffType":"open", "gearRatio":3.78, "diffTorqueSplit":0.50}],

According to what the beamng site says, it is an open diff with a gear ratio of 3.78 and an equal initial torque split.

Exhibit 2 - Geared

		["shaft",            "rearDriveshaft",      "gearbox", 1 ],
	["differential", "rearDiff", "rearDriveshaft", 1, {"diffType":"lsd", "gearRatio":3.78, "diffTorqueSplit":0.50}],

Same as before, 3.78 gear ration but this time type is LSD. However there are some extra parameters, BUT those are commented in the code (the // before each line) meaning they do not have any effect.

        //lsd specific
    //"lsdPreload":50, //Nm of preload torque. Clutch diffs 50-250, Torsen around 0
    //lsdLockCoef:0.2, //On throttle locking ramp. O.1-0.2 for clutch type, 0.3-0.4 is more like torsen
    //lsdRevLockCoef:0.2, //=lsdLockCoef by default. O.1-0.2 for clutch type, 0.3-0.4 is more like torsen

    //viscous specific
    //"viscousCoef":5, //Viscous locking stiffness in (Nm/rad/s)
    //"viscousTorque":5, //=viscousCoef * 10 by default. Max torque it can transmit from side to side

Exhibit 3 - Viscous

		["shaft",            "rearDriveshaft",      "gearbox", 1 ],
	["differential", "rearDiff", "rearDriveshaft", 1, {"diffType":"viscous", "gearRatio":3.78, "diffTorqueSplit":0.50}],

Type is set to Viscous. Again, the other parameters are commented in the code ( // ) and have no effect…

        //lsd specific
    //"lsdPreload":50, //Nm of preload torque. Clutch diffs 50-250, Torsen around 0
    //lsdLockCoef:0.2, //On throttle locking ramp. O.1-0.2 for clutch type, 0.3-0.4 is more like torsen
    //lsdRevLockCoef:0.2, //=lsdLockCoef by default. O.1-0.2 for clutch type, 0.3-0.4 is more like torsen

    //viscous specific
    //"viscousCoef":5, //Viscous locking stiffness in (Nm/rad/s)
    //"viscousTorque":5, //=viscousCoef * 10 by default. Max torque it can transmit from side to side

Exhibit 4 - Electric Diff

It is the same case as the geared diff. Difftype is set to LSD and the rest of parameters are commented.

		["shaft",            "rearDriveshaft",      "gearbox", 1 ],
	["differential", "rearDiff", "rearDriveshaft", 1, {"diffType":"lsd", "gearRatio":3.78, "diffTorqueSplit":0.50}],
    //needed values to tune diff behavior
    //locked specific - nothing needed, auto calc now

    //lsd specific
    //"lsdPreload":50, //Nm of preload torque. Clutch diffs 50-250, Torsen around 0
    //lsdLockCoef:0.2, //On throttle locking ramp. O.1-0.2 for clutch type, 0.3-0.4 is more like torsen
    //lsdRevLockCoef:0.2, //=lsdLockCoef by default. O.1-0.2 for clutch type, 0.3-0.4 is more like torsen

    //viscous specific
    //"viscousCoef":5, //Viscous locking stiffness in (Nm/rad/s)
    //"viscousTorque":5, //=viscousCoef * 10 by default. Max torque it can transmit from side to side

EXTRA: autolocking diff

		["shaft",            "rearDriveshaft",      "gearbox", 1 ],
	["differential", "rearDiff", "rearDriveshaft", 1, {"diffType": ["open","locked"] , "gearRatio":3.78, "diffTorqueSplit":0.50}],

The rest is commented.


There is a difference between open, viscous, geared and auto locker, but electric seems to have the same effect as geared.

All the extra parameters are deactivated/bypassed, and it seems to do some auto calculations based only on the diffType.
I’m curious tho, and I will definitely will uncomment and change the extra parameters from the differentials and see how it changes.

From the beamng site:

Types (diffType)
The differential device has several types available, to simulate the different common differential types in the real world.

  • “open” differentials distribute nearly equal torque (except for small frictional effects) to each output, while allowing a difference in outputs speeds. This type can be used to replicate the standard differential on a car.
  • “lsd” differentials induce cross torque between outputs dependent on how much input torque is applied. This type can be used to replicate “clutch limited slip” or “gear limited slip” differentials.
  • “viscous” differentials induce a cross torque between outputs dependent on the relative speeds of the outputs. This type can be used to replicate “viscous limited slip” differentials
  • “locked” differentials enforce that both outputs maintain the same angle, leading to complete bias of torque from one side to the other depending on traction level. In reality it is not really working as a differential at all but as a very stiff spring between the two outputs. This type can be used to simulate “spool” or “welded” differentials.



Yes - I had this exact same issue.

I believe this was the same problem I experienced when beta testing and I’m not sure what causes it. At approximately 140 mph, even when accelerating slowly and in a perfectly straight line, all the weight suddenly shifts to the right, causing the car to steer left and lose control.

It is definitely a bug, because while getting up to that speed, I can see the weight increasing on the front end (consistent with the all the front downforce I applied, via the lip), right up to the moment all the weight shifts to the right.

