The only thing thing that I care to see in .24

[vexx32]

The difference is that PhysX can only run on NVidia cards, not AMD. In addition, the PhysX libraries that come with the latest version of Unity are still outdated and do not actually run any calcs on the GPU whatsoever.

[John FX]

I made this and meant to post it earlier. I decided to make a silly rocket.

Javascript is disabled. View full album

It has so many tanks that it can only just lift itself (TWR=1 or so). It has six parts, it has zero struts. I cannot make it bend. Even when full throttle and nearly empty.

It actually turned into a huge amount of fuel storage at about a 250km orbit SSTO and I may do something with that.

I see a lot of problems coming from using 5 tanks where 1 would do and where NASA for example did use 1. Using 5 parts makes an object much more than 5 times more floppy...

As I have said earlier, what replica makers really want is procedural, paintable tanks. Procedural, paintable, fairings and interstages and probably procedural wings. Nobody actually wants to struggle with a limited amount of predefined parts when trying to recreate a real world rocket that had bespoke fuel tanks unless they are a masochist or maintaining elitism. They want to be able to make bespoke fuel tanks just like NASA does.

Solves the problems with lots of tanks in a single stage, reduces physics calculations, part numbers and increases rigidity.

win-win.

I like how you guys change the argument to point to "bad designs" when even the most reasonable of them would wobble without struts.

http://hostpicture.eu/photos/20131128138559761109433.png

Look at that one. It's even a replica of a real rocket (tronador 2). The 2.5m tanks would wobble like crazy and the same goes for the top stack joined by the adapter.

http://www.imgnook.com/41JGgo.gif

My Dnepr-1, which is also a replica, needs strutting between all it's tanks. and the ASAS.

Now tell me how are those designs insane or unrealistic or whatever the hell you want.

As I say, I notice you are using many tanks where one should be. That is your problem.

This is as a result of trying to get an exact look from limited parts. You would be better replacing the four fuel tanks in the first stage with a single procedural one which would eliminate your problem without needing to adjust node strength.

Myself I am trying to say the problem is a real one but the cause might not be the one you think and could have a different solution.

Edited January 7, 2014 by John FX

[Sethnizzle]

I want 64 bit.

[Fleb]

Need More science for struts!

[PDCWolf]

I made this and meant to post it earlier. I decided to make a silly rocket.

http://imgur.com/a/rQovf

It has so many tanks that it can only just lift itself (TWR=1 or so). It has six parts, it has zero struts. I cannot make it bend. Even when full throttle and nearly empty.

It actually turned into a huge amount of fuel storage at about a 250km orbit SSTO and I may do something with that.

Too low TWR, too little torque to bend. By the time the twr is good (for normal rockets) you are not fighting any force.

I see a lot of problems coming from using 5 tanks where 1 would do and where NASA for example did use 1. Using 5 parts makes an object much more than 5 times more floppy...

Well, given this is an abstraction it is correct to assume that 5 tanks together form a bigger one (I'm talking about internal tank structure) than just act like 5 different tanks. Wobble is still a problem and a bug, even if avoidable.

As I have said earlier, what replica makers really want is procedural, paintable tanks. Procedural, paintable, fairings and interstages and probably procedural wings. Nobody actually wants to struggle with a limited amount of predefined parts when trying to recreate a real world rocket that had bespoke fuel tanks unless they are a masochist or maintaining elitism. They want to be able to make bespoke fuel tanks just like NASA does.

Procedural everything kills the challenge in creating replicas, at least for me. Thank you for calling me an elitist just for liking a bit more challenge than you.

Solves the problems with lots of tanks in a single stage, reduces physics calculations, part numbers and increases rigidity.

win-win.

But wobble is still there and it's going to show sooner or later.

As I say, I notice you are using many tanks where one should be. That is your problem.

This is as a result of trying to get an exact look from limited parts. You would be better replacing the four fuel tanks in the first stage with a single procedural one which would eliminate your problem without needing to adjust node strength.

