Why do we have such a low FPS in multiplayer(MP performance vs SP performance)

[DancZer 65]

I haven't found an active discussion about the MP performance vs SP performance so here it is. The Low CPU utilization & Low FPS topic is too general for this issue. It has to be discussed separately.

Definitely there is something wrong with the MP. I think the issue is serious and affect lot of players. Even those who have very good peace of rig. Try to stick to the topic and not OFF it with CPU/GPU utilization in SP mission. Keep your SP performance as a reference of your MAXIMUM.

We just wan't the same or almost the same FPS as we have in SP.

My rig: Lenovo y510p with i7 4700MQ, 16GB@1600 DDR3, NVIDIA GT 755M and Intel SSD 520 Series 240GB,

Internet connection: 15Mbit down, 1Mbit up

I have 40-60 FPS in SP games and 16-25 in MP games. Sometimes even low!!

Why do i have such a huge difference? How about you? What is your rig+internet connection?

Mostly i play KotH on high populated servers, but if there are ~10 players my FPS isn't better.

Does the server FPS affect the client FPS? Why?

[R3vo 2654]

Mostly i play KotH on high populated servers, but if there are ~10 players my FPS isn't better.

Does the server FPS affect the client FPS? Why?

Yes, the server affects the client FPS, actually it does alot. The main reason for that are bad scripted/built missions. Many missions I've seen are using way to many AI or alot of unessesary script or unefficient written scripts. In addition, some servers haven't got proper hardware, or aren't running an Arma 3 Server alone, but also game servers of other games at the same time.

[HorbeySpector 164]

Does the server FPS affect the client FPS? Why?

Yes, and it sucks. Heritance of the amazing RV engine.

Now there will be a lot of people who come in here and talk about badly optimized missions, as if you will have 60 fps on a BIS mission..

nope.

[DancZer 65]

Now there will be a lot of people who come in here and talk about badly optimized missions, as if you will have 60 fps on a BIS mission..

nope.

If this is true, then BI should make a guidance, how to make a proper MP mission and what's are the bottlenecks.

[St. Jimmy 272]

Would be interesting to see some server benchmarks how some Intel CPU running 4GHz combined with good memories and connection runs same mission with same amount of people compared some CPU running at 3GHz. Just to learn how much a fast CPU server matters.

[blackpixxel 53]

I always hear that there is only bad MP fps because everyone can host missions on his bad PC. But why isn't it better when I play servers with high performance? And why should I get low FPS at all??

In theory, my PC only has to receive the positions of other players and AI and place them, and send the server the position of my own player. How should this consume more performance than when I play a SP mission with ton of AI??

My PC should not care if there are many AI in a mission, because my PC does not have to calculate anything AI related. This is the job of the server. A client's PC just places the units then, nothing else. The only thing that a client PC would have to do is interpolation of the position data, since you don't receive new packages from the server every frame. And to avoid stuttering AI and players the client has to interpolate their movement, which should not consume as much performance as calculating AI.

Something is definetly wrong with the whole netcode/multiplayer code. But BIS don't care. Since 13 years.

[fareast 20]

yea mp fps for me slightly lower than sp but sp performance also crap..i tested it on dual core with HT the cpu usage still pathetic..gpu usage also sometimes goes below 10%..

[havatan19 10]

I have 40-60 FPS in SP games

40-60 FPS in SP on a empty terrain (without AI) or with many AI and huge firefight ?

if its on a empty terrain, then i understand that the FPS is high, but if you have AI and firefight, you will notice that FPS in SP is shit aswell so you wouldnt wish that the FPS in MP should be the same as in SP.

[DancZer 65]

There is no AI in PVP games that i play and i still have low FPS.

60 FPS in the editor and 40 FPS during the campaign.

[Sterlingarcherz101 15]

I find at the moment dysnc is worse on the new patch. I can be getting 40 fps multiplayer but then horrible dysnc occurs.

[TSAndrey 1]

I always hear that there is only bad MP fps because everyone can host missions on his bad PC. But why isn't it better when I play servers with high performance? And why should I get low FPS at all??

In theory, my PC only has to receive the positions of other players and AI and place them, and send the server the position of my own player. How should this consume more performance than when I play a SP mission with ton of AI??

My PC should not care if there are many AI in a mission, because my PC does not have to calculate anything AI related. This is the job of the server. A client's PC just places the units then, nothing else. The only thing that a client PC would have to do is interpolation of the position data, since you don't receive new packages from the server every frame. And to avoid stuttering AI and players the client has to interpolate their movement, which should not consume as much performance as calculating AI.

Something is definetly wrong with the whole netcode/multiplayer code. But BIS don't care. Since 13 years.

Exactly. Why should the server impact my FPS? The only thing that should impact the FPS are large amount of objects/effects in the mission, but that's obviously not the issue hee.

[DancZer 65]

Why should the server impact my FPS?

I think that some things inside the engine are forced into turn base logic. Or the client wait the server response in synchronized way.

[Greenfist 1863]

I've been reading these forums a lot over the years but I don't think I've ever seen a definitive explanation for this FPS dependency of server/mission and client.

Would be interesting to see some server benchmarks how some Intel CPU running 4GHz combined with good memories and connection runs same mission with same amount of people compared some CPU running at 3GHz. Just to learn how much a fast CPU server matters.

Someone should try this with the client. I've only heard very contradicting opinions about it; does a better cpu really matter on a bad server?

I mean, if you have a powerful rig (~4.5GHz CPU, fast ram etc.), just join some low performance server and check the fps. Then downclock to 3GHz and join again.

