FPS issues since 1.20 (Kart DLC Patch)

[Nashable 10]

Hey all,

My current rig:

• CPU: i7 930 OC'd to 4ghz
  
• GPU: GTX 760 4gb, OC'd also (used by 337.50 and 337.88 with same effect)
  
• HDD: Arma is running on its own dedicated SSD
  
• RAM: 12gb Dual Channel
  
• OS: Win 7 64 bit Ultimate
  
• Settings:
  
  • General: http://i.imgur.com/fU3NWwf.jpg
    
  • Display: http://i.imgur.com/OFVoA5K.jpg
    
  • AA&PP: http://i.imgur.com/m6u0zg3.jpg
    

  [*]Startup Params: -cpuCount=4 -noLogs -noPause -noSplash -showscriptErrors -malloc=system

I've always gotten crappy performance out of Arma 3 but for some reason but since 1.20 its taken a nose dive. I'm capped at 15 fps during MP (Before the "its the server/gamemode", I regularly ask other players and they have 40+ frames at the same time on the same server) and can barely get 30 FPS in SP. I've just spent the entire day running every performance suggestion out there and reinstalling (such as http://day0.com.au/forum/arma/638-arma-3-performance-tweaks-and-settings-guide), also in my profiling Arma is utilizing about 20-30% of my CPU and 50% of my GPU. I even for science totally underclocked my system back to stock and I got exactly the same performance :O

I was wondering if anyone has experienced anything similar since the patch or if anyone has a similar rig that could help me understand my performance here.

Thanks,

---

Solved!

So I was able to identify the root cause of my issue. My Liquid Cooling pump had soft failed resulting in my CPU underclocking itself to 1.5 ghz :( under load.

With a new cooling solution in place I'm running at even better performance than before.

Edited June 2, 2014 by Nashable
Solved

[IT07 10]

do not use cpucount and the malloc launch parameter. Especially the malloc one should be removed. It can cause a lot of issues.

about the cpucount thing: it is bullshit too. It does not do anything to the performance of ArmA 3. So you might as well remove it.

[Fushko 59]

do not use cpucount and the malloc launch parameter. Especially the malloc one should be removed. It can cause a lot of issues.

about the cpucount thing: it is bullshit too. It does not do anything to the performance of ArmA 3. So you might as well remove it.

Correct, I'm fairly sure that the -cpucount command is only useful if you want to experiment running the game on less cores, otherwise it's automatically set by the game.

[IT07 10]

well, the ArmA 3 engine only uses 1 core. Not more. So cpucount is useless no matter what.

[nikiforos 450]

Especially the malloc one should be removed. It can cause a lot of issues.

about the cpucount thing: it is bullshit too. It does not do anything to the performance of ArmA 3. So you might as well remove it.

What kind of issues?

[byku 13]

Disable VSync.

[IT07 10]

What kind of issues?

Issues described by OP, memory write/read issues, game engine calculation issues, game stability issues, multiplayer availability issues, and so on....

[jumpinghubert 49]

well, the ArmA 3 engine only uses 1 core. Not more. So cpucount is useless no matter what.

-thats wrong. 45% minimum and 65% average with my quadcore, logged with afterburner in realtime.

-no issues at all with fred´s malloc tweak.

Edited June 2, 2014 by JumpingHubert

[IT07 10]

you do not have to believe me. All I know is that I did some really precise testing and simple common sense and it all came down to one thing: ArmA 3 is only using one core.

[Der_Richter 1]

you do not have to believe me. All I know is that I did some really precise testing and simple common sense and it all came down to one thing: ArmA 3 is only using one core.

Has been debated over and over again, and you ARE wrong. Simple as that.

Precisely measuring on my end shows pretty even spread on all 4 cores (measuring by processes and per core), with threads spawning all over the cores nicely. The cpuCount parameter is however useless, but for the simple reason that the engine assigns the threads automatically.

To return on topic, i am seeing slight changes in FPS as well, but not as major as described... Have anything else changed in your setup recently?

[IT07 10]

Has been debated over and over again, and you ARE wrong. Simple as that.

Precisely measuring on my end shows pretty even spread on all 4 cores (measuring by processes and per core), with threads spawning all over the cores nicely. The cpuCount parameter is however useless, but for the simple reason that the engine assigns the threads automatically.

