16 replies to this topic

[FORCELLO]

Hey all,

I get 120 fps when dropping in, but as soon as the battles start my fps tanks. I know MWO uses an older cryengine and isn't very optimized, but I find it hard to believe that my fps tanks to 40 fps when I'm fighting someone.

I have tried user cfg's from these forums and have set everything to low except for textures.

I have an I7 7700K overclocked to 4.9 GHz and a 1080 Ti.

How do streamers like baradul run this game smoothly?

I don't know what to do at this point and would like some help since I have absolutely no performance issues in any other game.

[UnofficialOperator]

When all else fails... more RAM!!!

But seriously I went from 8gb to 16gb and my stuttering issues seem to have gone away from my 580 8gb. You have a way better cpu and card too

[Kotzi]

1. Up to date drivers?
2. Run the cpu non overclocked.
3. Wlan? If so try Lan to see if there are any problems.
4. Buy a mechpack.

Edited by Kotzi, 28 May 2019 - 04:01 AM.

[O L L O]

Connection perhaps? Are these matches missile heavy?

Edit:
How is your CPU holding its clocks? Check if it is throttling.

I always run HWMONITOR in the background and check my lowest clocks after each gaming session to see if there was any throttling.

https://www.cpuid.com/

Edited by Ollollo, 28 May 2019 - 04:28 AM.

[justcallme A S H]

Don't run everything on Ultra High. 3/4 the settings don't really make "that" much of a difference but totally slam performance.

This is how I have mine setup --> HERE <--

I average 100-130FPS most maps @ 1440p. Solaris / River City - around 80FPS. MWO is entirely CPU bound.

[GoodTry]

justcallme A S H, on 28 May 2019 - 04:36 AM, said:

MWO is entirely CPU bound.

That's not true. Even at minimum settings, I saw a decent framerate jump at minimum settings going from a Geforce 980 Ti to a 2080 Ti, and keeping the same CPU.

I agree that it's largely CPU bound, though.

Edit: I'll also say that, even with a high-end Intel CPU and a Geforce 2080 Ti, there are still framerate hiccups during the day-night transition on River City...

Edited by GoodTry, 28 May 2019 - 04:59 AM.

[justcallme A S H]

I went from a GTX1070 to a RTX2080 and saw zero FPS increase with a 8700k @ 4.9ghz at 1440p.

So CPU, GPU - It's CPU in most instances.

Edited by justcallme A S H, 28 May 2019 - 04:59 AM.

[dario03]

At this point I think the game just randomly chooses if it is going to run good or not
I have played with all settings turned all the way down and still get times where the fps drops badly but more annoyingly stutter and pauses because of frametime spikes of 100+ms. And that's with no cpu core, the gpu, the ram, or the vram being any where near maxed out, usually around ~40% usage with low or mostly low settings.

However its never consistent, like last time I played it was mostly fine, some spikes here and there but mostly fine. Time before that though it was all over the place with fps drops and frametimes that don't even make sense. I made no changes, it just did. I had thought maybe it had something to do with communication with the game server but that's not usually something that would cause these kind of issues, and my ping is usually fine.

So I don't know, I figure it must just be a combination of things. Maybe the connection does effect the games performance and even though ping is fine some of the data is being transferred weird through my ISP. Maybe the game doesn't like something about my hardware despite it being better than recommended and works fine for anything else including other CryEngine games.

Edited by dario03, 28 May 2019 - 05:23 AM.

[MrVaad]

The correct answer is because, depending on the map, the Mechs you are facing, and the weapons used, you can be :
- CPU limited
- GPU limited
- fillrate limited
- draw call limited
Fillrate and drawcalls are tricky because you can't see them with classic monitoring software.

I'm trying to build a list that we could give to PGI.

1. There is some very specific particle effects that are not done right. I have identified at least the launch smoke of missiles and some bubbling puddles in caustic valley This can be checked in the training grounds (and in the cryengine editor).
I modified those effects with the cryengine editor to be sure (and no, i did not used this in match ).

2. There's a few places in some maps with veryyy small grass, you could think it's a flat texture but it's 3D grass. I can make my 1080 go to 65% in the training grounds in those places (a few spots at least in solaris).

