Private server lag

Conan Exiles General Discussion Feedback
1

Hey folks, currently running a private server on a full dedicated machine, it’s the only server running on the entire box. We maxed out at 32 players, I slowly increased player slots. Even at lower player caps, the server like many others have insane lag, rubber-banding and sliding thralls. The server is chewing through over 20GB+ of memory. On a test server with just me on it, it was at 10GB+

I have tried a few things such as different tick rates, lowering player cap, clearing up old bases and I’m now going to be trying disabling Living Settlements temporarily to see if its any help.

I see the issues on the official servers are pretty much the same and on many other private servers.

Does anyone have any tips to help lower it whilst we wait for stability patches?

EDIT: disabling Living Settlements helped, but very little. It has maybe saved like 500MB of memory and some CPU processes but still doesn’t help with lag or rubberbanding.

2

It would be interesting to see if you could get everyone on the server to shut off every chest/bench on the entire server to be accessed by the new system that allows you to build directly from them, and at the same time as you say living settlements off and the same with bars, that i would be very interested in hearing the results from.

3

What I can say is that turning off Living Settlements has made a difference, albeit very very little.

4

I’m seeing something very similar on my side.

I tested a clean Conan Exiles Enhanced dedicated server on a private Windows Server 2022 machine, with no mods loaded, just to remove mods from the equation.

Hardware:

Intel i9-10900KF
128 GB RAM
20 logical processors
Dedicated machine
Server tick rate: 30
One local player connected during the test

The real server process was:

ConanSandboxServer-Win64-Shipping.exe

What concerns me is that, even though the server process creates many threads, the active load still seems to concentrate mostly on one dominant thread.

With the server idle / one player standing still, the main active thread was around 17-19% of one logical core, while most other threads were near 0%.

Then I spawned several NPCs from different factions and let them fight each other while I moved around. During a 30 second PowerShell thread sample, I got roughly:

CPU equivalent total: 114.9% of one logical core
Total system CPU usage: 5.74%
RAM usage: 8619 MB
Threads: 36

Top active threads:

Main active thread: 84.22% of one logical core
Second thread:      21.35%
Third thread:        8.28%
Most other threads: near 0%

So from the outside, it still looks like the dedicated server is mainly limited by one heavy game/server thread, with some helper threads doing smaller work. That would explain why the machine can still have plenty of unused CPU overall while players experience rubberbanding, delayed combat and desync.

I’m not claiming this proves the whole replication/netcode system is unchanged, but it does look like the server is still very main-thread bound under gameplay load.

I also tested disabling Living Settlements, lowering corpse lifetime, reducing NPC caps, keeping tick rate at 30, and avoiding higher building replication distances. These may help a little, but they don’t seem to solve the core issue.

If anyone has found server settings that actually reduce rubberbanding/desync in Enhanced, especially on private dedicated servers, I’d be very interested.

5

I’m running a private server on HostHavoc with 20GB of RAM. All the stats look good server side. We rarely go above 70% of RAM usage, and not more than about 15% of CPU usage. but anytime we get above 20 players on the server we have to run hourly server restarts due to unplayable lag. We’ve disabled living settlements and today I limited the number of followers, but we’re still getting major desync and lag issues once we hit 20 players on the server. Does anyone have any other ideas?

6

hosthavoc is almost the same as GPortal very cheap and bad hosting,

but yeah game have problem with lag, devs need to fix it

7

It’s been years.

Best to play other games.

8

Hey! Thanks for all that detail! You and I are surely in the same boat. No matter what I try it will just continue to spike. I think it is a waiting game for Funcom to do something about it. :man_shrugging:

9

Thanks for the reply. Yeah, that is exactly what I was afraid of.

I’m still going to keep testing different server-side settings, but so far it really feels like most tweaks only reduce the symptoms a little rather than fixing the underlying issue.

If I find anything that makes a measurable difference, I’ll post it here. At this point I think the most useful thing we can do is keep sharing actual test results and server configurations, so Funcom has more concrete data to work with.

The worrying part for me is that the server can look almost idle in total CPU usage while one thread is already doing most of the work. That makes normal CPU monitoring very misleading.

By the way, one thing you could try, if you haven’t already, is increasing the server network rates in Engine.ini.

I’m currently testing this on my side:

[/Script/OnlineSubsystemUtils.IpNetDriver]
MaxClientRate=500000
MaxInternetClientRate=500000
NetServerMaxTickRate=30
InitialConnectTimeout=120.0
ConnectionTimeout=80.0

I would not recommend jumping to 60 tick rate yet, because if the server is already main-thread bound, that may make things worse rather than better.

In my case, I’m keeping NetServerMaxTickRate=30 and only raising the client rate limits to see if it helps with replication/network saturation without increasing tick pressure too much.

If you test it too, it would be interesting to compare results.

10

Thats actually a good idea to test, please let me know how it goes. I’m currently away at the moment and can’t rdp to the dedi, but I’d be interested to see what your results look like.

11

That’s exactly the main limitation on my side right now: I don’t currently have a large enough player base to test this properly under real load.

I can test that the server starts correctly, that the settings are accepted, that there are no obvious connection issues, and maybe compare basic behaviour with a few players. But the real problem only appears when there are many players close together, PvP situations, modded areas, bases, thralls/NPCs and a lot of replicated actors.

So my results will probably be limited unless I can test it during a proper populated session.

If you or anyone else with an active server can try the same values and compare before/after behaviour, that would probably give much better data than what I can gather alone.

12

I’ve had to lower my player cap and may have to lower it again and we’re thinking about limiting building size and even turning on minion population limits.

It feels like we’re being starved of fun and having to play a very basic game due to the stability and lag.

