PSA: Performance Issues? Try Verify Integrity of Game Cache

[dichebach]

I've got a pretty stout rig. Game has always had low-frequency performance issues of various sorts even on lowest "Oh My Okran! My Eyes Are Bleeding!" setttings . . . After playing enough to get the gist of the game, I reset all my graphics up to maximus and restarted with a crew of 5 (I have around 200 mods installed). Everything was fine, only low frequency issues until I got my outpost in Hub up to about Research Bench II stage and as I was bringing the original 4 back from an expedition while the new 4 were puttering around the base.
Started getting major slow downs when I would alternate between these two remote squads. Then crashes, repeatedly. Did some searching on how to report bugs and found some old posts by Hatsune and in one of those she suggested doing the "Verify Integrity" thing using Steam's library functionality. I also turned down all the graphics settings from 100% with Mega to 90% with "High." And that seems to have fixed my problems for now.

[Rhah]

I have had the same experiences, you will hit some more slow downs at least I did. Then when I created a base, i noticed a big hit. I basically kept decreasing settings except object and view distance I have kept them maxed out. I only get maybe 2 or 3 crashes, but game was slowing down alot. I added the mod for clean and speed and then I was able to put back to ultra high settings but about 50 days into building base and messing around in swamp i had to keep lowering settings im about medium now for about 30 days no slow downs really. Just everynow and then if playing many hour, need to exit game a reload. It might have a bit of a memory leak or not releasing resources properly.

Oh and import seems to clean a bit up as well. although I only have import when I hit buggy situations

[bunny de fluff]

I would recommend these:
enable fast zone hopping
reduce texture quality
play in borderless mode

[Rhah]

oh I forget about borderless mode. that speed up game alot at the beginning. for some reason on alt tab it would load extra resources for native resolution and the resolution game kept mucking with.

basically with fullscreen your using a bunch of extra resources for the 2 resolutions game was fighting with.

[dichebach]

Yep, I'm already using borderless mode.

[dichebach]

One other thing I will add: if you use mods, locate your mods.cfg file, you can open it with any text editor application. It will generally be in the . . .\Steam\steamapps\common\Kenshi\data directory. Open that up and save it with a new name like "modsCOPY.cfg" or "modsListSaved.cfg."
Reason for this are multiple, but most notable: when you let Steam update the game it will wipe out all your mod activation and reset the load order. In terms of manipulating the load order, it is much easier to work with it in the text file. Basically, just exit the game, open the mods.cfg file edit your load order and save and that will be what the game loads as the load order.
About fast zone hopping, I thought that that HARMED performance, no? I'll check it out though.
I've got textures on about one notch below max, actually no there are several texture settings (some they call decal). I think I have at least one of those at max, but the rest at ~90%. Will keep it where it is for now, and dial down stuff progressive as I see it degrades.
Shame the game has progressive memory management issues, but I cannot fault them for it. Given the complexity of the app I have no idea how many instances it is shuffling around at any given time, tens of thousands when one considers models I suppose? In my little bit of experience developing it was a game with, on the order of 10k structs each with anywhere from 1 to 500 instances in play at any given time (and most of those structs had child structs of child structs, etc., so these were dense data structures). The game had almost zero performance issues, BUT! turn-based, no physics, no 3D graphics, old-school C architecture--which was probably both a pro- and a con- for performance. I still don't know how that one guy got that thing to run on multi-core given the engine was from the 1980s! 3D graphics and physics seems to be the bane of good performance in general, which is a shame because they enrich games immensely. Imagine Dwarf Fortress if it tried to aspire to even half of the 3D naturalism of this game!

[Rhah]

for the zone hop feature I didn't really noticed a difference. I still zone less than 1 second either way. but with zone hop off is use 3.9 gb and with on memory usage jumps to 7.9gb (i think it was discussed few days ago in forum).

btw engine only uses 1 cpu. they are working on kenshi 2 with will be mulit core (im assuming multi threaded). but that is our main issue with performance I think as we are only using 1 core.

and the code they did is great for running so smooth on 1 cpu.

I'll try to find the other thread and link it to this post.

thead on zone hop and i did some monitor of system resources for about 30 minutes while playing.

https://steamcommunity.com/app/233860/discussions/0/1640917196992438466/

oh and thank you the info on mods cfg. Ill do that now I defiantly don't want a update to cause more work than it should.

[dichebach]

Yeah, I had heard that it only uses one core. If they really want to try to fix their existing code, they should PM me, I can probably get them in touch with the dude who figured out how to get that old C source game (an early aughties game, but the engine was created in like 1986 . . . you can work with it as a C++ solution, but it is effectively all C code, not a single class in it if I recall) to run multiple cores. It might well be a different matter given Kenshi is a 3D game, but maybe not. My knowledge of programming does not yet penetrate that far!
ADDIT: and it IS a great game, great app for a small Indie group, and it does run remarkably well given it is bottlenecking on one core.
However, with that said: it may well be that solutions which are not very time-consuming or costly could be fairly readily at hand for getting the app to utilize multiple cores. Like I said, I don't personally have the expertise to advise them on this, but I have an acquaintance who is a colleague of my former programming mentor who (supposedly) got that old engine to run multiple cores. So it might be doable.

[Rhah]

its pretty intensive from my understanding, but I am from a configuration management enterprise level software, not gaming.

Oh sorry I left out after kenshi 2 released the new engine that is created it will be back ported to kenshi 1.

[bunny de fluff]

c only has structure, but class is just fancy structure with pointers and function calls

[dichebach]

Originally posted by Rhah:

its pretty intensive from my understanding, but I am from a configuration management enterprise level software, not gaming.

Oh sorry I left out after kenshi 2 released the new engine that is created it will be back ported to kenshi 1.

Ah, well there you go! No point in them ♥♥♥♥♥♥♥ around with random people from the internet then

[Rhah]

oh haha nothing like that.

just there is a ton of issues trying to modify an engine to do something it was never designed to do and even if you can get it stable and working with the client it will probably never function as well as a newly coded engine. with trying to force fit the engine your looking at a lot of work plus probably a ton of problems with the client (or interface).

in my experience you will save yourself tons of time headaches and problems just by starting from ground zero. something like multi thread/core this a major overhaul that pretty much affects the entire engine.

now with going to a whole new engine for kenshi 2 is also another thing.
kenshi 2 client will be coded to work with the new engine and hopefully in doing so they are strictly thinking of kenshi 2. because if started thinking of how it will fit kenshi 1 as well in the design and creation can limit kenshi 2 in its end result.

trying to back port it will not be easy either but will be way more easy than trying force into something was not designed to do, working through all the problems, and then try to force fit client and all its issues can take many years of headache.

[dichebach]

I know what you mean! That project I worked on, that old solution has something like 1.5 million lines of code in it, and the project we were exploring was left dead on the road exactly for the problems of "reverse configuring" you are talking about! Legacy code is generally a nightmare, and it seems that, even if a programmer has 15 or 20 years of thorough experience, AND has specific experience with a set of code for years of work: once they set it down and leave it behind for a while, and if it is a large and spaghetti-like code base, then it will be a challenge even for them to come back to it.