ME-12 - Turbo-Spec.car (18.1 KB)

Since at 140 mph, according to Automation the car is producing 202 lbs of downforce at the front, and 97 lbs of downforce at the rear, I’m going to try turning down the downforce to where it has just under ~200 lbs of downforce nearer to its top speed and report back as whether or not that’ll make the car behave itself. (stay tuned)

edit: ME-12 - Modified Wings.car (18.1 KB)

OK - so I feel like I’ve only partially solved the mystery of the undriveable car, but it seems to be getting better. By keeping the downforce (in automation) just below 200 lbs front and rear, the car no longer exhibits the tendency to suddenly shift its weight to one side. So provided you’re considerate with the throttle, the car’s top speed can be achieved without losing control. Which leads me to my next point…

While it makes sense that getting aggressive on the throttle in a 1,000 horsepower mid-engine supercar can make it tail-happy, there seems to be fairly extreme weight transfer to the rear (expected), but again, at lower speeds the weight shifts to the right side of the car (not expected) even when the front tires are set perfectly straight. I’m curious as to whether or not it has anything to do with the way the differentials are calculated (or not calculated), but as it’s waaaaaaaaay past my bedtime (4am NY time) and I gotta get up for work tomorrow, I’ll let someone else investigate.

TL;DR = Keep automation downforce below 200 lbs* (front and rear) to eliminate weight shift bug. Someone needs to figure out what the story is with differentials to see if that’s affecting driveability as well.

  • NOTE: 200 lbs might not be a hard limit. It’ll probably vary from car to car. So since this is anecdotal, just keep an eye on the corner weights while accelerating slowly, until you see the weight shift to one side. Note the speed in BeamNG when it goes haywire, and go back to automation and check the downforce graph to see how much weight is being applied to the car at that speed. Then reduce the downforce so that at its top speed, you don’t exceed that limit.


What really got me was that even in my cars with stupid downforce, sometimes the weight thing would happen… And sometimes it wouldn’t!


Front tire size is key. Using Automation’s “recommended tire size” (shown when you mouse over the wheel/tire) is a good place to start. Generally speaking, lightweight cars and especially mid/rear engine cars are pretty sensitive to this at higher speeds.
@MasterDoggo, I was able to get the ultralight pretty drivable by simply dropping tire widths to 175 front, 225 rear. The Crustacean became much more drivable by dropping front tire width to 250 front (though it still picks up a weird vibration at 200ish, which dissipates, then returns at 220ish, then smooths out again - I saw 247mph before I ran out of map. Whee!).
@Slim_Jim, I think you’re on the right track with aero. I usually try to replicate weight balance with downforce, or bias it from there slightly to the rear. I ran yours at 398lb front @214mph, 629.3lb rear and bumped the front tire width down to 235 (it may benefit from less, or more rear tire via flares, I’ll leave that to you). It still moves around a bit, but at least you can steer through it now and it isn’t just diving left. This one also picks up some weird high speed vibrations.
Just spitballing here, but I wonder if this is what’s going on?
Round tires are not round
(link to BeamNG forum post)
tl;dr tires are a bunch of nodes, not actually round. I could see this turning into a high speed oscillation at certain speeds that transmits to the chassis. Basically, a chassis acts as an undamped, very stiff spring which receives inputs from the tires via the suspension. Supporting this hypothesis is the fact that you can change the “dive speed” via shock, spring, sidewall height, downforce. Also, lightweight cars seem way more sensitive about unsprung mass. Disproving the thesis, messing with the chassis quality slider seems to have no effect (which it would were my hypothesis true).
shrug Hope this helps.


Adjusting the downforce to more closely match the weight distribution made a huge difference and definitely allows for higher numbers. Big thanks @Obfuscious for pointing that out. I think I had the same problem when beta testing only the downforce was at the opposite extreme (too much rear downforce, and not enough front downforce - relative the car’s weight distribution).

A wider rear tire, AWD lol, and tweaking the gearing also helped with stability.
ME-12 - AWD.car (18.5 KB)

That is interesting. :thinking: I’m wondering if it’s related to the frame rate bug that’s affecting the turbos.

I tried this on one of my test cars, and it “crashed” the game.


The problem is not only at the high speed. It´s very difficult to make a driveable realistic car with reardrive. A series car with maybe 150hp have no superwide reartires or different tires in front and rear.

When i think about my Opel Senator from 1982, it has 180hp, 3gear automatic, 195R14 tires and reardrive was very driveable. Also my Senator from 1990 with 204hp, 4gear automaitc, locking differential and 205/65R15 tires was only oversteering when you realy provocate it or the streets are slippery.

So i think the exporter needs a little finetuning in this area…


Here we have different experiences. The Automation cars in Beam are actually quite forgiving. I haven’t encountered any undrivable car at normal speeds.

When you really push them, sure, but those old rear wheel drives would go round as well when pushed and not driven expertly. Know when to not apply full throttle, don’t steer too suddenly, don’t forget to heel-toe brake, let the car become stable before putting the throttle all the way down.

You make it sound like you expect 80s and 90s rear wheel drive cars to be point and squirt. They weren’t.


I agree with Private_Miros. The only problems I have enocunterd is with +800hp cars where the front might lift up (plus weird stuff with very stiff suspension settings). I usually make old cars with no traction aids and you have to be very sensitive with driving.


It helps if you play with a wheel and pedals rather than a controller.