Lag in ksp

[SpacePirat]

so i have actually a very stable laptop with 8 gigabyte of ram ad a intel core i7 processor but ksp lags very much. even if i play in the lowest graphic settings and without any mods.

i tried to change some settings in the settings.cfg but theres no way to fix the lag

Pls help

Greetings from Austria

[The_Rocketeer]

Welcome!

Does your graphics processor use system ram or dedicated video ram? Do you use the laptop for any other games?

Have you tried using a performance monitor while running the game to see which system resource is tapping out? Windows has one built into Task Manager.

[Gaarst]

KSP doesn't use multithreading (for now) and is very CPU demanding. Laptops CPUs, even high end, tend to have slower frequency than desktop ones; so that may be the reason why you're having lag.

The full specs of your PC would be helpful to determine if there's anything abnormal.

[Red Iron Crown]

Keep your part count as low as you can manage, nothing affects performance in KSP more than that.

You could also try turning your "Max Physics Delta" slider in the settings up a bit, that will trade some simulation accuracy for performance.

[Delta-Cheese]

I also have a laptop (and it is definitely NOT high end) and I can run KSP fine without lag (with max graphics), unless I do something a little to big.

It is a Dual Core 1.8Ghz i7-4500U (please don't laugh) with 8Gb ram, my graphics card is Geforce GT 750M. Compare my PC with yours, if it is better there is something wrong.

Maybe you need to update your graphics drive? Go to Nvidia (http://www.nvidia.com/Download/index.aspx?lang=en-us) and dowload a drive, I think that may be your problem.

Edited October 18, 2015 by Delta-Cheese

[Wild Cobra]

Yep.

I took mine off of default, and went to max graphics, pixels, etc, as well on both my computers.

If what I read is correct, and my interpretation, the limitations are how fast a single core can process the limitations of the unity 4 graphics engine in the game, which only handles 32 bit.

[Snark]

One thing that would be very helpful would be to narrow down whether this is a CPU problem or a graphics problem.

Indications that it may be a CPU problem: if it's most noticeable with high part-count ships, but doesn't change much based on what you're looking at (camera angle). If Task Manager shows your CPU pegged. If turning on time warp (which shuts down physics simulation) makes the framerate go way up..

Indications that it may be a graphics problem: if what matters most is what you're looking at. Here's a simple test: put a small, very simple (with few parts) satellite into low Kerbin orbit. Now try looking up at the satellite from below (so that the backdrop is just naked space, and you can't see Kerbin at all), and compare that with looking down on the satellite from above (so that Kerbin's surface fills the background. Does that make a big difference in your speed such as framerate? If it does, this is likely a graphics problem.

[sal_vager]

It's the physics, rather than graphics, that slows KSP down.

While Unity4 is multicore and does spread a few things to your other cores such as control I/O and audio, the physics engine is not able to divide the work across your cores.

This will change with 1.1 and Unity5, which can break up the physics into separate tasks for each game object to which physics apply, and divide these tasks amongst your cpu cores.

No one outside the Squad office knows how much of a boost to performance this gives yet, but they have hinted that it has improved the speed of KSP.

Unity5 will also make a 64bit Windows version of KSP viable as the Unity editor is 64bit, and issues that appear in the 64bit builds can be reproduced in the editor.

Currently Windows KSP is 32bit-only for this reason, it's complete hit or miss as to whether a 64bit bug can be fixed in Unity4 because you can't see what you're doing.

Luckily, the Linux 64bit builds worked, and Unity5 also allows for 64bit OSX builds for the first time.

[Snark]

It's the physics, rather than graphics, that slows KSP down.

That's often the case, yes, but not universally. It depends on the machine as to where the bottleneck is.

I was playing KSP on a machine with a plain-vanilla, slightly elderly graphics card, and was getting really crappy framerate whenever something pretty was on the screen, e.g. looking down at a planet from low orbit. I was having to keep my camera pointed at boring deep space if I wanted a decent framerate. Then l went and got a better graphics card, and the experience was much improved.

A lot of otherwise powerful laptops have very mediocre graphics cards in them, unless they're specifically kitted out as a gaming rig.

