Physx planetary destruction?

I was wondering if when planets explode, or are destroyed, that they will fracture and fragment into peaces such as this image? http://www.cgcookie.com/assets/tutorial ... ete-12.jpg

And this image http://www.superbwallpapers.com/fantasy ... net-10684/

finally: http://d2f8dzk2mhcqts.cloudfront.net/Pr ... _large.jpg

I think that would be such a cool feature to add, the debris could stay persistant and either help or hinder you by catching the pull of other planets and crashing to the surface or something else, so perhaps when destroying planets, you will careful how and where you do as to not end up getting bombarded by your targets debris!

What are the chances of something like this?

If they decided to do something like this, I REALLY FREAKING HOPE they use anything BUT Physx. I really detest that proprietary junk. Something based on OpenCL, like bullet.

Just wanted to vent that. Tired of being locked out of Physx because of the graphics card brand I use, not because of any inherent inability of my hardware to do it.

if they decided to use physx.. you should get mad at Nvidia for making a better product than Uber for implementing it.

while i do wish Nvidia would open up physx to AMD cards, i can see why they dont, and the changes are large enough that they factor in my decisions when buying graphics cards. if you dont think they make a big difference, play Batman Akham asylum both with and without it, it really is AWSOME!

This isnt meant to a physx bash or bum, just an IDEA to, for now forgetting the "physx" tech, of planetary physics shattering into huge interactable fragments, like my images suggest, i only mentioned physx as a possible feature, im sure other physics engines could achieve such sexy magic.

I asked about GPGPU stuff earlier and

I'm against Physx myself. You can implement the same thing without making it Nvidia only, and that would definitely make me happy.

If they do add it, they better at least do what Mafia II did, which was allow ATI users full use of Physx through emulation.

This a VERY good argument. Lets keep the gamelogic on the cpu.

I highly doubt the physical accuracy of those images anyway... you might get those kind of fracturing when you accelerate two marble spheres into each other, but not planets and other such objects...

Who cares? like neutrino said, there not going for realism, there going for awesome!

Like i said, no turning this into a physx bash/bum thread, it was just as example of a tech that is widely known to be capable of it, like i said im asking about the possibility within the physics used in the engine the guys are using.

I'd like to give my input here; I don't think it's even POSSIBLE with PhysX. This game is dealing with AU as units. You'd get major physics accuracy issues after only a few km, and space is much bigger, even when scaled down. A floating origin system couldn't work very well due to there being hundred of units, so there's no one spot where you can re-position to keep accuracy great... I don't even think Bullet could manage it without same modifications, which happily is possible due to it being 100% open source.

If this happens, I'm going to get really pissed off.

Easy way spitting on AMD users.

In this case - for me myself at least - only physical accuracy could be awesome... not being accurate would just look stupid.

Remember, they are shooting for awesomeness, not accuracy.

Besides, if I were them, I would implement the game with several different scenes that seamlessly transition between each other. In other words, the scene representing the planet is different from the scene representing the solar system. The model used to display a planet would be representative of the state of the planetary scene.

So with separate scene graphs, you can use any scale you like. They could have the solar system in 1M scale, with planets 10M wide, and nobody would know without a reference point. What I mean is, scale in computer graphics is completely arbitrary.

There also will likely not be nearly the same amount of space between planets and asteroids as there would be in real life.

Yeah, I know you didn't want it to be a physx bash/burn thread. But I really stress that I seriously disapprove of using physx as it stands right now.

However, I agree that a server probably wouldn't have a nice gpu and therefore gpu-powered game-changing physics engines shouldn't be used. But if they used gpu physics for special effects client-side, why use the option that locks out a potential third of your customers?

We read a lot of dynamical destructionable battlefields these days and now physical simulation of the destruction of a planet?

Of course all sounds awsome, but I wonder how a little company with little budget could ever create sutch stuff just a few others with more budget tried less then more successful in the past?

The question for me is how they ever will get this synced over the network in online games and how the hell it ever could be balanced.

Even if they really can bring the technology, with sutch massive dynamic aspects evrything becomes mutch more complex, like pathfinding, line of sights, up to complete possible gamestoppers.

On my personal view, they already have massive work to do just to make the "multi map" concept playable. Stuff like 360° maps with orbits of other 360° maps is insane enough. Cause no one did it, you have to imagine evrything yourself and can't rely to things that worked in the past.

I backed this game cause I m a huge fan of SupremeCommander. For me there was never a better RTS for online play and I hope to find a successor in this that make me forget SupCom2.
I just hope they don't forget to make an excellent RTS besides all the "awsome planetary destruction stuff". Sometimes I'm a bit worried cause it always seems to be in the focus and not the RTS itself.

I'm talking about literal physical accuracy. As in, a floating point number only has so many digits, and needs to start rounding up sooner or later. This gives severe jittering issues.

That could be rather difficult to code, especially when you consider being able to see explosions from orbit on the planet's surface and the like. Also, interplanetary rockets and the such would feel major jittering issues, as that's relative to space. It doesn't matter if you have tiny model sizes (which will have less accuracy due to trying to move stuff tiny amounts) or massive ones (which will lose accuracy quickly after a distance), it'll still happen.

This is true, but completely irrelevant. As I said, floating point accuracy is independent of scale; for example, I remember a billiards game where the balls were represented as multiple meter sizes in the physics engine (which was Bullet) due to smaller sizes not having enough accuracy.

Again, true, and again, irrelevant due to many reasons shown above, and the fact the game will probably include a solar system and planetary editor.

This problem isn't even just limited to physics engines. It's a major computer problem everywhere. Some older RTS games can actually have your funds in the game roll over due to hitting the upper limit of an integer. I'm not sure how the guys at Uber are going to get around this, whether it uses some automated system to create origins at relevant places, or if they use arbitrary sized values to allow for a theoretical infinite amount of detail.