Not Turbo Boost Button It Is A Bug

[Moesly_Armlis]

Want to improve your game play with less game slow down and lag.

Before

Stable and normal time yellow performance at 53% Physics Time Ratio.

After

Stable and normal time green performance at 99% Physics Time Ratio. Physics time warp to 4x and then back to normal is only difference.

I confirmed this "turbo" after installing KerboKatz PhysicTimeRatio.

Always suspected that this increase was happening but was given clear evidence that there is something going on under the hood.

Anyone else have the same performance boost?

Build a vessel that has the indicator at the top left have yellow performance throughout flying around.

Restart or reload and before take off use physics time warp to 2x or 3x and return to 1x normal.

Fly the vessel and now the performance is better and now the indicator blinks between yellow and green.

Reload and physics time warp to 4x and then back to 1x normal.

Fly the vessel and now the performance is improved and the indicator in the top left should turn green.

This works on my old laptop and without this trick my game play was all lag.

I will work on a video.

Why does physics time warp grant increase game performance?

Edit:

The unusual issue is the performance continues after time warp is returned to normal. Game play is horrible without doing this and at the beginning of each of my games the physics time warp up goes up to four and then back to normal physics time warp which somehow grants improved performance from then on.

Javascript is disabled. View full album

Edited September 12, 2015 by MoeslyArmlis
Normal Time 1x

[Nansuchao]

That's really interesting. I hope someone of our experts can explain this behavior.

[CalculusWarrior]

Is the game progressing slower (i.e. in-game seconds take longer than IRL seconds)? I know of a way to increase framerate at the cost of slowing down the in-game time, and it's by changing the physics delta-t, which seems similar to what that mod's doing.

[KasperVld]

Can anyone confirm this behaviour on a stock install?

[Moesly_Armlis]

Can anyone confirm this behaviour on a stock install?

I just tried it with the stock game and it does the same.

The performance is altered after a physics time warp.

Even animation of kerbals which would be very slow are sped up using this "trick".

Here is a video.

Skip to 1:30 for the test.

The animation would stay slow all day and performance would remain at a yellow indicator.

This shows the performance increasing after advancing and returning physics time warp. The non-physics time warp has no impact.

After returning from 4x physics time warp the indicator now flickers green, the animation is quicker and game play more responsive.

This was recorded at 7 frames per second. The distance traveled in about thirty seconds was measured.

The last trial the distance was much further than the initial. The game is running closer to real time and more actions are being accomplished. This would also be true for vehicles.

This shows the game is playing faster after advancing physics time warp and then time warp is returned to normal.

The trick to look up at the sky to increase performance is similar to this trick. But this trick broadens the scope.

Edited September 3, 2015 by MoeslyArmlis

[Chaos_Klaus]

The game has two different cycles for graphic rendering and for physics calculation. Depending on the situation it will favor one over the other. I guess this mod makes KSP do less physics ticks, which means less stress on your CPU but also less accurate physics. This could potentially invoke the kraken.

[Claw]

I'm not seeing this on my system, but I am intrigued. I do recall seeing lag similar to what is in the video, but I can't produce that lag on demand through launching a rocket or sending a kerbal EVA. Perhaps you can provide a few more details so we can try to narrow some of this down.

Can you guys tell us what version of KSP you are running? And tell us a bit about your systems (processor, video card, ram, etc...) Also, are you on a laptop or desktop?

Cheers,

~Claw

Edited September 3, 2015 by Claw

[problemecium]

I don't trust this. Whenever performance mysteriously spikes for me, I worry it's because the game has forgotten some job it's supposed to be doing, and for all I know it might decide atmospheres don't exist or that I have no money.

[John FX]

This is odd but interesting. If the effect is real then it might lead to a performance improvement, without the need to do the workaround for want of a better word, after a code examination.

[AlphaAsh]

Interesting observation. I'd postulate that the warp not only suspends nearly all physics calculations but effectively 'resets' a bunch of physics related variables being tracked.

How long does this 'performance boost' actually last? To be honest, I would expect it to dissipate after a short time as physics tracking is required to do as much work as it did before being reset.

[Majorjim!]

The OPs craft seems to have around 25 parts. If 25 parts causes his game to run 'in the yellow' then it is a very slow system and this may make the 'effect' more noticeable.

[_stilgar_]

The OPs craft seems to have around 25 parts. If 25 parts causes his game to run 'in the yellow' then it is a very slow system and this may make the 'effect' more noticeable.

Or maybe it's a laptop on battery. CPU throttled down to save energy. Turing physics warp caused the CPU to switch to higher frequency, improving performance even after the warp was disabled.

[*Aqua*]

Does Mono has a JIT compiler? If so changing the time warp could trigger on-the-fly optimization of some the code.

[Majorjim!]

Or maybe it's a laptop on battery. CPU throttled down to save energy. Turing physics warp caused the CPU to switch to higher frequency, improving performance even after the warp was disabled.

Interesting, could be this.

[peachoftree]

Does Mono has a JIT compiler? If so changing the time warp could trigger on-the-fly optimization of some the code.

I don't think so, DLL's seem to be compiled code.

[*Aqua*]

Compiled? It's not IL code?

[peachoftree]

Compiled? It's not IL code?

I feel stupid now that I think of it. of course its IL so it might have a JIT compiler.

[Moesly_Armlis]

The turbo will last throughout my play time. Have no issues flying, driving and going to orbit. Plus I am involved in a surface docking experiment that has not had many Kraken attacks. Usual two to three hours play time and the turbo works just fine. Without the turbo this experiment would be very slow and not enjoyable. If this is a "bug" then slap a feature label on it and do not fix it. The turbo seems to be like that button on an old PC that increases the clock.

