Huge Disappointment

[seanjenkinsjr]

lol tolad

[PopcornLadder]

You should really wait until US2 is actually complete to compare the two versions of Universe Sandbox.
[/quote]
I already said I realize it's an early alpha.

[PopcornLadder]

Tartar sauce. Messed up the quote.

[DiamondMiner10]

Nyaaaaauuuuuughhh!!!!!!

And here we see the elusive Xriqxa, active once more after a long hibernation.

[Greenleaf]

-The only thing improved was the textures
-On the other hand, the explosions are overdone
-everything looks very hard to find

Not really sure how to respond to your first point... new physics engine with improved performance, adaptive integration giving user controllable accuracy, collision effects (which were not there before), composition of planets and stars, stellar evolution, vastly improved graphics, various numerical integration modes and much more.


As to your second point. The collisions (their effect) are largely based on this paper https://www.purdue.edu/impactearth/Content/pdf/Documentation.pdf and is not just picked out of thin air in a gaming studio. Also remember that you don't generally see "pebbles" in the simulation, but kilometer large fragments.


And finally, perhaps everything is easier to find in the old version because that is what you are used to? Whenever an old piece of software gets a big UI redesign, the old users are often initially unhappy. I still remember how much of a pain it was to change from DOS to windows ;-)


I hope you will give US² a second look and be open to the changes made. Perhaps you will change your mind... perhaps you will not, but then at least you seem very happy with the old version :-)

[PopcornLadder]

I hope you will give US² a second look and be open to the changes made. Perhaps you will change your mind...

I might give it a second look, if they give the US1 community a significant discount. I mean, although it's improved, why buy it if you have the first one? Unless it's really cheap, I'm not paying for crap.

[tepidbread]

The game is not even finished yet... In fact it seems to be in the middle of a massive physics rewrite. Most of the integrators have been taken out of the game to be reworked. I can't say this for certain, but I am pretty sure that Universe Sandbox 1 only has Euler. (which is pretty bad). Also Gpu computation is in the process of being re-added to the game. We are about to see a massive improvement of accuracy and performance.

[Greenleaf]

The game is even finished yet... In fact it seems to be in the middle of a massive physics rewrite. Most of the integrators have been taken out of the game to be reworked. I can't say this for certain, but I am pretty sure that Universe Sandbox 1 only has Euler. (which is pretty bad). Also Gpu computation is in the process of being re-added to the game. We are about to see a massive improvement of accuracy and performance.

That is true. I still have a bunch of integrators to plug back in after this rewrite. Most users will generally use verlet which is fast and rather stable, but the others are quite interesting as well. Both to get very high accuracy and simply to compare their different performances for educational purposes.


US1 had what is called Euler and which was supposedly explicit Euler but was in reality a variation called semi implicit Euler. Much more stable than explicit Euler but not more accurate, and yes, pretty bad.

[tepidbread]

Most users will generally use verlet which is fast and rather stable, but the others are quite interesting as well. Both to get very high accuracy and simply to compare their different performances for educational purposes.

Will their ever be an option to change the default integrator from Velocity Verlet to something else?
 I am personally a fan of RK4.
Sorry If I am off topic. It seemed like a good opportunity to ask about integrators.

[Greenleaf]

Will their ever be an option to change the default integrator from Velocity Verlet to something else?
 I am personally a fan of RK4.
Sorry If I am off topic. It seemed like a good opportunity to ask about integrators.

Well, it is a topic which it is nice to be "off", so that is ok :-)
You will certainly get back the option of changing to rk4 again, if that is your desire. If it will be possible to change the default,t hat is less clear. Rk4 is not really "better" for planetary integration. About 4 times slower, two orders "better", than verlet, but unstable, since it is loosing energy.

[tepidbread]

Rk4 is not really "better" for planetary integration. About 4 times slower, two orders "better", than verlet, but unstable, since it is loosing energy.

That is interesting. I always have heard that Rk4 is more accurate. However, that may have been a misconception because it does not allow for conservation of energy. This is not my area of expertise. I am studying to become and electrical engineer. 

Thanks for the information though. I may try to revert back to alpha 14 in order to test this out. (I have never bothered to try it before)

Hopefully I can learn something interesting.

[Dante Kieth Bell]

"On the other hand, the explosions are overdone and basically light up the universe if one pebble hits Jupiter."

If your on earth and drop a pebble, Do you see a blinding explosion light up the universe? no you dont. You just hear a click and some dust would fly.

[Greenleaf]

"On the other hand, the explosions are overdone and basically light up the universe if one pebble hits Jupiter."

If your on earth and drop a pebble, Do you see a blinding explosion light up the universe? no you dont. You just hear a click and some dust would fly.

If you drop it at 10 km/s I am pretty sure you will feel like you lit up the universe ;-)

[Greenleaf]

That is interesting. I always have heard that Rk4 is more accurate. However, that may have been a misconception because it does not allow for conservation of energy.

The question is in part "more accurate than what?" and in part "more accurate for what?"
For a linear system, explicit euler, verlet and rk4 are both perfectly accurate.


Really roughly speaking; rk4 is a fourth order method and verlet is second order. For nbody both will show a truncation error and the order essentially means that whatever the error, it will be (2^2)=4 times less for verlet if you halve the timestep while it will be (2^4)=16 times less for rk4 if you halve the timestep. Thus you can take larger steps with rk4 and get the same error (minus rounding errors) or you can take the same size steps and get less error.


So, yes, rk4 is generally more acurate... but it has a systematic error loosing energy, which verlet doesnt have, so a stable orbit will over time degrade for rk4.


At the same time, verlet can get by with one derivative evalation per step where rk4 takes 4, and that being the expensive bit, rk4 can end up losing there... depending on your requirements.


My personal favorite is PEFRL which is energy conserving (commonly termed symplectic for nbody) and also fourth order but actually simpler to implement than rk4.


All that said, we also support both fixed stepping where a frame is always a fixed step length and only one step and a globally adaptive mode where an error tolerance is given, and the integration then takes steps which are juuust small enough to respect that tolerance. That is the default mode.
Here you will not see one integrator as being more accurate than another. Instead you will see one as being faster, because it can keep the error below the tolerance with longer (fewer) steps. This can be seen in the physics step rate.


you might be interested in this video
https://www.youtube.com/watch?v=IJ2MhXUDZ6o