I was hoping the hotfix yesterday would have at least some update or message about the issues, but there was nothing, not even a “we’re still looking into it.” my other worry is that they’re going to focus only on Official server stability.

There is very little left of us to try at this point, but if I get time and when I’m back I’ll try shoot some results across to you. The sweet spot in terms of population of player cap seems to be 20 - 40, anything above 40 and it just makes things even worse.

13

I’ve noticed that most of the lag on our server has to do with actual building, whenever someone replaces a piece of their build the game still recalculates every single piece to check if something should break or not. This literally creates a 2-5 second lag spike depending on thee build size. But lets bee honest… no one wants to build a shack… the game pushes you to build big with their big stations and enormous map room.

Btw this issue was in legacy as well, same issues persist.

14

It’s looking more and more like many little things adding up and creating just a web of stability issues.

15

That building recalculation point is very interesting, and it might be worth testing something specific.

If the 2–5 second lag spike really happens when replacing/removing building pieces because the server is recalculating stability and checking what should collapse, then disabling building stability entirely could be a useful test.

I don’t know whether this will fully bypass the calculation or simply make the result always stable, so I would only treat it as a controlled test, not as a guaranteed fix.

For anyone willing to try it, make a backup of ServerSettings.ini first, then under [ServerSettings] try:

StabilityLossMultiplier=0.000000
BuildingValidationEnabled=False

Then restart the server and test the same action again on a large base: replace/remove pieces and compare whether the 2–5 second spike is reduced or gone.

If the spike disappears or is heavily reduced, that would strongly suggest that building stability recalculation is one of the bottlenecks. If nothing changes, then the spike may be coming from another building-related system: replication, land claim, ownership checks, placeables, save/update logic, or something else.

This is also why I think this issue is not only about bandwidth or MaxClientRate. Network tuning may help with replication and client update rates, but a global 2–5 second server hitch when modifying a building piece sounds more like a server-side blocking calculation, possibly on the main game thread or another critical synchronous path.

To be clear, I am not saying Conan is literally single-threaded. But it does feel like some important server-side systems still behave as blocking operations. When that happens, adding more CPU cores does not necessarily help if the bottleneck is one heavy serialized task.

This is also what worries me about Enhanced. The UE5 port is promising and visually much better, but many server-side bottlenecks seem to behave very similarly to Legacy. UE5 by itself does not automatically fix building stability, actor replication, AI, thralls, inventories, huge bases, or modded server load if those systems are still using mostly the same underlying logic.

So from the outside, it feels like Enhanced improved a lot on the visual/client side, but some of the deeper dedicated-server problems may still be inherited from Legacy.

For private RP/PvP servers, especially modded ones, the real problem appears when many systems stack together:

  • 30–40+ players,

  • PvP or many players close together,

  • large bases,

  • building edits/replacements,

  • many placeables,

  • thralls/minions/NPCs,

  • modded areas and systems,

  • and many replicated actors.

At that point, it feels less like one magic setting will solve everything and more like a web of server-side stability/performance issues.

If anyone from Funcom reads this, even a small acknowledgement that private/modded server performance is being looked into would help a lot. It would also be very useful to know whether systems like building stability recalculation, actor replication, and main-thread/server hitches are being profiled specifically, not only Official server stability in general.

16

I will try this as soon as I’m able, but if someone else is able to test before me or Sonidu, please let us know! I’m locked in now, determined to get some sort of stability. It’s hard for me to test as my server is normally always full at 54/54 players.

I really hope Funcom help with private server stability too otherwise the game will just continue to die like it is now! Please listen to your players and hosters! We want to keep the game going, we want to create content and we want to have fun!

17

Will be interesting to see how this works out.

18

:brazil:
Conan Exiles servers pc PC Discussion > PC Servers & Clans
1
Bem, se seu servidor estiver em seu próprio computador PC, ou seja no mesmo local em que tem o jogo de conan exiles, deverá apenas atualizar o game e jogar e verificar os arquivos do jogo…

2
Se está fazendo um servidor ON-LINE terá que atualizar o HOST para devido a uma correção dos servidores da UE5

3
Para um servidor ON-LINE você poderá instalar a versão mais atualizada que poderá resolver

4
Se está usando o servidor no mesmo computador em que está jogando, recomendo que jogue em outro computador deixando uma máquina dedicada para o servidor ON-LINE

19

I haven’t tested the “BuildingValidationEnabled” setting, but we did try disabling stability, the server still does recalculate even if it’s off. Is “BuildingValidationEnabled” real? like is it a setting we can modify?

20

That is very useful information, thanks.

If you already tried disabling stability and the server still recalculates, then that probably means StabilityLossMultiplier=0 does not bypass the stability/building update logic. It may only make the final result “stable”, while the server still runs the recalculation path.

About BuildingValidationEnabled: yes, it is a real setting in the official Enhanced server settings documentation.

From the docs, it causes pieces that load with 0 or invalid stability to be removed. It is off by default, and it also says it relies on StabilityLossMultiplier > 0.

So I may have been too optimistic about that setting. It sounds more like a validation/removal setting for invalid pieces, not necessarily a way to disable runtime recalculation when replacing/removing building pieces.

In other words, this test is still useful:

StabilityLossMultiplier=0.000000
BuildingValidationEnabled=False

But if you already tested stability disabled and still got the same spikes, then the bottleneck is probably deeper in the building system: stability graph recalculation, building topology updates, ownership/land claim checks, replication, save/update logic, or some other synchronous building operation.

That would also reinforce the idea that this is less of a network-only problem and more of a server-side blocking calculation issue. Network settings may help with replication limits, but they will not solve a server hitch if the building system itself is blocking the game thread or another critical path.