Is it just me who wants multithreading?

[suzin.felipe]

i guess SQUAD, being new to the gaming industry when this KSP started, didnt know the importance of choosing a potent engine.

(i have no idea as i aint in the industry, but i think it's probably that licensing unity was considerably cheaper than other options)

KSP is develop by 6 dudes with no budget, previous experience, in a company with a completely different business focus.

We are lucky that they had an engine that they could use.

We can only hope that they finish this game soon, cash a really big profit and make KSP II with the better engine money can buy.

[SFJackBauer]

http://bulletphysics.org/wordpress/?p=340

Learn that Unity's physics engine is years old, can't do multithreading nor use the GPU. Read this. Cry a lot.

"Note that the ATI Radeon 7970 can simulate 110k rigidbodies at 30fps"

(whats more, bullet engine is free and has integration solutions in unity, afaik)

Its easy to simulate hundreds of thousands of balls who have no user input, no sound, no texturing, no fancy lighting...

The problem is not that it is impossible to tap into the multicore and gpu power. It is possible. The real problem is how to tackle the exponential complexity that is needed to keep everything synchronized as a coherent system that does something more useful than rendering a bunch of balls in the screen, or a bunch of cloth hitting one another.

[lammatt]

KSP is develop by 6 dudes with no budget, previous experience, in a company with a completely different business focus.

We are lucky that they had an engine that they could use.

We can only hope that they finish this game soon, cash a really big profit and make KSP II with the better engine money can buy.

this is exactly what i want to say

i still recall how barebone this thing was when i just got it when it was 0.12

it's been a very long way this game has gone already.

[Kerbart]

i guess SQUAD, being new to the gaming industry when this KSP started, didnt know the importance of choosing a potent engine.

I'm pretty sure they knew the importance of picking the right engine. And I'm pretty sure that, given the scenario KSP started out from, Unity seemed like a decent choice, or maybe even the best choice. Then there's a million constraints: cost, familiarity, etc. I doubt it was ignorance that drove Squad to Unity. And I'm not sure Unity is not a powerful engine. I'm not familiar with the other engines, but maybe Unity offered some options (physics modeling the way KSP needs it) that others didn't, or not in a sufficient way?

[Infineon]

yup. its just you. its not like this has ever been discussed before. I sure as hell do not want a feature that would greatly increase the part limit bottleneck, no that would be totally pointless

Ok, if everyone wants this, how can we convince Squad into spending time and resources on it?

Let's say: how much are we ready to pay for this to be implemented?

There are no official numbers, but lets assume 100 000 copies of KSP were sold. I would bet that this number is much higher, but I'd go safe side here.

I think at-least half loved KSP, and wants to get it faster (again, my own feeling is - you would not buy this game in the first place if you would not like it, but I will go safe side again)

How much are we ready do pay for this extra work and effort? Call it "Pro upgrade" or "KSP 2", doesn't really matter. I'd be happy to pay some $20 to get rid of the performance issues, and enjoy KSP beyond current part-limit. But let's take half of that: $10.

100 000 copies * 0.5 of fan-base * $10 = $500 000

I sincerely believe Squad is missing some serious profit opportunity here.

Edit:

Don't get me wrong, I like new features and I appreciate all bug-fixes and tweaks Squad is working on, but there is little point of all that additional work done if we can't use it on real complex aircraft and rockets.

Edited October 9, 2013 by Infineon

[Whackjob]

I have no doubt whatsoever that if Squad started up a kickstarter project to have the engine moved to a differing platform, they would have more than sufficient funds that would permit them to do different things:

1.) License a very high end game engine.

2.) Hire on additional staff to handle the porting while they continue their current objectives without interruption

3.) Shut up all and sundry at the office who might complain about a port with coffee and whiskey and limited edition premiere boot polish

4.) Buy the company from whoever

5.) Buy a company sportscar

6.) Buy a decade supply of fresh tires and hire a small team of stuntmen to continuously do donuts in the company parking lot with the company sportscar, 24/7/365. "Deliveries through front entrance at extreme personal risk."