3. Some solaris ads could have problems. And perhaps too many of those ads can be destroyed (ie physics enabled on a great distance)

4. There's one cryengine variable that doesn't have the correct value regarding occlusion (ie "testing if something is hidden"). This variable is locked.
The way it is set, only big objects are tested for shown/hidden. I've never seen a cryengine game use this value.

Edited by MrVaad, 28 May 2019 - 09:53 AM.

[dario03]

MrVaad, on 28 May 2019 - 08:59 AM, said:

The correct answer is because, depending on the map, the Mechs you are facing, and the weapons used, you can be :
- CPU limited
- GPU limited
- fillrate limited
- draw call limited
Fillrate and drawcalls are tricky because you can't see them with classic monitoring software.

I'm trying to build a list that we could give to PGI.

1. There is some very specific particle effects that are not done right. I have identified at least the launch smoke of missiles and some bubbling puddles in caustic valley This can be checked in the training grounds (and in the cryengine editor).
I modified those effects with the cryengine editor to be sure (and no, i did not used this in match ).

2. There's a few places in some maps with veryyy small grass, you could think it's a flat texture but it's 3D grass. I can make my 1080 go to 65% in the training grounds in those places (a few spots at least in solaris).

3. Some solaris ads could have problems. And perhaps too many of those ads can be destroyed (ie physics enabled on a great distance)

4. There's one cryengine variable that doesn't have the correct value (but is locked ) regarding occlusion (ie "testing if something is hidden")

I was wondering about things that don't show up in monitoring like that. Though in my case I don't think its gpu since I can turn up graphics settings a decent bit and not get more issues. However I have wondered about something like cpu cache. I had a program way back in the day that was very limited by that. The program was single thread only but when I upgraded from a dual core to a quad core the performance jumped up a ton, like 100%. Thing is the quad core single thread performance in most things should have been about the same so I asked about it on their forums. Figured it was the cache since the cpus were about the same clock speed and architecture but the quad had much more cache.
Also thought maybe ram speed but when I messed around with that in this game I didn't see much and I don't think its usual for my cpu generation to.

And yeah heavy LRM usage on the field seems to often times drop performance.

[MrVaad]

Yes, CPU cache years ago made a big difference between cpus, but i don't know if it's still the case today

From what i've seen in the editor, 4 effects are spawned for EACH launched missiles :
- smoke launch
- smoke prime (smoke while flying ?)
- smoke remaining (smoke left on the missile path)
- fire on the rear of the missile itself

3 of those are OK, but the missile launch smoke is a BIG puf of smoke that stays way too long.
So you end up with multiple smokes layering on top of each others (because you're not launching a solo missile ).
This tank your system and there's not real way around that. You need to fix your particle effect.

I can be wrong, but i think, most video games never use smokes on top of each other.
Using only one/two big smokes and moving them chaotically can give the illusion.

Edited by MrVaad, 28 May 2019 - 10:46 AM.

[O L L O]

justcallme A S H, on 28 May 2019 - 04:36 AM, said:

Do you have any changes to the startup ini-file (whatever its name is) to recommend?

[Eatit]

In this video the streamer is on the new FP PTS and the Dev is making changes to the game physics. You can hear people complaining that when the changes happen the FPS drops dramatically. 1:06:45 is where you can see this in the video.

I imagine that if the server is under high load you will experience FPS drops. This seems to be the case here.

Video >Here<

[MrVaad]

I imagine the PTS is also running with more logs, more telemetry, and tools to edit things live.

[justcallme A S H]

Ollollo, on 29 May 2019 - 03:24 AM, said:

Do you have any changes to the startup ini-file (whatever its name is) to recommend?

I don't use anything in that department. Just the in-game stuff!

[Tarl Cabot]

Quote

but I find it hard to believe that my fps tanks to 40 fps when I'm fighting someone.

Define the above, ie 50m or closer spamming SRMs, RAC, etc?

Control Panel, Power Options set to High Performance (always double check after a major WIndows 10 update)
Nvidia control Panel, Power Management Mode,, change from adaptive to Max performance.
And last time system heatsinks were clear of dust buildup?

[0Jiggs0]

Tarl posted a tip in similar thread that worked great for my laptop. Might be worth a try.
https://mwomercs.com...wer-fps-online/