Graphics and CPU both van be bottlenecks. It depends on a lot of things, including hardware, graphics settings, and ship design (and thank you for the well-reasoned and -explained detail about CPU issues!).

Edited October 18, 2015 by Snark

[Wild Cobra]

I also have a laptop (and it is definitely NOT high end) and I can run KSP fine without lag (with max graphics), unless I do something a little to big.

It is a Dual Core 1.8Ghz i7-4500U (please don't laugh) with 8Gb ram, my graphics card is Geforce GT 750M. Compare my PC with yours, if it is better there is something wrong.

Maybe you need to update your graphics drive? Go to Nvidia (http://www.nvidia.com/Download/index.aspx?lang=en-us) and dowload a drive, I think that may be your problem.

I have the 750M in my laptop, and it doesn't even seem to work hard with KSP. It also has the i7-720HQ 2.6/3.6 ghz.

It's single thread rating of 1578 is nothing to sneeze at. My single thread is only 23% faster at 1947.

My tower uses the i7-4790 which has a single thread rating of 2293! It was a $950 tower new, and my laptop was $1,050.

[Clipperride]

With my rather ageing laptop, it was the graphics settings that removed most of the lag. Now, apart from launching when the exhaust is interacting with the ground and when a station gets up above ~450 parts, the game clock is usually green. It may take a little trial and error to tweak the settings to something your set up is happy with, but it's worth an hour or two tinkering with them in turn.

[Wild Cobra]

I think I doubled my CPU capacity for KSP.

I disabled hyperthreading in my BIOS. Since KSP uses only one thread for unity, it only had half of a core available. Now it has a full core available.

KSP was using 15.04% - 15.05% of my CPU capacity according to my resource monitor. 12.5% of that would be fully utilizing one logical core. Now that I made the logical cores equal to the physical cores, it is utilizing above 27% of the CPU. 25% plus a little extra instead of 12.5% and a little extra.

My large orbiting station is now at about double the frame-rate!

[Curveball Anders]

I think I doubled my CPU capacity for KSP.

I disabled hyperthreading in my BIOS. Since KSP uses only one thread for unity, it only had half of a core available. Now it has a full core available.

KSP was using 15.04% - 15.05% of my CPU capacity according to my resource monitor. 12.5% of that would be fully utilizing one logical core. Now that I made the logical cores equal to the physical cores, it is utilizing above 27% of the CPU. 25% plus a little extra instead of 12.5% and a little extra.

My large orbiting station is now at about double the frame-rate!

Hyperthreading doesn't lock half the core out in any way, shape or form.

Extensive testing with strict single thread software can show as much as a whopping 1-2% in performance gain.

But only for that process, and with just system default processes running disabling HT actually slows the computer down.

So if you get that kind of performance boost in KSP then you should look for some other explanation.

Edit:

Here's just one of the tests that's published over the years.

https://semiaccurate.com/2012/04/25/does-disabling-hyper-threading-increase-performance/

Edited October 19, 2015 by Curveball Anders
added link

[cantab]

Hyperthreading itself indeed doesn't work like that. However bad scheduling by the operating system can put two threads needlessly on the same physical core. If you have an older OS or one that you haven't kept up-to-date this is more likely, but it shouldn't happen with a modern and updated Windows or Linux OS.

[soulsource]

Servus!

It might eventually help if you could describe your issues in more details. There are several things in KSP that can cause lag or stuttering, and not all of them can be helped.

Let's start with the most annoying thing, the Unity garbage collector. Every few seconds it'll scan for no longer needed memory in order to free it. This causes noticeable lag on almost any hardware. In the worst cases (typically with buggy mods) it can cause the whole game to freeze for half a second or so.

The next thing is the physics engine. This is the part of the game that will cause issues on processors with low single thread performance. To have it cause troubles on an i7 one has to build massive ships, comprising hundreds of parts. I'd bet on this if your laptop had a Pentium, or an AMD A-Series CPU, but not with an i7...