[Umlüx]

imagine all the crafts we could do with a multithreaded physx enabled x64 physics engine..

i'd gladly pay $10 for that.

but i fear squad is not experienced or big enough for a giant step like an engine switch.

[Whackjob]

Tell you what, Squad. I'll go halfway with you here. Do the kickstarter, get the money, hire the staff, port the engine, and I'll let whomever donates the most shave off my beard. Second place gets to shave their name into my back.

I'll take the hit, but I want results!

[suzin.felipe]

Change engine in the middle of development is the same as throwing all they did till now and start from scratch.

The only game that tried that was Duke Nuken and we all know how did that go.

My opinion is that the last feature to be implemented is the career mode, after that a big push in polishing, adding parts for science and improving the physics model a bit.

Ship the game as ready, cash in as much money as you can and start working in KSP II with the Cry engine or something like that if you want.

[Rakshasa]

Some things just can't be multi-threaded.

How physical forces traverse through a structure might be one of those things.

Sure, the graphics can be off-threaded, but if that physics problem (not sure if it exists, just suspect it does) can't be multi-threaded, we're SOL. Without a lot of work, anyway.

Wrong, physics simulation is not that difficult to multithread and has a very long history in computer science.

[mellojoe]

Don't forget. Optimization comes LAST in game design. No sense to optimize something that is only half-finished, only to have to throw out parts that aren't working. Optimization will happen. Later.

[3_bit]

Well I'm building an AMD-based rig in December, and while their processors heave extremely high value, Intel wins at single-threads. I hope the issue of threading is sorted out before .24 or so.

[LtHeckard]

Your Quad core CPU is quite good enough. You may need more RAM or a better Graphics card. This game requires about 4GB RAM on high setting. You will also need a Graphics Card if you want to assist your framerate. the CPU is really important, but a Graphics Card and RAM is very important too. You may want to upgrade your Graphics card. the EVGA GTX 650 is about $100-120 and it should give you a good boost. 8GB of Corsair Vengeance RAM will cost about $70. Although, make sure the RAM fits into your Motherboard.

Edited October 9, 2013 by LtHeckard

[magnemoe]

Don't forget. Optimization comes LAST in game design. No sense to optimize something that is only half-finished, only to have to throw out parts that aren't working. Optimization will happen. Later.

Yes, add that Unity might change before release making it even more wasted.

[cullyn]

imagine all the crafts we could do with a multithreaded physx enabled x64 physics engine..

stop right there.

leave the physx out. there are other solutions that don't lock a substantial proportion of the user base out of said features. Its actually one major reason I REFUSE to buy nvidia for a GPU. physx should be usable by any GPU.

[numerobis]

Unity doesn't use the GPU for physics on any video card.

The choice of game engine involves a lot more than just runtime speed. There's money, training, and development time considerations that are often more important. Unity is cheap to start with (free as in beer), not very expensive when you're going ($75/month per seat, no royalties), relatively well documented and increasingly well known, and the Unity editor UI is pretty good so you can get both devs and non-devs working efficiently, and you can hire people and get them up and running fast. Cross-platform work is pretty easy; change a setting and you build on Windows, Mac, Linux, or various mobile platforms (obviously not everything maps perfectly between them, but 95% of the work is done for you). Progremming is mostly in C#, which is easier to program in than say C/C++.

And it's all a bit slow in the performance department. You win some, you lose some.

[Eric S]

stop right there.

leave the physx out. there are other solutions that don't lock a substantial proportion of the user base out of said features. Its actually one major reason I REFUSE to buy nvidia for a GPU. physx should be usable by any GPU.

While I'd agree that I'd rather see a GPU-neutral alternative, Unity is already using Physx, it just isn't using a recent enough version of Physx to use GPU assistance. Yes, the first versions of Physx were CPU only, and that may be what he was referring to.

