Server CPU Performance vs tickrate @ Ubuntu 18.04 Engine 4.51

Discussions about Servers
Post Reply
LuckyDog
Novice
Posts: 5
Joined: Wed Jul 22, 2020 4:06 pm

Server CPU Performance vs tickrate @ Ubuntu 18.04 Engine 4.51

Post by LuckyDog »

Hi there,

this is more or less a question i got stuck with over the past couple of months.

I am running 3 Servers
1x VPS 6 Core 16 Gigs of Ram
1x Dedicated Server Intel Xeon E3-1230v3 4Core / 8 Threads, 3,3 GHZ, 32 Gigbyte DDR 3
1x Dedicated Server Intel Xeon E5-1650v3 6Core / 12 Threads 3,5 GHZ 128 Gigabyte DDR4

All Servers run Ubuntu 18.04 LTS and each one of them hosts 2 Instances of Tactical Ops 3.4 which is a mod for UT99.

What i have seen is, that the VPS seems to spread the Load of the UCC-BIN over multiple vCores where the dedicated servers just simply use one core to run the thing.
Where the VPS produces minor glitches during heavy load with 30 players, the dedicated servers seem to do the job just fine (of course theres gonna be lagg during high pinger calculation etc.)

But what i dont understand is: How the hell can a 15 Year old Server version (I believe thats what it is) cook one core of my Xeons at roughly 80 to 90 percent of load while having the tickrate set to 60 with 30 players?

I know that on UT some weapons rely on a stable tickrate, this doesnt seem to be the case on Tacops.
Even if i run the thing at 120 ticks, with lets say 18 to 20 Players no issue but as soon as playercount rises the CPU starts to stall because it cant deal the amout of calcuations.

How the hell did old Athlons deal the thing back in the day? Cant be that servers only allowed 4 players right?

Any ideas here?


Thanks and great Week all,

LuckyDog
User avatar
esnesi
Godlike
Posts: 1018
Joined: Mon Aug 31, 2015 12:58 pm
Personal rank: Dialed in.

Re: Server CPU Performance vs tickrate @ Ubuntu 18.04 Engine 4.51

Post by esnesi »

LuckyDog wrote: Mon Nov 09, 2020 10:19 pm
What i have seen is, that the VPS seems to spread the Load of the UCC-BIN over multiple vCores where the dedicated servers just simply use one core to run the thing.
Where the VPS produces minor glitches during heavy load with 30 players, the dedicated servers seem to do the job just fine (of course theres gonna be lagg during high pinger calculation etc.)

But what i dont understand is: How the hell can a 15 Year old Server version (I believe thats what it is) cook one core of my Xeons at roughly 80 to 90 percent of load while having the tickrate set to 60 with 30 players?
Have you ever seen it functioning on a VPS, or localhost, without the core overload issues?
Iam not sure if TF (TickFix) is a thing for TacticalOps, it might be compatible.
But i've read somewhere that without it, it can cause high load issues as well in some situations.
User avatar
sektor2111
Godlike
Posts: 6410
Joined: Sun May 09, 2010 6:15 pm
Location: On the roof.

Re: Server CPU Performance vs tickrate @ Ubuntu 18.04 Engine 4.51

Post by sektor2111 »

TickFix as far as I know it's not for Linux.
SC]-[WARTZ_{HoF}
Adept
Posts: 426
Joined: Tue Feb 21, 2012 7:29 pm

Re: Server CPU Performance vs tickrate @ Ubuntu 18.04 Engine 4.51

Post by SC]-[WARTZ_{HoF} »

Here is some info Higor gave me when I was asking about Client FPS, MaxClientRate and Server TickRates. Hope this helps.
Higor: when it comes to more competitive games, it depends on the player count
Higor: 1v1 best at 120TR
Higor: Siege at 55-65TR
Higor: iCTF at 80-100TR
Higor: DOM somewhere between 40 and 60TR
Higor: ctf 20 clients keep it at 40-45TR
Higor: dm 16 somewhere 50-60TR

Higor: a client playing at 180fps hits someone at a certain spot
Higor: if the server is running at 30TR...
Higor: there's roughly a 17% chance that the position the client hit the other isn't anywhere on the server

Higor: create a server, put 2 players in it
Higor: make them stand next to each other
Higor: hit F6 on each player
Higor: start running with player A in circles, check how bandwidth usage changes on player B's F6 stats
Higor: that varies with tickrate
Higor: if your MaxClientRate is 20k, then figure out a way that the BASE + PER_PLAYER_BW * (maxplayers-1) never goes above that
Higor: if a running player adds 2kb/s
Higor: then you get (base is 3kb/s i think) this:
Higor: (20kb/s - 3kb/s) / 2kb/s = current supported players
Higor: about 7 in this case
Higor: increasing/lowering tickrate will alter both base and per_player_bw
Higor: uk Siege is at 25k rate
Higor: and caps at 17 players i think
Higor: it's set to 22 lol
Higor: uk Instagib do cap at 11 players or so
Higor: they're set to 12
Higor: they're at 20k rate and 85TR

Higor: so basically, you pull out a calculator, decide how many players you want
Higor: and start trying tickrates with the 2 dummy players until you reach a formula that barely hits the net rate cap
Higor: BASE + PER_PLAYER_BW * (maxplayer-1)
Image
Image
Image
Post Reply