Why couldn’t async physics and networking be implemented using PhysX?
2 Likes
Probabaly the same reason why 64-bit support is not implemented using PhysX.
2 Likes
I downloaded it unzipped it over the default EA version files compiled the engine but its still not working do i need to disable the chaos plugin in the editor?
Here it is a bit of tuning of chaos physics (ragdolls and ponytail).
Cloth feels great. However, maybe I’m a bit blind, but I remember apex cloth collided with physics assets by default?
Also wondering if adding inner pressure model is considered to maintain volume in soft bodies.
Please replace the Vehicle Advance template for something more real, the mini go-kart doesn’t realy help anyone due to the incorrect size of the vehicle, replace it with the simple vehicle template that is in ue4.
ps doesn’t work on my computer , but i can sometimes get the wheels to steer right but nothing else
2 Likes
Yeah, it’s about time to update the template, no?
I would appreciate if you can show more sophisticated types of vehicles, such as tracked vehicles with many wheels, more realistic settings so that we can call it Advanced Vehicles. Chaos Vehicles has negative reactions associated with it currently and you will really need prove it to loose it. My two cents. ^^
2 Likes
Great information, thank you! I had created a basketball benchmark with PhysX and was able to get 2000 basketballs before slowing down but was only able to get about 600 with Chaos Physics.
Looking forward to all the new amazing things that are coming with UE5!
https://drive.google.com/drive/folders/1XTaibVEGMe_w0eA8S6sDDbQoEGbMTexm
4 Likes
What is the possibility of getting new or extended trace shapes with chaos? Seems like a great opportunity to look into them. Oriented capsule (very useful for a wide variety of things; I’m shocked this isn’t already standard in both CPP and BP), frustum or cone, closed cylinder, line trace with thickness (same as closed cylinder, but no rotation if that proves too difficult to implement). I know you could use a collision mesh in the shape you want, but it hardly has the utility or use-cases of actual traces.
The dream would of course be more customization options to the base Character Movement component along with the above changes (any gravity vector, better physics and moving object interaction out of the box (expanding on this: being pushed around by moving objects anywhere but from the bottom requires custom workarounds to not be glitchy/janky looking), etc), but I’ll not be too greedy.
3 Likes
There’s still similar problems to what I was having with 4.26 Chaos. Large amounts of objects are either sleeping very slowly or not at all if there’s enough of them, especially if they’re on each other. Everything gets stuck in a permanent state of jittering with 100 dice in a box.
Someone PM’d me before in October saying they worked with Epic and they were never seen again after.
Thanks for the opportunity to post feedback. From my tests, I noticed very notable instability in multi-rigid-body interactions.
Bricks stability test:
Chaos: Imgur: The magic of the Internet (sped up)
vs PhysX: ShareX GIF | Gfycat
Rolling sphere into bricks test:
Chaos: ShareX GIF | Gfycat
vs PhysX: ShareX GIF | Gfycat
As I’m developing a physics-based project, these results were concerning, although I completely understand if Chaos is still underdeveloped. Is Chaos currently meant to support this usecase? Can we expect stable results before the switch is made?
6 Likes
Chaos should be meant to support every use case before even considering to deprecate PhysX as a replacement for a general purpose physics engine. These videos are completely laughable. Can’t understand what epic is thinking presenting this broken and slow physics engine as a serious alternative and somehow expecting people to be happy about it.
By any chance, did you measure performance during those tests?
5 Likes
For what it’s worth, Chaos is in a bad state now, but it’s a nessecary shift. PhysX is not a good solver - it’s everywhere because it’s free, not because it’s good. Determinism, more advanced time step options, async, and networking are all things that physx is bad at, even in later versions of the API.
Improving the UE4 implementation of physx would be polishing a t*rd that is better bagged and binned sooner rather than later. Getting rid of it is an excellent and much appreciated move - bring on the death of PhysX.
The biggest problem imo is the messaging - the way epic talk about chaos in marketing materials, you’d assume it’s much further along than it is. The number of people I’ve spoken to who, by no fault of their own, don’t realise that chaos is nowhere near usable let alone production ready is quite immense. In reality it’s untennable to ship physics driven gameplay with it today, and (IMO) probably will be for quite some time. Based on my testing, I can see a world where it’s not ready for primetime by 5.0. Setting some more reasonable expectations on this front might be worthwhile.
5 Likes
Or for gods sake.
Show me all those games made with unreal where you spawn 600 balls in single place ?
That’s purely YouTube measuring contest, not an common game development use case.
Chaos is optimized for what is commonly used in games and more importantly for what is used in Fortnite. Scene queries, Cloth, Vehicles.
It will also first support use cases that are actually useful (like networked physics) or 64 bit precision.
2 Likes
You say this when Epic’s showcase of it was more severe than just 600 balls. Stability, visual or otherwise, is currently struggling with <50 objects.
It’s supposed to be at parity or better than the regularly sim-exploding PhysX.
1 Like
Hi,
Chaos is meant to replace the old Physics, so, it is in development and UE5 is a Early access.
Being patient and testing and keep going the post and report and all useful thing for them is our contribution part.
I am sure Chaos will evolve and come in place at right time.
About Chaos Vehicles, i second Geodav and ChrisK post here, it will be awesome to replace the Vehicle template for UE5 with a refreshed template, more realistic, various multi wheels, (motorcycle too ?).
And/or a sample project to learn from, like the Valley of the Ancient.
And/or more integration tech Fortnite style, like the Water system.
Etc… Anything to propulse us in line with the Future of Chaos.
Thx.
3 Likes
While I understand that Chaos is currently focusing on a subset of usecases, if I as a developer want to spawn 600 balls in one place in my project, I should be able to.
2 Likes
Except it’s also much slower than PhysX in scene queries, both updating the scene and querying it. Fun, right? Besides, a fortnite specialized thing is not acceptable as a general purpose physics engine, which is what a general purpose game engine should provide.
Physics replication could be made with PhysX just fine if only they tried. But they’d rather pretend it’s impossible to modify PhysX. The only point you have here is the 64 bit support - though I wonder if PhysX 5.1 will do something there too…
2 Likes
Nobody’s claiming PhysX is the ultimate perfect physics engine that can never be beaten. However, it’s light years ahead of the current state of chaos. Thus when epic is actually considering to replace it without having proven they can even get anywhere near it at all, and enabling chaos by default in UE5 while it’s not even usable, people are expectedly going to be really annoyed.
Chaos is completely raw. Get it out of the UE5 preview and put it back in the oven for another couple years.
3 Likes
I’m sure epic is trying to build a system that works for all there needs, but really why not keep Nvidia PhysX also and update it to the latest version as a plugin, I find chaos’s workflow to be for simple things like destructible meshes for rocks overs overkill.
2 Likes
It’s a benchmark. In a real game, there would also gameplay happening alongside the simulation. In his test, the only thing happening was the simulation of the balls. In a real game, for 600 balls in chaos, your ball budget would be lower than 600 (e.g. 200, 1/3rd of 600), whereas for 2000 balls in physx, your ball budget could reach 600 (little more than 1/3rd of 2000) or even higher.
1 Like