[jfx]

PhysX is already dead, it just has not stopped moving yet. (And nVidia refuses to pull the plug with the recent $ Million buy in for two AAA titles).

[DaRocketCat]

Tell you what, Squad. I'll go halfway with you here. Do the kickstarter, get the money, hire the staff, port the engine, and I'll let whomever donates the most shave off my beard. Second place gets to shave their name into my back.

I'll take the hit, but I want results!

AND you'd have to set your shaved self as profile pic for every single site you use

Guaranteed success.

[Dispatcher]

Perhaps I should just show up at the homes of the unity developers and just stare at them for a while.

Until unity update happens.

Whackjob, are those ear plugs (i hope)?

[jwenting]

Am I crazy or is KSP able then to launch multiple instances of Unity engine to work on different pats of flight physics (thrust/weight/gravity/part links/air resistance/etc.) and combine results into a smooth running simulation?

yes, it's you. Unity can't do that.

[Infineon]

yes, it's you. Unity can't do that.

Unity can't do what? Calculate only current air pressure? Or current weight? Or current thrust vector? Current center of mass? It is unable to calculate one parameter? And it's impossible to retrieve these results and use them to calculate acceleration, part integrity, overheat, etc.? And then it's impossible to present us a highly optimized game?

[numerobis]

Theoretically, it's trivial. In practice, it's a lot of work.

Oh look, an example of how physics engines are hard to get right.

Edited October 9, 2013 by numerobis

[veniteo]

Even if unity supported multithreading, it might not work. It might not be possible to multithread the physics calculations, because the calculations of each part of the ship are not independent of each other. However, it should be possible to use GPUs to accelerate the calculations

There are two reasons GPUu acceleration for physics doesn't work. This is the opposite side of the spectrum from Bitcoin mining. GPU's are notorious for being low-level calculation workhorses. 3d physics calculations will never be efficient on a GPU, simply because GPUs are designed for Repetitive simple calculations. its why we don't have CPU graphics. A CPU can't hope to match the speed with the low level calculations. CPUs are like think tanks, they are great for short bursts of high level calculatons, but they are incredibly inefficient when it comes to long drawn out operations. The issue lies squarely in the clock speed of your CPU at this point. If you have an overclockable system, That is your best bet. Only overclock if you know what you are doing. (I would not recommend overclocking on a MoBo that you paid 70 bucks for)you can boost your performance in game by 30% or more. I don't see Unity Becoming Multithreaded anytime soon, and with the overall trend being lower clockspeed, multicore CPUs, There isn't really another solution.

Unity can't do what? Calculate only current air pressure? Or current weight? Or current thrust vector? Current center of mass? It is unable to calculate one parameter? And it's impossible to retrieve these results and use them to calculate acceleration, part integrity, overheat, etc.? And then it's impossible to present us a highly optimized game?

Unity Can't efficiently process all of these at the same time. hell it can't process most of this efficiently period. I actually suggest that the Dev team slow down Unity Engine development and Get Havok. about 900x better for physics.

Your Quad core CPU is quite good enough. You may need more RAM or a better Graphics card. This game requires about 4GB RAM on high setting. You will also need a Graphics Card if you want to assist your framerate. the CPU is really important, but a Graphics Card and RAM is very important too. You may want to upgrade your Graphics card. the EVGA GTX 650 is about $100-120 and it should give you a good boost. 8GB of Corsair Vengeance RAM will cost about $70. Although, make sure the RAM fits into your Motherboard.

RAM is not that much of a concern here. I run 20 GB of 1333 and i have maybe 10 FPS @launch. thats with an FX-6100. I'm about to upgrade to Crossfired HD 7770's and an FX-6300 Edited February 19, 2014 by Aphox
Merged multipost; please use edit function

[qromodynmc]

I believe ksp performance mostly depends on single core clock speed,I mean 5 ghz single core modern cpu would calculate faster than 2 ghz eight core cpu.