Newton vs Bullet demo

[Richard Simpson]

Wow, that speed increase is impressive.

[Canardia]

Actually Bullet is much faster, I just noticed a huge inefficiency in my Bullet class

Now it positions and rotates all cubes in each frame, but it only needs to do that when the physics body has moved.

I tried to get Bullet deterministic by removing all srand() functions and making all rand() calls to return 0 always, but it didn't work (it acts just like before, works fine, but is not deterministic), so that's something where Newton still wins.

 

The Bullet lead developer, Erwin Coumans, lives also in California, just around the corner where Josh lives, so maybe he should have a visit some day

[Richard Simpson]

Actually Bullet is much faster, I just noticed a huge inefficiency in my Bullet class

Now it positions and rotates all cubes in each frame, but it only needs to do that when the physics body has moved.

I tried to get Bullet deterministic by removing all srand() functions and making all rand() calls to return 0 always, but it didn't work (it acts just like before, works fine, but is not deterministic), so that's something where Newton still wins.

 

The Bullet lead developer, Erwin Coumans, lives also in California, just around the corner where Josh lives, so maybe he should have a visit some day

 

 

Ooh. I done a quick test, and with 500 newton boxes, I get 4fps. To get 4fps with bullet I can have 5000 boxes. I think the speed gain and the amount of objects you can have at the same time wins over the deterministic issue. But I cant see myself making online game xD. It would be great if there could be an option within the editor to change which physics engine is used.

[Canardia]

By the time LE3 comes out, I think Bullet will be quite a close match to Newton in accuracy. It already wins in many other features, but I want to see that it wins in all features before I can clearly say that Bullet is better than Newton. It would be kinda silly to have 2 physics engines in LE3, especially when you have the full source code of the physics engine, so it's your own fault if you can't unify them.

[Newton Dynamics]

On 11/9/2010 at 1:39 PM, Canardian said:

By the time LE3 comes out, I think Bullet will be quite a close match to Newton in accuracy. It already wins in many other features, but I want to see that it wins in all features before I can clearly say that Bullet is better than Newton. It would be kinda silly to have 2 physics engines in LE3, especially when you have the full source code of the physics engine, so it's your own fault if you can't unify them.

How many games has you already released doing 5000 rigid bodies? 

Plus what you are saying there is simple not true. There are tools that compare Newton, Physx and Bullet side by side doing the same task, you know. 

when you do that kind of physic that  bullet does yes you can move tousands of objects, but physics is a lot more that that.

At some point you want your physic simulation to do more that just boxes colliding and repealing each other like they are pile of goo without consistency.  You would probably want to control your object over a large range of parameters and that's something that neither Bullet, Havok or Physx can do. 

You need a commercial quality simulator like vortex to get the quality that newton provide at real time. 

 

[Newton Dynamics]

well, seven years ago he made very strong statements. I was curios as to what he accomplished with 5000 boxes seven years later. 

[macklebee]

Canardian/megatron/mika/cant-remember-previous-name was well known for making "strong statements" about everything that typically resembled what people nowadays call "alternative facts". So don't take anything posted by him to heart.