Great GameEngine

[MrESA]

KSP needs this: http://unigine.com/en/industries/simulation/aerospace

Maybe for KSP 2.0

[Red Iron Crown]

That looks like a Serious Business Simulator engine, I wonder what the pricing is like?

Also: That is some deep, deep lurking.

[marach]

That looks like a Serious Business Simulator engine, I wonder what the pricing is like?

Also: That is some deep, deep lurking.

It's Unigine of the kinda popular "Heaven" benchmark fame it's not too expensive but is more than Unity up front from what I remember and since the change to cryengine and unreal4's licensing costs that makes it one of the more expensive out there (if they haven't changed it in the last yearish since I asked them)

[EladDv]

If KSP 2.0 is ever made or a fanmade conversion happens this is the one engine to go to- performance looks good, features that compliment the KSP gameplay and a fully fleshed out C# API that would make converting existing mods quite a breeze

[Theysen]

Oh my god, are there currently any games working on this engine?! I thought these were real life pictures

This would be overkill tbh.

[Red Iron Crown]

Oh my god, are there currently any games working on this engine?! I thought these were real life pictures

This would be overkill tbh.

http://unigine.com/en/industries/games

I don't recognize any of the games in the screenshots, but it looks like some games are using it.

Edited September 16, 2015 by Red Iron Crown
Typo

[Temeter]

KSP most probably wouldn't ever work in that environment. Expensive physics on top of that graphic would kill the performance, not to mention KSPs parts or just the way of ship construction would clash hard with the more realistic artstyle.

I'd rather expect KSP to be worked for a while. There is enough potential to improve the game, especially with the Unity upgrade (which is also going to introduce many new bugs).

Not that such a thing would be awesome, don't get me wrong.^^

Edited September 16, 2015 by Temeter

[KerikBalm]

I think I'm going to need a new computer.... those images look amazing

[EladDv]

KSP most probably wouldn't ever work in that environment. Expensive physics on top of that graphic would kill the performance, not to mention KSPs parts or just the way of ship construction would clash hard with the more realistic artstyle.

I'd rather expect KSP to be worked for a while. There is enough potential to improve the game, especially with the Unity upgrade (which is also going to introduce many new bugs).

Not that such a thing would be awesome, don't get me wrong.^^

an engine that's capable of realistic graphics dosnt mean you cant have non realistic ones.

this engines performance looks much better then any other engine in the operating conditions that KSP requires- complex orbital and atmospheric physics are this engines' bread and butter so i dont see why wont you see it as the most suitable engine to do the job. also it looks like this engine is able to run physics calculations on GPUs which will make the game much more fluid even if vessels will still operate on 1 thread at a time.

[Red Iron Crown]

also it looks like this engine is able to run physics calculations on GPUs which will make the game much more fluid even if vessels will still operate on 1 thread at a time.

A GPU is a large array of comparatively weak processor cores, if the physics is a single thread then it will perform worse on the GPU, not better.

[timsfitz]

It says it can render whole solar systems. Sounds perfect to me. On the other hand, KSPs graphics are good enough. I still play the original XCOM from 1994 every once in a while.

[EladDv]

A GPU is a large array of comparatively weak processor cores, if the physics is a single thread then it will perform worse on the GPU, not better.

GPUs are better for some operations and with the way some of the physics could go (like air stream occlusion and the way atmospheric heat and VFX are calculated) some aspects would obviously benefit from running on the GPU and stuff like GUI would also be better running on the GPU clearing proccessing power on the CPU for more intensive loads like part physics and actual aerodynamic forces calculation (please note that i am not a professional programmer or a hardware designer and i am probably wrong and would love to be proven wrong)

[Red Iron Crown]

GPUs are better for some operations and with the way some of the physics could go (like air stream occlusion and the way atmospheric heat and VFX are calculated) some aspects would obviously benefit from running on the GPU and stuff like GUI would also be better running on the GPU clearing proccessing power on the CPU for more intensive loads like part physics and actual aerodynamic forces calculation (please note that i am not a professional programmer or a hardware designer and i am probably wrong and would love to be proven wrong)

A single GPU compute unit is less powerful than a CPU core for all operations. Full stop.

GPUs are really good at parallelizable tasks because they have so many of the compute units; the number of "cores" is so high that they overwhelm the per-core performance disadvantage. For non-parallelizable tasks the CPU, with its more powerful cores, will beat it every time.

[Temeter]

an engine that's capable of realistic graphics dosnt mean you cant have non realistic ones.

this engines performance looks much better then any other engine in the operating conditions that KSP requires- complex orbital and atmospheric physics are this engines' bread and butter so i dont see why wont you see it as the most suitable engine to do the job. also it looks like this engine is able to run physics calculations on GPUs which will make the game much more fluid even if vessels will still operate on 1 thread at a time.

How do you compare performance, tho?

I think you're confusing what I ment: If it provides more potential/tools for physics, then only for calculating physics for rigid bodies. E.g. the presented aircraft are - as in real simulations - basically one big part with some movable addons (control surfaces, etc). KSP works completely different, it has a bunch of somewhat loosly conected parts that might or might not imitate how a real craft would work. So you'd probably need to completely rewrite half the code anyway, and who knows how performant that is gonna become.

Not sure about GPU physics calculations, I don't think they are really made for the complex calculations you'll get in a simulations. GPUs are great, but there are reasons we still use CPUs.

GPUs are really good at parallelizable tasks because they have so many of the compute units; the number of "cores" is so high that they overwhelm the per-core performance disadvantage. For non-parallelizable tasks the CPU, with its more powerful cores, will beat it every time.

So we'd basically get the same issues that prevent multicore processing?

