What 'lags' more ?

[micr0wave]

Which one uses more cpu, one 300-part vessel, or two 150-part vessels ?

[K3achas]

I would think that two vessals would use more computing power because there would have to be twice the orbital calculations.

[Sharpy]

Within the same physics bubble?

*shrug* hard to say. I'd say probably two. In one hand, there's less cross-dependencies to calculate as strains on the end of one 150-part vessel don't need to be distributed over the other vessel; any force travels only through 150 parts, not 300. OTOH the game must check for collisions between the two. Within a single vessel things like clipping are possible because the game doesn't check for self-collisions (although there are exceptions, esp. with moving parts - e.g. the ramp lowered using landing legs pushing it down works basing on collision between the ramp and the legs and these belong to a single vessel. OTOH the ramp based on airbrakes requires to be detached; if it belonged to the same vessel the airbrakes would just phase through.)

So... do the reduced physical complexity of the body outweigh the overhead of collision-checking? Honestly, I don't know.

edit re: orbital calculations. These really don't create much lag. The game can handle hundreds of objects in orbital flight at the same time just fine, adding one more does very little.

Edited July 18, 2015 by Sharpy

[micr0wave]

yeah, within physics of course.

[Feradose]

I think the game uses one core per vessel inside the same physics bubble.

It would lag a little bit more than half of it when youre using two 150 parts instead of 300 parts, since 300parts will be stuffed in one core and two 150 part ships will be using two cores.

[Violent Jeb]

I'm sure it's the latter. They both have the same part counts, but it's two big objects to track instead of one.

FWIW, I imagine 2x100 part crafts and a 300 part craft would be closer in footprint. multi-part space stations always crash me with much lower counts vs. a single ship.

[Laie]

I never noticed any improvement after docking (two large vessels before, one huge vessel after). That doesn't mean that there is no difference at all... but it's certainly not huge.

(hold on a sec)

Had a savegame from before docking to mid-sized vessels (nothing huge, but enough to make the clock go yellow). Enabled FPS counter, docked, and well... if I squint really hard I can maybe convince myself that there is a very very slight improvement. It's certainly not on the order of magnitude Violent Jeb imagined.

[pincushionman]

I think the game uses one core per vessel inside the same physics bubble…

I'm not sure that's true in the current version of Unity/PhysX being used - I think there's only one physics thread total available to the game. I'm pretty sure that will change in U5, though.

[Eric S]

I'm not sure that's true in the current version of Unity/PhysX being used - I think there's only one physics thread total available to the game. I'm pretty sure that will change in U5, though.

Correct on both counts. In 1.1, it will almost definitely be the single 300 part craft that lags more. Prior to that, I haven't noticed much of a difference.

[Wolfox]

I believe that single craft will lag more, since orbital calculations aren´t that complex (in comparison to real-time physic interaction between collision meshes) in term of CPU usage, its part interactions what melt your CPU. You can see this phenomena in action when you lift off with huge rocket with many, many boosters. As soon as you ditch them all out, your FPS rise considerably, even though that all those parts are still on screen and still in physics bauble, but no longer affecting each other, but each flying on its own. Atmospheric drag calculations are still very simple as opposed to part interactions.