To return on topic, i am seeing slight changes in FPS as well, but not as major as described... Have anything else changed in your setup recently?

ArmA 3 is not the only one using your cpu...

[Greenfist 1863]

ArmA 3 is not the only one using your cpu...

You're saying that the usage on other cores is something other than Arma? Don't think so. I see the A3 process using all 4 cores.

[ProfTournesol 956]

You're saying that the usage on other cores is something other than Arma? Don't think so. I see the A3 process using all 4 cores.

Same for me.

[IT07 10]

You're saying that the usage on other cores is something other than Arma? Don't think so. I see the A3 process using all 4 cores.

Ok let me get this straight: ArmA 3 UTILISES only 1 core. That means that it will use only 1 core to provide all the calculations that the engine is using.

All of the other things in ArmA like clouds and grass or whatever are being processed by other cores. So, ArmA itself IS only using one core for it's main frame: the engine.

[ProfTournesol 956]

Ok let me get this straight: ArmA 3 UTILISES only 1 core. That means that it will use only 1 core to provide all the calculations that the engine is using.

All of the other things in ArmA like clouds and grass or whatever are being processed by other cores. So, ArmA itself IS only using one core for it's main frame: the engine.

I'm curious to know how you could check that.

[IT07 10]

It's called experience and knowledge.

[Greenfist 1863]

So things like clouds, grass, AI aren't parts of then engine?

[IT07 10]

It's not just clouds. Clouds consist of 2 things: 1. cloud without texture, 2. cloud with texture.

The cpu does 1, gpu does 2.

So, the clouds are a part of the engine, but are rendered at a different level within the game. All of the important calculations by the engine are done by one core.

The rest of the stuff is done by other cores. If you see even performance on all of your cores; then you have set the in game settings to a certain level that it keeps your other cores busy too. But it does not mean that ArmA 3 is using all of your cores for the main frame: bottom level calculations.

[ProfTournesol 956]

It's called experience and knowledge.

Lol. True that im lacking of both.

[jumpinghubert 49]

you do not have to believe me. All I know is that I did some really precise testing and simple common sense and it all came down to one thing: ArmA 3 is only using one core.

so your overall cpu-usage is not more than 25%? And IF its more than 25% its windows system usage or other programs? When afterburner shows 65% overall cpu usage then arma3 uses obviously 2 cores @100% and a third core @15% (spreading over all cores). Compare arma3-benchmarks in the i-net on quadcores with deactivated cores, with only one core, two cores and so on. Try it and run for example helo´s benchmark. :)

[ProfTournesol 956]

For those who want to learn how RV is using Cores, particularly what concurrency means : http://www.bistudio.com/index.php/company/developers-blog/91-real-virtuality-going-multicore

[Greenfist 1863]

so your overall cpu-usage is not more than 25%? And IF its more than 25% its windows system usage or other programs?

No no no; Arma3.exe process uses all cores but the "core engine" uses only one core. Even though the game can't work with just this "core engine" alone, we can still say the game engine uses only one core.

(What he meant to say is that game's performance depends on one main thread which runs only on one core. Which is true. But the game still uses all the cores with its other threads.)

[Nashable 10]

How did this turn into a debate on how RV uses cores? :)

So I was able to identify the root cause of my issue. My Liquid Cooling pump had soft failed resulting in my CPU underclocking itself to 1.5 ghz :( under load.

[maddogx 13]

So I was able to identify the root cause of my issue. My Liquid Cooling pump had soft failed resulting in my CPU underclocking itself to 1.5 ghz :( under load.

So for once it would have been truly legitimate to say "it's your hardware's fault!", but nobody said it. BI community drops the ball again. ;)

[jumpinghubert 49]

No no no; Arma3.exe process uses all cores but the "core engine" uses only one core. Even though the game can't work with just this "core engine" alone, we can still say the game engine uses only one core.

(What he meant to say is that game's performance depends on one main thread which runs only on one core. Which is true. But the game still uses all the cores with its other threads.)

aaaahhhhh.....!!! Misunderstood it. Thanks for clarification. And sorry for the long run in offtopic.