Also, I think most of this is done by the CUDA chips only Nvidia cards have?

Edited September 16, 2015 by Temeter

[Val]

... not to mention KSPs parts or just the way of ship construction would clash hard with the more realistic artstyle.

That doesn't make sense.

As far as I understand, KSP would look exactly the same in that engine as it does in the current, unless they updated the parts and environments with new different looking textures.

The engine should not dictate how the graphics look. It only dictates what performance you can have with the look/style/detail level you have chosen for your graphics (as a developer).

Edited September 16, 2015 by Val
Duplicates removed

[Temeter]

That doesn't make sense.

As far as I understand, KSP would look exactly the same in that engine as it does in the current, unless they updated the parts and environments with new different looking different textures.

The engine should not dictate how the graphics look. It only dictates what performance you can have with the look/style/detail level you have chosen for your graphics (as a developer).

Of course, an engine is little more than a base with a bunch of tools. I assumed he suggested the engine to make use of the graphical capabilities and scale it's able to work on right out of the box. If not, why would we even know if the engine is better than any other?

[EladDv]

How do you compare performance, tho?

I think you're confusing what I ment: If it provides more potential/tools for physics, then only for calculating physics for rigid bodies. E.g. the presented aircraft are - as in real simulations - basically one big part with some movable addons (control surfaces, etc). KSP works completely different, it has a bunch of somewhat loosly conected parts that might or might not imitate how a real craft would work. So you'd probably need to completely rewrite half the code anyway, and who knows how performant that is gonna become.

Not sure about GPU physics calculations, I don't think they are really made for the complex calculations you'll get in a simulations. GPUs are great, but there are reasons we still use CPUs.

yeah i agree that KSP is different to the examples on their site but really any rigidbody system will able to calculate a KSP craft tree, yeah some tweaks might be in need but it really depends on the engine and the programmers on how they are going to use it- i mean the devs could probably work on a welding system to merge the rigidbodies and just unweld once a part is supposed to detach-heck i am thinking about implementing that right now (having problems with vessel generation) KSP has gone "Gameplay first, Optimisations second" so i wont hold my breath for it to become a stock reality

So we'd basically get the same issues that prevent multicore processing?

yeah but with more smaller cores we could see smaller tasks going on the GPU rather than taking processing power on the CPU, which will be left for complex physics calculations

[magnemoe]

A single GPU compute unit is less powerful than a CPU core for all operations. Full stop.

GPUs are really good at parallelizable tasks because they have so many of the compute units; the number of "cores" is so high that they overwhelm the per-core performance disadvantage. For non-parallelizable tasks the CPU, with its more powerful cores, will beat it every time.

Yes GPU cores are weaker and far more limited than CPU cores.

PhysX is using the GPU https://en.wikipedia.org/wiki/PhysX , think this has an benefit or they would not do it.

Yes its difference between modeling an explosion and static loads like in KSP where the last has plenty of dependencies.

[Red Iron Crown]

Yes GPU cores are weaker and far more limited than CPU cores.

PhysX is using the GPU https://en.wikipedia.org/wiki/PhysX , think this has an benefit or they would not do it.

Yes its difference between modeling an explosion and static loads like in KSP where the last has plenty of dependencies.

Agreed, I was responding to "this engine is able to run physics calculations on GPUs which will make the game much more fluid even if vessels will still operate on 1 thread at a time". For the GPU to deliver a performance benefit requires the physics in question to be multithreaded.

[cantab]

Unigine don't appear to be targeting video games at all. That's not to say it couldn't work, but it's possibly lacking in game-useful features and big on working simulator features.

Also it doesn't really state how it handles aircraft mechanics and vehicle structural dynamics. Most flight simulators rely on "canned" handling values which works for fixed well-known aircraft design but won't work for a build-your-own-vessel game like KSP. KSP and X-Plane are the notable exceptions to that.

[EladDv]

Agreed, I was responding to "this engine is able to run physics calculations on GPUs which will make the game much more fluid even if vessels will still operate on 1 thread at a time". For the GPU to deliver a performance benefit requires the physics in question to be multithreaded.

oh you might have misunderstood me...i was referring to stuff other than the main physics thread for a vessel i know a thread is better to be run on one core and thus is better on the CPU i was talking about small stuff like CoM calculations and heating not big physics threads that are single threaded

[Red Iron Crown]

oh you might have misunderstood me...i was referring to stuff other than the main physics thread for a vessel i know a thread is better to be run on one core and thus is better on the CPU i was talking about small stuff like CoM calculations and heating not big physics threads that are single threaded

Why bother? If your big physics thread is dominating processing requirements then everything else can run on another CPU core without affecting performance (this is exactly how KSP is now).

[Motokid600]

I don't think Squad has enough ambition for that.

[ModZero]

http://unigine.com/en/industries/games

I don't recognize any of the games in the screenshots, but it looks like some games are using it.

I do, and Cradle is ridiculously beautiful (even though that's largely due to the artwork, not engine power). But note, that this is an image generator (that's what they call themselves on the professional pages) intended to be ran on server farms to do real-time generation of very high resolution imagery (why? Simulator cockpits, that's why). They have a physics engine, but you actually have to try and google to make sure of that, and as for visual quality, just look at demos and screen shots for Unity or Unreal Engine. I've even heard that one of those has a game that simulates a solar system.

[marach]

That looks like a Serious Business Simulator engine, I wonder what the pricing is like?

Also: That is some deep, deep lurking.

Found my notes on it, the cost last year was on average $30k for as many users as needed for a project so the point where it came out ahead of Unity pro was 20ish devs not sure on this years pricing so can't say more than that.