Another reason could be graphics. Here the question is, what series of i7 we are talking about. Recent i7 variants have rather powerful graphics (which should run KSP in Full HD at maximum settings easily), but that has not always been the case. Yet, KSP is not very demanding when it comes to GPU utilization, and with lower graphics settings it should run even on Sandy/Ivy Bridge graphics units.

Another thing noone mentioned yet: Are you playing on Windows or Linux? If you play on Linux and your Laptop has an nVidia GPU (in addition to the integrated Intel GPU), make sure to use the proprietary nVidia drivers, and make sure that the dedicated GPU is actually enabled.

[Wild Cobra]

Hyperthreading itself indeed doesn't work like that. However bad scheduling by the operating system can put two threads needlessly on the same physical core. If you have an older OS or one that you haven't kept up-to-date this is more likely, but it shouldn't happen with a modern and updated Windows or Linux OS.

I know it doesn't lock half a core out. What is does is keep they system from splitting a core into two logical cores.

If an application is can only use one core to operate in, and is utilizing that core's capacity 100%, then it does indeed help. Maybe it's the doubling of the cores cache? I only know my results improved, and recall doing it to automation equipment PC's at work to greatly improve them about 5 years ago.

I will restart and change my BIOS twice, and put up the same orbital scene with my lander of 1395 parts, and a docked refueler. I will snapshot my resource monitor. I will do this within an hour. Meanwhile, I have other things to do before I restart my system.

Edited October 19, 2015 by Wild Cobra

[Alshain]

Hyperthreading itself indeed doesn't work like that. However bad scheduling by the operating system can put two threads needlessly on the same physical core. If you have an older OS or one that you haven't kept up-to-date this is more likely, but it shouldn't happen with a modern and updated Windows or Linux OS.

To be more specific, when cantab says "modern". He means Windows Vista or higher or equivalent Mac/Linux. Pretty much any OS that came out after Dual Core processors were in development. Even Windows XP is fine if you have it up to date.

[Wild Cobra]

Here are my results:

http://imgur.com/a/xv1sw

CPU usage went from 14.885 to 29.35% in these snapshots.

[Red Iron Crown]

Looks like near-identical utilization, it's just reporting it differently.

[Wild Cobra]

Looks like near-identical utilization, it's just reporting it differently.

well, even if the reporting is different, I see a difference.

When an application is utilizing a core to 100%, isn't it possible it is also utilize 1/2 or all of the core cache too, making a larger difference than that test on the generation 3 processor?

I downloaded the Passmark system performance tester yesterday. Maybe I'll see what it says, but doe it fully utilize a logical/physical core in testing?

[Red Iron Crown]

Modern Intel CPUs have per-core L2 cache, the same amount will be available to a single core whether HT is enabled or not. (HT "cores" aren't really cores, but telling the OS they are makes scheduling tasks for them easier.)

When you say you see a difference, what do you mean exactly?

[Racescort666]

Well, if we're all chiming in about our dismally underpowered laptops we use to play KSP, I have an HP with 2.4/2.0GHz AMD Quad-Core A8-6410 APU and AMD Radeon HD 8210 graphics. Not terrible. I get about 6-12 fps during regular gameplay which varies a little depending on part count. I occasionally get as high as 15 fps with like 10 part satellites but average ships and space planes in the 50-200 part range I'll get around 8 fps. It's noticeably slower during atmospheric flight.

I have been holding off building a new computer out of some leftover desktop parts I've got until 1.1 comes out. Hopefully the performance boost brings me from the verge of hating myself to merely being cautious about high part count space stations.

[Wild Cobra]

Modern Intel CPUs have per-core L2 cache, the same amount will be available to a single core whether HT is enabled or not. (HT "cores" aren't really cores, but telling the OS they are makes scheduling tasks for them easier.)

When you say you see a difference, what do you mean exactly?

I have less lag for my over 1,000 part count builds. A higher frame rate. I should measure it, but I'm busy doing other things. I can tell there is a difference. My camera angle speed change for my lander with 1,395 parts in orbit, when it's doing those part stress calculations, is faster.

At least the CPU in my computer is in 5th place for single thread benchmarks:

http://www.cpubenchmark.net/singleThread.html