Jump to content
stupor-mundi

Trying to understand multiplayer low fps and microstutters

Recommended Posts

I use IL-2 on a laptop with a 1070 eGpu, on a 1080p non-gaming monitor (60hz). Vsync never worked well for me, I get the (relatively) best results by setting graphics to 80 fps, combined with dynamic resolution factor (currently 0.7)

 

Always I had problems with fps and spotting in multiplayer. After the recent update, and MS forcing the 1903 update on me, and performance further degrading, I installed the latest Nvidia driver, moved startup.cfg to another folder, and restarted IL-2, attaining a new startup.cfg, presumed default.

 

I'm trying to understand properly what exactly happens in multiplayer that causes the performance to be so bad. On ultra settings, in single player, I can fly against 8 AIs, have them all around me in immediate vicinity, and get around 60 fps. In multiplayer, I see a strong degradation on the "big" servers, with big maps and many players. My spotting is terrible.

 

So when I fly in multiplayer I usually go on "EU official", which has icons and smaller numbers of players. Even there the fps is much lower (than single player) with additional microstutters, but not as bad as on the 'big' servers.

 

I find it kind of hard to grasp how and why multiplayer even impacts my GPU (i.e. fps and stutter). If it is the multiplayer data transfer ("netcode"), which causes a bottleneck, what is the nature of this bottleneck? I can only imagine it would have to be the cpu. I run on a mobile cpu (4 core, 8 threads), hence weak-ish.

 

Yet when I fly on multiplayer with the resource monitor running, (and I get into a low fps situation), none of the cores shows above 70% load. If none of the threads is even fully loaded, how can the networking impact the FPS negatively?? Clearly it does though.

 

The IL-2 multiplayer settings GUI used to have a field for upload and download limits, I believe. It's disappeared at some point?

 

My new startup.cfg has got these, default, settings:
    client_download_traflimit = 1024
    client_upload_traflimit = 1024

 

My old startup.cfg had:
    client_download_traflimit = 4096

 

Which I had set, based on some info on one of the threads, that you should do a speed test and divide that by roughly four, to determine the setting (or something like that).
    
Is it still advised to edit those settings, even though they've been removed from the GUI?    
Does that old rule still hold?   

Edited by stupor-mundi

Share this post


Link to post
Share on other sites

I see a number of views on this, but no answers. Is it an exotic issue that others don't have? Is the topic in the wrong section, should I try to get it moved?

I'll try to give some background to why I bring up this question:

 

I'm considering:

- buying a gaming monitor to get refresh > 60 hz (in this case I would go > 1080p)

- building a new gaming PC

 

Yet, I'd be quite content with the fps I get in offline/single player. My 1070 eGpu seems to be sufficient to run Il-2 GB graphics-wise, if it wasn't for the multiplayer / network induced crapness.

 

Equally, if it were clear that my laptop's cpu is too weak, that would resolve the question in favour of building a gaming PC. But in this case I would expect to see at least 1 thread maxed out in a low fps situation, which I don't.

 

I'm not really looking forward to the effort and expense of that, only to maybe find that the bottleneck lay elsewhere. To illustrate, here is a screenshot of resource monitor cpu cores/threads. I took it about 10, 15 seconds after being a 10fps situation on the tank server. None of the threads look maxed out.

I'm actually quite surprised at the even load across all 8 threads, I had expected there would be one thread bottlenecking.

 

Capture.thumb.PNG.bf1b2d4920cdf92a009a8027d2d5ae6e.PNG

Share this post


Link to post
Share on other sites
9 minutes ago, stupor-mundi said:

I see a number of views on this, but no answers. Is it an exotic issue that others don't have?

 

The MP stuttering seems to be something that not everyone is experiencing, or at least not severely enough to complain about it. I don't really have much to add to that discussion.

 

12 minutes ago, stupor-mundi said:

To illustrate, here is a screenshot of resource monitor cpu cores/threads. I took it about 10, 15 seconds after being a 10fps situation on the tank server. None of the threads look maxed out.

 

The usage shown by Resource Monitor is misleading, as it is an average for each core over a certain time. In reality, Windows is constantly moving the load around, and the game can be limited by the performance of a single thread at any instant despite no cores showing an average of 100%. Similar behaviour can be seen with other software, such as PassMark single thread benchmarks.

 

It's also a good idea to monitor GPU usage to make sure that you are not hitting limits there.

