True or false: A client FPS can never exceed that of the server?

[Tankbuster 1736]

Apologies for being here for 8 years and not knowing the most elementary of facts about this game and it's engine, but rather rashly, in clan teamspeak the other evening I said that 'client FPS can never exceed that of the server they are connected to' and someone asked if I was sure, and now I'm not actually sure that I was right.

Should I order my humble pie?

[Minoza 11]

I'm pretty sure you can. I mean, using "#monitor" command you can get server FPS info in chat, it's usually much lower than my framerate. Oh and, in case you try it out, it's "#monitor off" to turn it off. :p

[das attorney 858]

Hey TB,

As far as I know Arma 2 server FPS is max 50, and you can definitely exceed that as a connected client.

[ctops 10]

In this topic;

http://forums.bistudio.com/showthread.php?154908-Aiming-unsteadiness-(server-side-lag)

I'm new to the ARMA world but understanding that OP question is basically the same is ARMA 3's features (aiming, soldier movements etc.) all fully controlled by server-side client ?, as I don't see an installation for alpha I might presume the full release may be different ?

Älení na nejjednodušší jazyka

[Harzach 2515]

Server framerate is unrelated to client framerate in the most general sense.

However! When the server is bogged down with lots of AI/scripts/etc, these things will appear to be running at low framerates on the client. That is, the client is running fine, but AI are skipping/warping/etc.

[kremator 1064]

Aye TB .. you were talking out yer jacksie !

[Tankbuster 1736]

Looks that way. :)

[max power 21]

I guess in a sense it's true. While you're getting a decent video frame rate, if the server is slammed, the server side simulation is jacked... this means all the crazy things you see when your sp game is running at low frames: ai walking through walls, vehicles crashing and flipping, weird physics, etc.

[tvig0r0us 27]

I guess in a sense it's true. While you're getting a decent video frame rate, if the server is slammed, the server side simulation is jacked... this means all the crazy things you see when your sp game is running at low frames: ai walking through walls, vehicles crashing and flipping, weird physics, etc.

I think the most noteable thing is the super slow reaction time of the ai. I run my test server on a pretty old machine so it can't quite keep up the way it should most of the time and other than aircraft seeming to have inconsistent speeds and the occasional oddity (ai walking through walls as you mentioned), the most obvious sign that the server is overloaded is the oblivious ai. Sometimes it's so bad you can literally walk right up to them and look in their eyes before you leisurely pull the trigger and blow them away.

[Harzach 2515]

Is AI walking through walls really a framerate issue, though? I see it all of the time and my dedi almost never drops below 50fps.

[max power 21]

Is AI walking through walls really a framerate issue, though? I see it all of the time and my dedi almost never drops below 50fps.

Pathfinding, prediction, and other problems happen more often at low frame rates.

[Beagle 684]

It is a well know fact that low server fps basically cuts client fps in half...and lag menu. The main reason is remote A.I. exeding what the server can bear...happens around 200 active remote A.I. for quite powerfull servers. Warfare missions are plagued by this for years.

[Tankbuster 1736]

I think (read hope) Max has sort of saved my blushes. In an MP environment, where the clients are dependent on the server for 'seeing' much of the action, by which I mean AI activity, the client FPS is irrelevent. The AI are bound by the server FPS, not the client FPS.

In short, no matter how uber your FPS are, the server controlled AI will only operate at the server FPS. This doesn't mean that as I said earlier client FPS< server FPS, but it might give that impression where the client is seeing nothing but server entities.

[kremator 1064]

It's OK TB ... we still love you :) (in the most manly sense of the word!)

[Tankbuster 1736]

It's OK TB ... we still love you :) (in the most manly sense of the word!)

Bless you! :)

[Harzach 2515]

Maybe I'm not getting accurate returns from the #monitor command. Is there better/more accurate way to measure server framerate?

[cifordayzserver 119]

A smarter person than I told me "A client's VM can not run faster than the server, but they can have higher FPS"

[Tankbuster 1736]

Maybe I'm not getting accurate returns from the #monitor command. Is there better/more accurate way to measure server framerate?

What makes you think #monitor isn't reliable?

[pawelkpl 29]

FALSE !!! I DO HAVE DEDICATED SERVER RUNNING NOW OF 14FPS (heavily modified xeno domination), As client my fps is 40-70fps.

[Harzach 2515]

What makes you think #monitor isn't reliable?

As I mentioned, my server rarely drops below a reported 50fps, yet AI walking through walls/objects is not an uncommon sight. In fact, my server often exceeds the stated maximum framerate of 50fps, which further leads me to doubt the veracity of the results I am getting from #monitor.

Maybe it's something to do with my server config or the mission I'm running or ... who knows what. I'm not doing anything beyond the norm, though.

[Predator.v2 10]

Harzach, what kind of server are you using? We tried different root servers with A3 and even with our test root server with 6 core on 3,2ghz, 32 gb ram and 100 Mbit on a 20 max player cap, it drops down to like 3 fps on invade&annex.

[Harzach 2515]

i7 2600K @4Ghz, 16Gb RAM, 50/5Mbit. It's a second PC, re-purposed for server duty only. I'm launching via TA2DST.

Server load is never that great (most I've had on was nine players), and I pretty much only run Insurgency (though highly modified), so enemy AI is generally well-managed.

[pawelkpl 29]

As I mentioned, my server rarely drops below a reported 50fps, yet AI walking through walls/objects is not an uncommon sight. In fact, my server often exceeds the stated maximum framerate of 50fps, which further leads me to doubt the veracity of the results I am getting from #monitor.

Maybe it's something to do with my server config or the mission I'm running or ... who knows what. I'm not doing anything beyond the norm, though.

I do have experience with dedicated server - domination mission more than 3 years and what I've noticed AI behaviours better (human alike) and do not see some glitches - walking through doors, etc if server runs abnout 15 fps instead 50fps, maybe it is domination mission related but I doubt it...

[gossamersolid 155]

Pathfinding, prediction, and other problems happen more often at low frame rates.

If the objects are local to the server and the server has a low framerate, this is to be expected.

[fred41 42]

As I mentioned, my server rarely drops below a reported 50fps, yet AI walking through walls/objects is not an uncommon sight. In fact, my server often exceeds the stated maximum framerate of 50fps, which further leads me to doubt the veracity of the results I am getting from #monitor.

Maybe it's something to do with my server config or the mission I'm running or ... who knows what. I'm not doing anything beyond the norm, though.

... it is of course very mission dependent and depends of the number of players.

Scheduled code execution, like spawned script code, code executed by execVM, condition evaluation in (AI) FSMs has nearly no impact to the server FPS, while non scheduled code execution like eventhandler processing, actions in AI FSM e.t.c. has high server FPS impact.

A high number of players will normally cause frequently triggered eventhandlers, AI actions e.t.c., so it impacts server FPS a lot.

That AI is walking through walls, indicates that the scheduler is heavy (over)loaded, which causes the delayed condition evaluatiions in AI FSMs, even if you see high server FPS values.

See here for details about scheduled/non scheduled code execution: http://community.bistudio.com/wiki/Code_Optimisation

I think, it is time for an extended version of the #monitor/diag_fps() command, to allow server admins and developers to monitor the current scheduler load too!!!

But to the thread topic question: Client FPS schould normally be independent from server FPS (except lags caused by synchronisation delays, still a problem in the current state of the game)

Edited May 17, 2013 by Fred41