Myself I am trying to say the problem is a real one but the cause might not be the one you think and could have a different solution.

I get procedural everything, and wobble is still unsolved so the next guy is going to come on the forum and complain about it. Do you plan on telling everybody to use a mod to overcome a bug or is it better to have squad finally fix it?

Even if there are workarounds (sacrificing the aesthetics) wobble is still a problem and is going to show up eventually.

[DaMichel]

The difference is that PhysX can only run on NVidia cards, not AMD. In addition, the PhysX libraries that come with the latest version of Unity are still outdated and do not actually run any calcs on the GPU whatsoever.

Ah, all right. Should have known that it's NVidia only :<

[metl]

That would totally explain why my framerate is worse on my ATI card over my wife's NVidia which is several years older! I knew I should have stuck with Nvidia.

[marach]

IIRC there's a version of the Bullet physics engine for Unity. I've heard a lot more good things about Bullet than PhysX. Of course, that all depends on how it's implemented. An aerodynamic system for instance still needs to be made from scratch, but from what I gather Bullet is generally much more reliable, and handles complex shapes better. Non-wobbly joints depend on setting the best settings for the joints themselves (which is all KJR does, I think?)

But of course, the game's been developed for coming on three years now using Unity's default PhysX implementation, so it would be non-trivial to go back and undo all of that...

A lot of indie devs have rolled their own bullet integration but iirc nobody has actually released a fully working implementation to the public...

I know the Rawbots implementation uses the same calls, variables, etc, as the unity standard so much of their change over was actually a case of re-tweaking parts to be weaker than they were.

[marach]

That would totally explain why my framerate is worse on my ATI card over my wife's NVidia which is several years older! I knew I should have stuck with Nvidia.

Nope nothing is run on the gpu except graphics, single core CPU speed is one of the major bottlenecks right now though...

[vexx32]

That would totally explain why my framerate is worse on my ATI card over my wife's NVidia which is several years older! I knew I should have stuck with Nvidia.

It shouldn't; last I heard of the matter, the PhysX modules used in KSP are still too outdated to use the GPU for physics calcs... I may have to recheck that, though.

[draeath]

physx is NOT nvidia-exclusive, only the GPU-acceleration of it is. It runs quite happily in CPU.

I'd love it if KSP could use CUDA/OpenCL though, I'd imagine that could help out when the part count gets hilarious. Probably not workable though, seems KSP uses floats a lot and GPUs like integer math.

[marach]

physx is NOT nvidia-exclusive, only the GPU-acceleration of it is. It runs quite happily in CPU.

I'd love it if KSP could use CUDA/OpenCL though, I'd imagine that could help out when the part count gets hilarious. Probably not workable though, seems KSP uses floats a lot and GPUs like integer math.

Bullet (the most requested physics change) uses opencl and is platform agnostic, unlike physX and CUDA which are a pile of steaming^H^H^H^H^H^H^H^H proprietary horribleness...

Tbh I hope Nvidia dies for all it's anticompetitive moves. They aren't even subtle about it! When they offered AMD physX compatibility one of the requirements was complete disclosure of hardware designs "to better optimize it"...

[metl]

Well I have an ATI 6670 with 1Gb VRAM on a dual core 2.4 Ghz AMD with 8 gigs ram running Win7x64. My wife has a Geforce 8800GTS with 320MB VRAM on a dual core 2.1Ghz with 4gb ram running Win7x64. This is the only game that she gets better framerates than I, and can even use higher settings. The only other difference is hers only has the Demo installed. Any suggestions? PhysX would certainly explain it.

[CleverClothe]

I don't know, a bit over a hundred tons seems like a reasonable amount.

That is a 6.6 meter structure. I'm not a structural engineer, so I can't tell you what exactly an equivalent 2.5 meter structure would carry. But I can tell you it would be less.

That thing also weighs about 2 tons, while the KSP large decoupler is 0.4 tons.

