Networking Universe Sandbox

[Daryl95]

I have older computer:
Optiplex GX270
2GB RAM
3.0Ghz Hyper Threading CPU
1GB ATI Radeon HD4600
480W Power Supply

As you can see, I've pretty much maxed this thing out, a little light on the ram though. I want to run a considerably large simulation, something a newer computer might be able to handle, but I think I speak for a lot of people when I say that I don't have the money nor the desire to go out and buy the newest or even a newer computer. Especially in this economy.
Second solution:
What if you could network several computers and then run the simulation on all of the computers simultaneously, distributing the program among all the computers, cutting the load for any one computer in half, at the least. This would greatly increase the group of people who can use the product as anybody with two or more older, yet 21st century computers could run the program and get much more out of it. And let's face it, if you're here, then you're a computer guy. If you're a computer guy, then you've got at least two computers from the beginning of this century.
I have posted this in the Community Idea List. I posted it here also because I think that this is something a lot of people would like have and wanted to make sure that the idea got out there.

[vh]

http://news.softpedia.com/news/Rent-Your-Own-Supercomputer-for-2-77-per-Hour-82166.shtml  

not sure how easy networking would be to implement, but for the time being you have a solution!

and by large how large do you mean? I can run several thousand objects with no lag (any my computer has around the same stats as yours) as long as i keep the time step low.

there really is no limit to the number of objects your computer can handle, it'll just take longer to calculate and go through each time step. The only real limit i would think of is when you can no longer move your mouse to make new objects and the lag is at 5 minutes per frame 

Oh and multi-core support would be nice..

and no, i only have a single $400 laptop

[Daryl95]

That's still expensive since it's per core and I'd like to have it run on my own computer anyway. Maybe if there was software that could trick the whole Operating System or just the program into thinking it was running on one computer while it runs on three that would also be a good solution, but it would also be nice to be able to open the sim on multiple computers at once if you were working on something fairly large like I am and wanted more than one person on it, but for now I have to make one subsystem at a time.

[vh]

exactly how large? can you say

its not a matter or how large really, but the amount of objects

and currently ubox only runs on single core, it doesn't utilize more than one. So until that is impletmented i doubt this networking thing is possible.

[Daryl95]

exactly how large? can you say

its not a matter or how large really, but the amount of objects

and currently ubox only runs on single core, it doesn't utilize more than one. So until that is implemented i doubt this networking thing is possible.

right now, it's our solar system, with all 8 planets, Pluto, Sedna, Nibiru, and the important moons + Alpha Centauri A and B + the Milky Way with all the dust particles rendered to make it look like our galaxy.
Originally, I kept a few surrounding galaxies because the novel I'm creating extends beyond our galaxy, I wanted to map any star systems in Andromeda and the Magellanic Clouds that my characters would be on. It seems that the distance, coupled with the dust rendered around the galaxy are the major contributing factors to the trouble, but I can't delete the Milky Way too because I want some sense of where my characters are in the galaxy in relation to the center.

[vh]

about the distance, i don't think it affects performance but thats just my opinion

if you don't want the galaxies to cause lag, try deleting the galaxy, turning the dust multiplier down to something like 0.2, and recreating it.
sure, it might looks less pretty, but it'll make your sim run faster. Even a couple milky ways at 1x dust multiplier doesn't cause me (and hopefully you) any lag.


Dust particles have no 3d'ness to them so it takes several thousand of them to start lagging at all. Objects need to be rendered in 3d, so anything more than 500 or so lags me. It doesn't sound like that is the case though, are you running any cpu intensive programs while having using ubox? anti-virus scanning? when i close them out, the fps generally goes way up.
 If your system gets past 500 planets, moons, and stars that'd be pretty impressive! (and i want a copy of the sim )

Also do you have preformance issues atm? or do you expect them to happen as you go along? Anything below 10 fps starts hindering me imo. Turn off trails and labels if you're not using them too, it halves my lag frequently

[Daryl95]

I don't generally run any extremely resource consuming tasks while running the sim. Right now I have to run the internet but that shouldn't be too bad. I've tried what you said. It does greatly improve performance, in some areas. Still drops to around 12-22 fps in certain areas, but when I zoom into Jupiter I get 34 fps. Up to 90fps zoomed between the stars and the galaxy. I may turn off trails, but I need labels as that is sometimes the only way to see the object when I'm trying to look at an entire system, such as Jupiter, since I'm looking at it now.

[Bla]

Dust particles have no 3d'ness to them so it takes several thousand of them to start lagging at all. Objects need to be rendered in 3d, so anything more than 500 or so lags me. It doesn't sound like that is the case though, are you running any cpu intensive programs while having using ubox? anti-virus scanning? when i close them out, the fps generally goes way up.
 If your system gets past 500 planets, moons, and stars that'd be pretty impressive! (and i want a copy of the sim )

It's not the 3D graphics on normal bodies which make the simlation much slower than the 2D particles, it's because they have mass. When you add a new body, it has to calculate both the force of the body on all other bodies and particles, plus the forces from all other bodies on the body. If you just add a particle, you only have to calculate the force from all other bodies on the particle.

I agree with the idea of running simulations on multiple computers. I wonder how it could work... I think it'd require improvements to the multicore support.

[vh]

oh really, thats new
because i know some programs which simulate upwards of 5000 particles with mass, and they run less cpu-intensive than ubox!

[Chris]

The next version of Sandbox comes with multi-core support.
On a dual core system, we measure a 30-40% performance increase already,
i think this will make a lot of people enjoy the application even more.

Network ditributed computing is indeed an interesting thing to consider,
but it usually lends itself best to computations that don't have to be visually presented in realtime.

Think of it this way:
rendering to screen at acceptable framerates requires an interval of roughly 16ms (~ 60Hz).
If we offload the simulation math to another machine, or multiple machines for that matter, we will have to make sure the trade is worth it. Latency suddenly becomes an issue, because every millisecond we wait for a network response is a millisecond lost.

That said, multi-core will already make all the difference for most of us.

Cheers

[FiahOwl]

The next version of Sandbox comes with multi-core support.
On a dual core system, we measure a 30-40% performance increase already,
i think this will make a lot of people enjoy the application even more.

Network ditributed computing is indeed an interesting thing to consider,
but it usually lends itself best to computations that don't have to be visually presented in realtime.

Think of it this way:
rendering to screen at acceptable framerates requires an interval of roughly 16ms (~ 60Hz).
If we offload the simulation math to another machine, or multiple machines for that matter, we will have to make sure the trade is worth it. Latency suddenly becomes an issue, because every millisecond we wait for a network response is a millisecond lost.

That said, multi-core will already make all the difference for most of us.

Cheers

HURRAH! HURRAH HURRAH! MULTICORE SUPPORT! HURRAUH!


...sorry, just wanted to use my duel core processor.