 

15 minutes ago, stupor-mundi said:

- buying a gaming monitor to get refresh > 60 hz (in this case I would go > 1080p)

- building a new gaming PC

 

What are the specs of your current laptop? If it has a typical low-power CPU, it is quite likely that you are severely limited by CPU performance. In that case, a new gaming PC would probably be the best option.

 

Higher refresh rates can mitigate certain types of stutters, but they won't help much if performance is abysmal. 

Share this post


Link to post
Share on other sites
1 hour ago, Mitthrawnuruodo said:

The MP stuttering seems to be something that not everyone is experiencing, or at least not severely enough to complain about it.

 

Now, this is really astonishing to read! We have several topics complaining stuttering in MP. There are some sever issues in MP performance-wise and

it is observed not only by users with low-mid range PCs, but also by people with new powerful rigs.

 

And btw. Mr. Jason Williams started a new topic about this very same subject to gather more information and data to finally analyse the issues in MP.

 

The issues with stuttering in MP is also very popular on the Russian forum, except that some of the dev team there are in constant exchange in some

way with the users, which I miss a little bit here.

Share this post


Link to post
Share on other sites
15 minutes ago, -=-THERION said:

...except that some of the dev team there are in constant exchange in some way with the users, which I miss a little bit here.

Well we got Jason...it's better than nothing I suppose.

Myself and a few friends solved most of the issue by forcing the game to run with only 60 fps (which sucks when you have a 144Hz monitor like me).

Share this post


Link to post
Share on other sites
19 minutes ago, -=-THERION said:

Now, this is really astonishing to read!

 

What exactly are you disagreeing with? I'm not saying that there are no stutters.

 

I'm well aware of Jason's thread about MP stutters. It currently has only 15 replies, so it looks like a complex issue that not everybody is able to comment on. That seems like a reasonable explanation for the lack of responses here.

Share this post


Link to post
Share on other sites
53 minutes ago, =FEW=Hauggy said:

Well we got Jason...it's better than nothing I suppose.

Myself and a few friends solved most of the issue by forcing the game to run with only 60 fps (which sucks when you have a 144Hz monitor like me).

 

Yes, I have been fiddling around this weekend and I run my game with 30(!) fps, which isn't gorgeous, but I got rid of stutters (mostly)...

Share this post


Link to post
Share on other sites
On 11/3/2019 at 9:37 PM, Mitthrawnuruodo said:

The usage shown by Resource Monitor is misleading, as it is an average for each core over a certain time. In reality, Windows is constantly moving the load around, and the game can be limited by the performance of a single thread at any instant despite no cores showing an average of 100%.

 

Hah! I should probably check again, using some more accurate monitoring software. However, assuming this is correct, everything falls into place.

 

On 11/3/2019 at 9:37 PM, Mitthrawnuruodo said:

What are the specs of your current laptop?

 

It's an ultrabook from end of 2017 IIRC. low wattage core i7. Certainly qualifies as a quite weak cpu in 2019.

Thus, assuming the cpu is the bottleneck in multiplayer, i.e. not being able to keep up processing the incoming data from the server, makes sense.

 

In the big thread about multiplayer stutter, Bauemer had suggested to do a fresh install to remove stutters. Since I had done a fresh nvidia driver install, and had gotten a fresh startup.cfg , I doubted doing this would have a big effect in my case.

 

But I did it anyway--Revo uninstaller, wiping entries in the registry, everything... with very interesting results:

 

Even though, in multiplayer, in busy areas where my rig bottlenecks in the client/server comms, my FPS were just as low as before (for instance 15fps in bad cases), they were suddenly devoid of the additional, irritating microstutter. Quite amazing. I wonder what it is that a fresh install does, or that the update process screws up over time... So, low fps don't look nearly as awful as what i had gotten used to, when the stutters are gone.

 

I'd make it part of my routine, after one or 2 updates, if it wasn't for the extreme clunky inconvenience of the fresh install process.

 

Also interesting is that there is the OTHER kind of stutter, which, now that one kind of stutter has disappeared, I can identify, and I believe it's not a multplayer stutter:

 

Even flying offline, on the rhineland map, when flying towards a town, there's that moment of stutters when the buildings are loaded, which the other posters on this thread have mentioned.

 

 

 

Edited by stupor-mundi

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×
×
  • Create New...