[CleverClothe]

As I have said earlier, what replica makers really want is procedural, paintable tanks. Procedural, paintable, fairings and interstages and probably procedural wings. Nobody actually wants to struggle with a limited amount of predefined parts when trying to recreate a real world rocket that had bespoke fuel tanks unless they are a masochist or maintaining elitism. They want to be able to make bespoke fuel tanks just like NASA does.

In complete agreement. I drool at the thought.

[marach]

is hers only has the Demo installed

There you go it's that simple. Since the demo we've had what 5/6 updates with numerous parts etc added?

[PDCWolf]

Well I have an ATI 6670 with 1Gb VRAM on a dual core 2.4 Ghz AMD with 8 gigs ram running Win7x64. My wife has a Geforce 8800GTS with 320MB VRAM on a dual core 2.1Ghz with 4gb ram running Win7x64. This is the only game that she gets better framerates than I, and can even use higher settings. The only other difference is hers only has the Demo installed. Any suggestions? PhysX would certainly explain it.

Since the last demo we got this feature called "ocean lag" implemented. We also got "performance fixes" and "Better scene distribution to avoid reloading everything when you change the scene (like from vab to space center or space center to flight)". I used the quotation marks because you, a 0.23 user can see those "improvements" over a 0.18 user.

[metl]

Huh. I thought all of those features were supposed to improve the overall performance. It sez so on da packages!

[draeath]

Well, it does for me! My .23 is a much better performing KSP than my old ones were.

[Bane1998]

The only thing that I want to see in .24 is an end to wobbly rockets. One moment It's flying straight then, BOOM, it starts to wiggly horribly and either throws off the orbit entirely or breaks up. They way connections work in KSP is that there is one point that connects to another point(very weakly) and that holds it together which is not good for keeping large things together. Please make connections more stable squad.

I don't care about wobbly. It means my ship isn't stable, and I should fix it with struts or a different design or something. It's a feature, not a bug.

However, the amount of lag it causes is unacceptable. Maybe I'm just very sensitive to lower FPS, but I find it extremely unpleasant to play with any sort of large structure, which is very unfortunate.

[PDCWolf]

I don't care about wobbly. It means my ship isn't stable, and I should fix it with struts or a different design or something. It's a feature, not a bug.

Nope, squad said it is a bug. Not a feature, not realistic, not intuitive, not a good mechanic, not a challenge, just a bug that must be dealt with.

[Seret]

Nope, squad said it is a bug. Not a feature, not realistic, not intuitive, not a good mechanic, not a challenge, just a bug that must be dealt with.

Can you point out which bug on the tracker refers? I've just had a bit of a search and can't find one.

[PDCWolf]

Can you point out which bug on the tracker refers? I've just had a bit of a search and can't find one.

Would you put "ocean lag" in the tracker after squad mentioned it? Maybe you would if they said they fixed it and it is still there. Its pretty much like resources and anything pre-skunky-being-fired. No evidence exists anymore. You'll have to trust me on this one, like with resources and squad saying multiplayer was not coming which everyone but the fanboys can confirm. Everything that remains is nova's/bac9/any other ex-dev's word and user memories.

[Seret]

"Ocean lag" is in the tracker, bug #1569.

Sorry, but "you'll have to trust me" is pretty unconvincing, you need to back that up with some actual evidence. If somebody in authority made a statement on it at some point, you should be able to dig that out.

My view: if it's not in the tracker, don't expect it to get fixed. There's no real reason they'd need to keep it out of the public tracker, since it's such an old issue. If you really want to prove your point, open a bug for wobbly parts and see what happens. I'll be standing by to eat my hat if need be.

[boxman]

As much as I hate to say it I think it would no longer be Kerbal Space if they completely took away the wobbling. Just look the newbies who just started playing the game. The random wobbling, explosions and such due to lack of struts is a part of the reason they find this game fun from the very beginning.

A large amount of the people who start out playing this game mainy does it for fun and the explosions and then later move on to make more serious ships and learning the physics behind it. If we take away this part then these people might not get to the point where they realize that space and the more serious simulator aspect of it is fun as well.