I would do it myself but cpu is already badly low-end. :(

Edited November 2, 2014 by Greenfist

[windies 11]

I've been reading these forums a lot over the years but I don't think I've ever seen a definitive explanation for this FPS dependency of server/mission and client.

Someone should try this with the client. I've only heard very contradicting opinions about it; does a better cpu really matter on a bad server?

I mean, if you have a powerful rig (~4.5GHz CPU, fast ram etc.), just join some low performance server and check the fps. Then downclock to 3GHz and join again.

I would do it myself but cpu is already badly low-end. :(

I pushed my CPU up to 4.9ghz once since I have watercooling, I don't like running it that high since This MSI board doesn't have offset voltages, and MP performance literally stayed the same as the normal 4.2/4.4 I run, tested on the same server. SP performance did go up, but again it was very very modest gains for the stress on the hardware like 4-5 FPS or so. Honestly when I switched from a Phenom II 965 to an i5-3570K, in multiplayer there was almost no difference in performance, I think maybe like 3-5 fps in multiplayer. I know some people will say "WOAW 3-5 fps that's massive for ArmA" but for my eye's it really feels like nothing. I can't play ArmA when I'm getting like 15-20 FPS. What's really annoying is when you have a pretty good rig that can run things really well, like I can run some games in 4K and still get 40-50 fps average.

Is PhysX sync'd across clients? I could imagine that would be a huge reason why I see such a large difference between ArmA 2 and ArmA 3 MP performance vs SP.

[Gera_CCT 12]

I was getting 15 fps with a old system with a q6600. Then went to a 4790k and now i got minimum of 30 fps in MP like King of The Hill with 70+ players with same settings than before.

[toysoldier 35]

The problem is arma don't use the full resource of your computer and its only 32 bit. FOR A PC ONLY GAME IS THAT A SHAME!

[brightcandle 114]

I asked the same question in Alpha and Beta. We were asked to provide a profile trace to show the issue, we did so and then BIS staff never looked at it or did anything with it. Its been the problem with the game since the beginning of Alpha, while at times things improved a little its been messed up for most of the time and never acceptable levels of performance. BIS have been given the full details of the traces in multiplayer showing the issue and have had them for over a year. Its inexcusable at this point for performance to have continuously got worse over time, considering it has been the number 1 issue since the beginning of alpha.

[Dwarden 1125]

server FPS is irrelevant to your client FPS unless it's too low (under 5fps) ...

when that happens it start to negatively affect script scheduling, AI processing, predictions, updates etc.

server FPS maximum at 50 is enough to process all the server needs to

anything more would be waste of resources

(e.g. prediction which is discarded due to newly received update or new prediction in next frame)

MP optimizations are also ongoing work in progress

---------- Post added at 23:32 ---------- Previous post was at 23:29 ----------

The problem is arma don't use the full resource of your computer and its only 32 bit. FOR A PC ONLY GAME IS THAT A SHAME!

64bit isn't going solve FPS problem (even if some specific 64bit operations're used it will be in fraction of %)

[Necramonium 10]

When you play Altis life you pretty much get used to it, get over 60 fps when the server has done a fresh restart, a hour later you wont be even going passed 20 fps.

[atomictim7 18]

I had terrible FPS on servers, but I saw somewhere that using a different memory allocater helps increase FPS...and it does...everyone I've shown it to, has said they have had a massive boost...and they are streamers/youtubers who have lower FPS anyway from recording/streaming.

Follow the instructions here if you want to try it for yourself:

https://github.com/fred41/tbbmalloc_arma

[Tajin 349]

This whole issue is really hard to judge without meticulous testing.

Dwarden already said it, MP does need some more optimization.

However, that is only the smaller part of the problem. From my experience, mission-optimization plays a much bigger role in this.

High amounts of different assets can be a problem, but especially poorly optimized scripts (of which there are plenty) can put a lot of strain on both the clients and the server.

The thing is, you can hardly compare the performance of one server/mission with another without knowing exactly what is going on behind the scenes.

I've recently played a TvT round with roughly 50 people and had no FPS problems at all. That mission however was without any vehicles and used only a few simple scripts.

[Dwarden 1125]

I (and you) can setup server for up to 50 players to be able keep ~50fps (with drops to 20-25) nearly nonstop

I tested this on multiple PVP, PVPVE and COOP missions on my CHIMERA servers

of course in moment I there is too much AI or script issues then I can't vouch for that

anyway I'm not gunna deny there might be some hidden or new bug

that's why good reports and reproduction cases/ missions helps a lot

now,for player numbers over 50 the server FPS drop is larger with each new player

Edited November 5, 2014 by Dwarden

[jandrews 116]

I was getting 15 fps with a old system with a q6600. Then went to a 4790k and now i got minimum of 30 fps in MP like King of The Hill with 70+ players with same settings than before.

I have had similar results. I went from AMD overclock'd quadcore to i7 4950, 8GB RAM to 32 GB RAM and yes better GPU x2. went from 20's to high 50's low 60's for COOP and MP (PvP). Some dsync is server side, BUT overall I noticed the MOST change when I upgraded my rig. BTW, new rig settings are set to ultra high.

[TSAndrey 1]

I (and you) can setup server for up to 50 players to be able keep ~50fps (with drops to 20-25) nearly nonstop

I tested this on multiple PVP, PVPVE and COOP missions on my CHIMERA servers

of course in moment I there is too much AI or script issues then I can't vouch for that

anyway I'm not gunna deny there might be some hidden or new bug

that's why good reports and reproduction cases/ missions helps a lot

now,for player numbers over 50 the server FPS drop is larger with each new player

So Arma isn't optimized for >50 player counts?

[Dwarden 1125]

So Arma isn't optimized for >50 player counts?

I was more of stating that the more players the faster the FPS happens