Low CPU utilization & Low FPS

[Groove_C 267]

Greenfist you didn't understand me. There is a default -cpuCount param applied if you leave your startup params line empty. I know I can fill out manually the number of cores I actually have. But what is the default one? Dwarden?

[Greenfist 1863]

Greenfist you didn't understand me. There is a default -cpuCount param applied if you leave your startup params line empty. I know I can fill out manually the number of cores I actually have. But what is the default one? Dwarden?

The number is detected automatically. So it's the number of your cores. There is no particular default number.

[Groove_C 267]

Greenfist have a look here:

patch 1.07.71143 Improved: -cpuCount=4 is now default on computers with more than four logical CPUs. If you want to use more CPUs, use -cpuCount=N to override this.

[Greenfist 1863]

I didn't realize you were talking about a different game and a change from 4 years ago. :)

-cpuCount=<number> Change to a number less or equal than numbers of available cores. This will override auto detection (which equate to native cores).

I'm guessing the "native cores" means physical cores. Because -enableht enables the "extra" HT cores.

Edited October 14, 2014 by Greenfist

[Groove_C 267]

I will than use only -enableHT :)

Becasue I didn't know they've implemented autodetection. Thought there was certain number of cores set by default as for ArmA II.

Thx anyway :)

[Greenfist 1863]

I actually use only -enableHT :)

Then remove the -cpucount parameter completely.

[nikiforos 450]

Anyone tested the latest update and if performance is the same or better ? ( or worse) :)

[mac d 10]

Anyone tested the latest update and if performance is the same or better ? ( or worse) :)

It could be just me ,but i'm having quit a performance boost with this update.

Went from 60fps average during altis benchmark to 70 something fps now.

During scenario's i'm noticing everything run's very smooth ,way better then before :D

Don't know if the update helped with multiplayer ,I never play multiplayer :)

[Dwarden 1125]

don't use -cpuCount if you want use -enableHT ...

[tonschuh 3]

don't use -cpuCount if you want use -enableHT ...

-exThreads as well ?

:confused:

[Greenfist 1863]

-exThreads as well ?

:confused:

I don't think that's related to cpucount or ht. And it's already autodetected so you shouldn't have to use it anyway. (it's exthreads=7 for you)

[tonschuh 3]

I don't think that's related to cpucount or ht. And it's already autodetected so you shouldn't have to use it anyway. (it's exthreads=7 for you)

Thanks for the info !

:)

[ruPal 143]

Cpucount + exthreads gave me better performance than -enableht

[hyzoran 17]

Cpucount + exthreads gave me better performance than -enableht

-Enableht is only implemented in the server performance binary ATM. The ArmA 3 client does not yet (if ever) support it.

[Greenfist 1863]

-Enableht is only implemented in the server performance binary ATM. The ArmA 3 client does not yet (if ever) support it.

I don't think that's true. I mean; it's in the wiki, the launcher and changelog. Even Dwarden commented on it and he obviously knew we were talking about the client and not the server.

[raptor2x 10]

Hello, I can't post any thread so i'm asking my question here : Am I the only one who have these problems :

Only in multiplayer, most of time when I enter in cities..

[St. Jimmy 272]

Hello, I can't post any thread so i'm asking my question here : Am I the only one who have these problems :

Only in multiplayer, most of time when I enter in cities..

What hard drive you've? HDD or SSD? When was the last time you defragmented your hard drive (unless you've SSD)?

Also what is your power plan set to? Setting it on high performance should help if it already isn't.

Edited October 16, 2014 by St. Jimmy

[raptor2x 10]

Thank you for your answer.

I don't have an SSD, but I have friends who don't have an SSD and don't have my problems.

My HDD is a "Seagate Barracuda 7200.14 SATA 6Gb/s 1 To"

I defragmented it few days ago.

Sorry if I do some mistakes in my english.

[mercenar1e 10]

Anyone tested the latest update and if performance is the same or better ? ( or worse) :)

its just about the same for me.. its pretty disappointing that after so many updates the games performance hasn't improved.

Hello, I can't post any thread so i'm asking my question here : Am I the only one who have these problems :

Only in multiplayer, most of time when I enter in cities..

that my friend is the ARMA experience, its "normal".

[smiley_ie 7]

Hello, I can't post any thread so i'm asking my question here : Am I the only one who have these problems :

Only in multiplayer, most of time when I enter in cities..

If your running Nod32 disable HIPS

[Polymath820 11]

Sorry reality is Arma 3 is just not efficiently mufti-threaded... It is mufti-threaded but not efficiently. If you've noticed Arma 3 has a tendency to thrash 1 CPU core at a time? Reason? Arma 3 sets up the primary thread "loop for video" on 1 thread, that thread then creates 3 others the texture loading, geometry, and file-operations. If the primary thread is overloaded the 3 sub-threads lose performance, arma 3 I've noticed loses a lot of performance in geometry draw. Texture mapping not so much.

How do I know it's not multi-threaded? Arma 3 uses threads for the wrong reasons. Scripts and threads are designed for Bulk not atomic

https://software.intel.com/en-us/articles/designing-artificial-intelligence-for-games-part-4

#Check FSM's and AI states

Overly complex systems. If your subsystem is tied in with too many other systems that constantly cause the subsystem or other systems to wait, threading might not be the answer. There is also a chance that the system itself might benefit from a redesign, if this is the case.

