Hey all, I've been wondering how relative server performance scales with the number of users connected. Let's assume that all Client PC's are perfectly capable at running a Single Player game at a constant 60fps at the users selected fidelity settings. Let's also assume that the server is similarly capable. The reason I ask is because I've been watching numerous excellent MP videos from the like of @spanj as he plays on @Thee Wolf 's servers - not mentioning you guys expecting a response, just wanted to give a quick nod to your videos . However, one thing I've noticed are the quite dire FPS numbers at times, making gameplay tricky. I personally, experience something a bit like motion sickness followed by eye strain if I try to play any game where the FPS drops so badly...seems my eyes are particularly sensitive to low fps - even the cinema isn't always so pleasant for me, especially if there are lots of panning shots. As a more concrete example of what I'm talking about, I have two mostly older PC's. One is my main gamer and is used for my Single Player games. On this I obtain a steady 60fps most of the time, despite the PC also being the host of course. My second PC is a similar spec, but with less ram and weaker GPU's. However, this is the machine I used to host my local Multi Player servers. Now in this Client / Server configuration neither PC was taxed at all CPU-wise, while the Client's GPU was often pushed, due to my use of generally higher settings. PC Specs: Client 2600k @ 4.4ghz GTX 1070 32gb DDR3 1600 RAID0 SSD's (OS & Steam) Windows 10 Pro Server 2500k @ 4.6ghz 2x GTX 680 8gb DDR3 1600 RAID0 SSD (OS & Steam) Windows 10 Pro So, my question: While my Server PC doesn't even break a sweat when hosting one Player, how do things scale with multiple players? I guess players per playfield and the amount of currently active playfields are a major factor, but I'd like to understand how this impact server Ram usage, CPU Load as well as network traffic. My final potential issue is bandwidth. Running on my LAN it's not a problem, but if I did invite some external players to join then how will my meagre 20 m/bit down and 1.2 m/bit up fair? Note: I've observed how weaker client systems can get a major FPS boost where they were perhaps CPU limited when running a Single Player game. Cheers, Scoob.
Hello Scoob! I can only tell you my personal experience! I think your server hardware will definitely not be the problem. What I've noticed in the past is that the ping of yourself and those of your teammates, as well as the complexity of the designs, is crucial to the MP. In your case, I also see the upload of 1.2 as a bigger problem to let more than 1 external join! As an an example: My internet has 10.5 MBit upload with a usable and stable 25ms ping! I can easily host two external people but if we spawn several larger structures (or larger CVs, SVs or HVs) then the FPS already likes to go <30. (And the worst are CV attacks with many turrets <15FPS ) My server (old AMD 8-core, 12 GB DDR3, SSD) on the other hand is bored with 2-3GB of used RAM and 10-20% CPU load! I hope I could help you with my assessment!
Hey @Scoob - glad you're enjoying the video's I cannot say what the spec of TheeWolfs server is, I don't know but the way it is setup currently is not good when it comes to limiting lag and dropped frames. The playfield we are fighting on at the moment has been modified in a not so graceful way to make a unique MP experience on the server. The aim as you may already know, is to enable the 2 PVP factions to have an all out war without the stress of endless grinding only to lose a load of stuff in a single engagement. The PVE faction will sell us their excess resources for gold which is the only mineable resource on the PVP planet. The PVE faction use that gold to buy their own planets and for general bragging rights etc. The problem is the gold deposits are so very enormous they lag out the entire area. Before we even put a base or any other stuff in the area, frames were down to about 25fps. Now we have a base with defences a ton of SV/HVs parked in it, the frames are around 10 -15. Add to that the number of rocks on the planet was drastically increased so we could get some basic resources without having to trade. And furthermore, Typically we build our bases underground to make them harder to attack but the process of digging out that much ground also has an impact on frames and general lag. So i guess what i am saying is, if you are planning to modify your server from vanilla, be careful; the simplest of changes can have a big impact on FPS. The Wolves server will be changed come A8 to eliminate a lot of the problems we are seeing now. The deposits will change to much smaller and better implemented to reduce lag, the number of rocks will go back down to vanilla numbers i think. Not much we can do or want to do about underground bases though, kinda needed in PVP. I do think Eleon are doing some optimisation with A8 which will hopefully improve some things but no doubt the amount of stuff they are adding will also introduce other performance hits.