My ancient laptop was suspected of throttle control.

Specs:

Windows 7 64 bit

Intel 2 Duo P7450 @ 2.13 GHz

Ram 4 GB

ATI Mobility Radeon HD 3650 Dedicated 512 MB

OS and KSP 1.0.4 installed on 64 GB SSD

Another observation concerning the Max Physics Delta Time per Frame and resetting the turbo.

Once the game is exited (such as an alt-F4) and restarted the turbo is lost.

Investigating this further reveals that the turbo continues to work even after returning to the main menu and starting another fresh game.

When changing the Max Physics Delta Time per Frame the turbo is lost and restarted with the same procedure of ramping physics time warp.

So I have to lean towards the Max Physics Delta Time per Frame being a lead suspect.

The game play is much improved over a loss of some calculations. So what is the harm and would be similar to changing the Max Physics Delta Time per Frame. Bring this turbo bug/feature under some control so that when more accuracy is required then it can be increased.

Edited September 3, 2015 by MoeslyArmlis

[Hotaru]

I've noticed something a bit like this. I keep Max Physics Delta Time per Frame set really low (like 0.005) because I prefer to have the game run in slow-motion at high FPS rather than low FPS in real time.

However, any time I use phyics warp (even just 2x), that stops working and the game seems to revert to a much higher physics delta per frame setting--everything runs in real time, but at reduced FPSe. Since the result tends to involve low-FPS-induced Krakens I generally restart when that happens and avoid using physics warp if I can help it.

Although I haven't tried on a stock install, my modded installs have been doing this since 0.90. I do have a mod installed that allows custom warp speeds (can't remember the name), I will try uninstalling it and see if that has any effect although I think this has happened since before I added that mod.

[Moesly_Armlis]

The frame rate was checked using the debug window.

The Frame rate was 33 frames before Physics Time Warp and after the frame rate was reduced to 20 frames.

The game is more responsive because the action is happening closer to real time, it does not feel like slow motion, the camera pans and zooms much quicker and kerbal animations are smoother (which means it takes less than one second for them to stand up compared to two and a half).

What calculations are being lost and how does that affect simulation. The time per frame is what is being altered. My experience is no difference other than the game now feels alive after ramping the physics time warp.

The issue that I am understanding is that everyone is doing this turbo by accident. Any time a physics time warp is used then from then on the turbo is engaged. Examples are using physics time warp to reach 70 km or decent with parachute and after that the turbo effect has reduced the Max Physics Delta Time per Frame and can be reset by exiting and restarting or going to the main menu and changing it at on the settings page.

This is then I presume a bug (read feature) and would need reporting and eventually fixing. I like the increase in performance and since this has been happening anyways it should be an option in stock, much like the mod Hotaru mentions.

TLDR:

If you never use physics time warp then this has no value. Performance increases by simply ramping up physics time warp and returning to normal, something that happens in everyone's KSP session.

The Max Physics Delta Time per Frame should have lower values available and can be changed easily from within a game. Both are currently happening in a stock game with almost nobody noticing.

Launch a kerbal in a command pod and ramp up/down the physics time warp and recover. Then launch that 600 part monstrosity and enjoy smoother game play.

Edited September 4, 2015 by MoeslyArmlis

[NecroBones]

Huh, this is fascinating. I've probably never noticed because I almost always end up doing a physics warp pretty early in any given mission. From a coder's perspective, this sounds like the non-accelerated settings are being reverted to an internal default, rather than the player's config-panel settings.

[Motokid600]

Interesting I'll have to try this. Another thing. I don't know if its just me, but I get a performance increase when I toggle the flight log on and off with F3. Its subtle. But it's there.

[Moesly_Armlis]

Everyone is experiencing this turbo effect because everybody uses physics time warp.

The Max Physics Delta Time per Frame is being altered the moment you use PTW and can only be reset from the settings page in the main menu.

My next question is what is the value of the Max Physics Delta Time per Frame after being altered.

From my experiments the MPDTF is being altered for each of the PTW and is probably not being restored when returning to normal. After any PTW the value is set to less than 0.03

The other issue is what impact does this have on simulation. I have used this turbo for vessels on the ground, in the air or space and have never noticed any thing.

Could someone provide technical details on the Max Physics Delta Time per Frame settings. What are the values for each of the physics time warp. Can anyone clearly explain how changing MPDTF value will alter how a large vessel's physics are altered or prevent it from "falling apart". Could this explain the sudden Kraken attacks?

From the Wiki:

• Max Physics Delta-Time per Frame
  
  • This setting determines how many physics calculations the program can run per game second. So, as the number increases, gameplay slows down, as more calculations happen every game second. It slows down, but not in the usual way. There is no lag, but now one second that passes on Kerbin takes more than one second in real time. This setting is good for keeping very large ships from falling apart. Additionally, in settings.cfg, this value can be increased beyond the range of the GUI slider.
    
  • Default: 0.04 sec
    

An experiment on the range of MPDTF could provide some useful information. The physics time warp is not to be used and the Max Physics Delta Time per Frame changed from the settings page.

My initial thoughts are to measure fuel consumption at high air speed conditions and use five different settings of the MPDTF.

What else should be investigated?

[metl]

I've noticed this. I never realized what was going on, but when I start the game for the first time, my first launch or two would stutter. Eventually everything would smooth out and run great. I will try to simulate this test and see if the physics time warp seems to be the culprit.

[Deddly]

Not everyone uses physics time warp. Those of us who spend most of our KSP time designing planes, boats or cars will very rarely use it because it will adversely affect the performance of the craft, especially if the wings fall off the plane or the car drags its engine along the ground...

Next time I load KSP I'll definitely try this.