Mods and performance.

[eberkain]

I started my new science mode playthrough and after a while I just couldn't deal with the terrible performance.   So I went through checked every mod with a simple test to see if it have a performance impact and looked at my graphics settings and got things optimized by dropping some mods and tweaking some stuff.   Currently I can launch a rocket with 45+ FPS the entire flight, thanks to the Hangar mod the complexity of the payload doesn't hit performance anymore.  Once in orbit things are great and running at 55-60 most of the time.    

I built a little remote lander that could transmit science and put 4 of them on a hauler and sent several to the mun and minmus,  At about 150 parts those hit my FPS to about 30ish.   

I now have a small station in orbit of the mun which is 3 vessels docked together.  Probably less than 100 parts total.  FPS is about 12 anytime I'm looking at it.    Just can't deal with it. 

I went into a completely stock install and stuck together several of the stock vessels and cheated it into orbit.  At least 300 parts.  60 FPS. 

So clearly some of the mods I'm using are killing my performance.  I'm going to revisit my mod list again and trim things down to only the bare essentials and do some more testing to see what mods are the culprits.   I need a good reliable way to benchmark performance as I switch mods around, what are your suggestions?  What mods in general will drag down performance?  I am almost tempted to just go back to a completely stock install.  

[eberkain]

I think I have things setup, here is how it played out.  

I did some thinking about exactly what I wanted from the playthrough and drastically shortened my mod list by dropping off alot of stuff that makes the game harder and more complicated.  I'm really just interested in building bases this time around.  

I then setup a save with a 350 part vessel in orbit of the mun.  One by one I added a mod loaded up the save and waited a bit to let the performance stabilize and the recorded the results.  Stock would give me 55 FPS, now I have 55 mods installed at its dropped to 45 FPS which I am happy with.  My initial test with my full list was giving me 10 FPS.  I discovered several mods that hit performance pretty hard in that situation that I always use. 

[Krakatoa]

On 2/23/2018 at 2:32 AM, eberkain said:

I think I have things setup, here is how it played out.  

I did some thinking about exactly what I wanted from the playthrough and drastically shortened my mod list by dropping off alot of stuff that makes the game harder and more complicated.  I'm really just interested in building bases this time around.  

I then setup a save with a 350 part vessel in orbit of the mun.  One by one I added a mod loaded up the save and waited a bit to let the performance stabilize and the recorded the results.  Stock would give me 55 FPS, now I have 55 mods installed at its dropped to 45 FPS which I am happy with.  My initial test with my full list was giving me 10 FPS.  I discovered several mods that hit performance pretty hard in that situation that I always use. 

Would you care to share what those might be? I imagine there's a few folks around the forum on older comps that could know what to avoid, or it may even be a bug that needs to be sorted, or even the ones you're using with little to no impact.

[Kerbal101]

On 19.02.2018 at 3:59 PM, eberkain said:

I need a good reliable way to benchmark performance as I switch mods around, what are your suggestions?  What mods in general will drag down performance?  I am almost tempted to just go back to a completely stock install.  

Speaking as a player...

1. The good method I know is to make a folder outside of gamedata, then move all elements, except "Squad" and ModuleManager.* there.  Afterwards, for each step, add the mod - either one by one; or add a half of original amount(paying attention to dependencies) and then if regression is detected, move the half after half outside of the gamedata. The "half" method is faster than one-by-one.

2. There are several bottlenecking possibilities, if KSP hits one of them - it will influence framerate. Mods that add contracts, cause background polling, modify drag calculations, vehicle distances, texture and geometrical complexities, increase graphic effects in all possible ways(shaders, smoke fx, extra models, extra textures etc), also shadows and light sources. Of course, also, open docking ports.

Some mods may actually increase memory usage,yet also increase framerate(!). This is happened to me with Stock Visual Terrain on surface of Val. I was very surprised! Also I got no penalty for Terrain Detail neither on  Kerbin, nor on Mun - but visually quality went down. I also recommend to test how stock terrain detail, scene rendering quality and aerodynamic effects influence the fps, when running fully stock.

This is a good way to find mods that affect performance. In general, the way current KSP works, single-thread performance of your CPU plays major role. Mostly because bottleneck happens in draw calls, which are connected with geometry complexity and amount of models. Here welding mod will not help, because it does not weld the models themselves. But if you happen to have a lot of joints - then either part-mod or welding is a way to go, as it would be physics calculation that drags fps down. However using mod which gives larger capacity parts (reducing total part count) or models with less polygons - should improve performance. "Should" because the OS environment also matters - the graphic card driver, graphic card itself, CPU scheduler of OS kernel, - they all contribute.

Then there is mod method and text-edit method to reduce the underwater details of the Kerbin at certain altitude, which will help performance.

[eberkain]

3 hours ago, Krakatoa said:

Would you care to share what those might be? I imagine there's a few folks around the forum on older comps that could know what to avoid, or it may even be a bug that needs to be sorted, or even the ones you're using with little to no impact.

I always maintain a mod list when I play games.  https://docs.google.com/spreadsheets/d/1iznWcub3iGpN_Xvd-UFHVwMrxjvcaKQiGOZvAUsWwRs/edit?usp=sharing

[linuxgurugamer]

1 hour ago, eberkain said:

I always maintain a mod list when I play games.  https://docs.google.com/spreadsheets/d/1iznWcub3iGpN_Xvd-UFHVwMrxjvcaKQiGOZvAUsWwRs/edit?usp=sharing

No, he asked which mods you found which had a severe performance impact. Which would be more useful as a list of mods to avoid rather than a list of mods which are efficient

Edited February 26, 2018 by linuxgurugamer

[eberkain]

56 minutes ago, linuxgurugamer said:

No, he asked which mods you found which had a severe performance impact. Which would be more useful as a list of mods to avoid rather than a list of mods which are efficient

At the bottom of the spreadsheet are the mods that I ended up removing because of performance problems with notes on the performance I observed.  But this is all dependent on my hardware and my system/game settings.   So I'm reluctant to post a list of mods and say these mods are bad, avoid them.  

[linuxgurugamer]

16 minutes ago, eberkain said:

At the bottom of the spreadsheet are the mods that I ended up removing because of performance problems with notes on the performance I observed.  But this is all dependent on my hardware and my system/game settings.   So I'm reluctant to post a list of mods and say these mods are bad, avoid them.  

But you can say that with this specific hardware and settings, these mods don't perform well.