Thanks for the responses @MEKNET1977 and @spanj I guess with 8.0 around the corner my question is likely more difficult to answer as we're not yet sure what potential problems, or indeed optimisations, this new version might give us. I recall the devs saying the larger playfield shouldn't impact performance on their own, but lots of stuff is changing. It does seem that my 1.2 m/bit upload speed is perhaps where I'll suffer the most. When sold the package I was told I'd get a solid 10 m/bit up, but they basically miss-sold me, and the local infrastructure simply cannot handle that sort of speed. They don't know why but it can't, so I'm stuck. I wasn't considering being a serious Server Host to the same level as Thee Wolf, more a semi-casual thing so a couple of friends could potentially join in. I guess it's more about "having a play" than anything lol. Also, as I do have a fairly decent spare system in the 2500k I thought it'd be a good use for it MEKNET1977, if the server isn't being taxed even during quite demanding encounters that are seeing the FPS drop hugely, is it then the client machines that are being pushed, or that the data traffic between the various players and Server cannot cope? Spanj, I did actually catch one of your more recent videos after posting this, mentioning that the server wasn't so great & will be tweaked when 8.0 drops. Are you saying that the FPS drops I see in your more recent videos aren't considered normal in MP? You mention that currently it's perhaps not set up ideally to limiting lag, are there any pro tips in this regard? In the past - with just me playing - I've set up mildly modified servers but I tend to stick to the default device limits, which likely made things easier. I guess I simply never pushed my own hardware hard enough to see any issues with just me connected. I think I'm best off just trying it when 8.0 is on the main branch - I'll do my Experimental testing in SP. Brain is fried this evening...not sure I entirely made sense lol. Scoob.
Hi @Scoob ! Good question! I am an IT expert (Domain, Exchange and Citrix) but unfortunately I can not explain exactly how the game handles this. These are just my personal observations paired with some expertise! But I suspect it is not the hardware performance of the server and the clients, but basically the quality and bandwidth of the Internet lines (of all participants). Originally I also thought I could host the server for 2-4 people but it is still no comparison with the larger professional vendors! I then decided (should not be an advertisement! ) to go to Nitrado even if I have to do without the EAH (which is very helpful in the MP). But it works well for smaller groups and is cheaper and more performant than doing it at home. Just for illustration: I have 4 player slots for 4 euros per month! The server is running 24/7 and I do not have to pay any electricity for it! In addition, the data centers have much better Internet connections which are not comparable with our private Internet connections! But you have short bearable framedrops when, for example, in the field of flying where several structures or vessels are. @spanj i like your LPs, at the moment i'm watching your Season 3! Although I am a pure PVE / COOP player I find your PVP game very entertaining! Please excuse my terrible English today
Very hard question Scoob. Not jusyjthe amount of playfields matter but what those playfields contain. I have seen playfields using 100mb ram and some 2gb ram
Definitely not normal - No we suffer worse FPS than normal i think. i'm not running the server so I dont really have any pro-tips more just observations on the current and past setups. This is the worst FPS lag we have had on the Wolves server and also the setup with the fewest playfields.
Yes, I can appreciate that. One persons game might be very different to another's based on what players have built, POI distribution etc. I've only tested mostly vanilla games - a few tweaks to Drone attacks and critters - and my Server never broke a sweat with just me connection, despite some fairly sizeable builds. Think large BA, multiple vessels docked. Still, as a lone player I can never really push things like a group of players could. Glad to hear this isn't a normal for a Multi Player server! While I have played around with my own server of course, I've given no thought to performance at all as, with just me, the server was never pushed. I think I will certainly play some more once 8.0 is available. Scoob.
Hey hey @Scoob I have been developing Empyrion playfields and servers for years now and its been a long hard road to test and keep the FPS up while having great Playfields to enjoy, let me tell you lol! But Out of everything I can tell you this the up rate you have will not support more than maybe 2 to 3 others with you, before you have a server choking, breakdown lol. I say this because we have tried on servers using 1.5 up rates and it just doesn't work. Sorry to say! As for playfields its all based on how much you have in and on them. The more you use asset wise the more the server draws on it. Now on that note its all also based on the assets you add and there draw. The ore deposits are evil, they take more stats than most my ship builds, go figure lol, don't know why but they do from testing. Then I can have trees cover every inch of a playfield and nothing is effected. Right know our server is in testing phase, as we call it, finding the bugs and issues, fixing what we can. Also the reason I have no new episodes for you guys yet, we making the server to record next season on. I don't have programs to help me make these playfields I do them all by hand 1 asset at a time till its right and working the way we want, I see it in my dreams now lol ;p All in all its trial and error when building playfields. Build it test it, Adjust where needed, and test again, then rinse and repeat. I hope this helps you get the idea of how things are affected when creating your server and playfields. Some Empyrion assets eats CPU and RAM like my kids eat candy on holidays, lol ;p While others have little or none. If you have more in-depth questions join or Teamspeak and will gladly have a chat, TS: wolves.lerdin.com TheeWolf WolvesCommunityServer Founder
Hi @Thee Wolf Thanks for the information. Today I'm barely getting 1.1mbps up stream so I suspect hosting a server here, even if just for a couple of friends, isn't going to work out so great. Ah well! I appreciate how different aspect of the game can tax different areas. I'd never have considered Ore Deposits to be something excessively taxing, though @spanj did touch on this in his reply. Also, you mention trees are not a resource hog, I'd have thought the opposite! I think, for me, the plan as much as there is one is to get my hands on 8.0 first. Perhaps once that's on the main branch I will test hosting a server. I might just take you up on the Teamspeak chat offer once I'm tinkering...you might live to regret it lol. Scoob.
Hey hey @Scoob That's what our community is all about helping each other to get the most out of our gaming so its no problem