#Check arma 3 uses atomic work-loads in all scripts run on the server and are run as individual threads

Atomic workload. If the work of the subsystem cannot be broken down, you might not be able to go parallel. The audio mixing task may work great as a thread, but the work of the task is to mix multiple sounds into the channels that are ultimately pumped to the speakers. If your system does calculations on individual chunks of audio before the mix, then it may be possible to thread that.

#Check arma 3 uses threading for the primary video-thread loop bad bad bad don't do this.

Repeating code. There are cases where multiple threads end up churning away at the same code, only to have some of it thrown out or ignored. Most of this can be avoided with redundancy checks before work begins.

#Check AND MATE

Costly overhead. There is a bit of extra work in some of these systems (pathing, for instance). If the overhead outweighs the benefits gained from threading, then it might be a good idea to avoid it or just turn it off. This may be true for systems with small numbers of elements (entities, paths, etc.).

Edited October 20, 2014 by Polymath820

[Gera_CCT 12]

its just about the same for me.. its pretty disappointing that after so many updates the games performance hasn't improved.

that my friend is the ARMA experience, its "normal".

Not normal to me. I have a WD Blue 1Tb and use O&O Degrag every week.

[niktator 10]

I wonder if we will ever see this thread beeing obsolete.

Performance issues can and will ruin lots of fun.

So why do we get DLC like kart stuff(Was meant as a joke, once upon the time on some random 01.04.), if there are way bigger issues?

If DLC stuff is not related to this problem, since different ppl are working on such things - I`d say the budget should be moved over to the "engine department" asap.

http://feedback.arma3.com/view.php?id=716

2421 votes at lets say in average (only)24 Euro... Thats 58104 Euros of paying customers to get this done. Well ppl like me do love the game, so we did not buy the solution of this bug - but how many ppl don`t even register at the tracker to vote/complain? I`d say more then 90+% of the customers do not vote in the tracker.

How about a feature/content freeze, until #716 is fixed or at least improved?

Kinda regards, Nik

[shankly1985 1]

I'll never buy another Arma game again. I am completely done with this JUNK!

I have not once had a chance to sit down and enjoy this game. The performance is just simply unplayable! No matter the settings I use I continue to run 20/30fps and CPU i7 2600k even @3.4ghz uses only 50% load

This is shocking, how much this as been reported and not a single word to us or a fix!

[BluesAdam 10]

Hi everyone,

As a fellow Arma fan, I've bought Arma 3 several days ago. The low fps really was an issue and no fix seemed to cure the fps, so I decided to make an in-depth analysis as to how the game utilizes CPU and had gathered some interesting data and felt like sharing with you. First of all, here are my system specs:

---

CPU: AMD FX6350 3.9ghz (no overclock)

GPU: Asus AMD R9 270 2gb 256bit DirectCU II (no overclock)

RAM: G-Skill Ripjaws 4gb 1600mhz DDR3 Single Stick

Motherboard: Gigabyte 990xa UD3

PSU: Cougar CMX 700w Modular

---

Graphics Settings:

Resolution: Native---------------------Bloom: Off

Texture:Ultra--------------------------Radial Blur: Off

Objects:Ultra---------------------------Rotational Blur: Off

Terrain:Ultra---------------------------Depth of Field: 100

Shadow:Ultra--------------------------Sharpen Filter: 100

Particles:High--------------------------FSAA: 2X

Clouds:Ultra---------------------------ATOC: All Trees

PIP:Ultra-------------------------------PPAA: Smaa Ultra

HDR:Standard-------------------------Aniso. Filtering: Ultra

Dynamic Lights:Ultra-----------------AO: HDAO Standard

Caustics:Disabled

Visibility: 2000

V-sync: Disabled

First of all, I should give information about several cpu features I tested in bios, they are as follows:

(Note: Those features might be CPU/Motherboard specific)

1- Core Performance Boost:

Simply put, when equal or less than 2 cores are in use, this feature boosts the core clock speed for those cores. It is disabled for me as Arma 3 uses at least 3 of 6 cores for me. You can set the boost value if enabled.

2- Cpu Unlock:

This feature is mainly for Phenom CPUs that come with locked cores. This feature unlocks those cores. In FX series CPU, there is no need to enable it.

3- Cool n' Quiet*/**:

This is a power saving feature. It simply changes voltage values of CPU depending on CPU load. I've tested this both enabled and disabled and results will be given below.

4- C1E Support*:

"C States" are idle and halt states for CPUs. When CPU is off-load, these feature stop CPU activity, thus saving power. No need to disable those, as C State functions do not affect CPU while it is under any given load.

5- SVM (Secure Virtual Machine Mode):

This feature has no affect on gameplay whatsoever. It only affects virtual machine usage (virtual OS etc.) So enable it or disable it, it's your choice.

6- CPU Core Control:

This is a feature I've not yet tested. It stays on automatic for me. To explain this feature, I should give a little information about how AMD FX series CPUs work. For 6 core FX CPUs, there are 3 units, with 2 cores each, which gives a total of 6 cores. Using this function, you can set it to "1 core per each computing unit," so you will essentially have 3 cores, but with quicker response times. Again, as ARMA uses at least 3 and at most 5 cores, I let this at auto.

7- Core C6 State*:

Again C State is related to idle and halt state power management. Leave this enabled.

8- HPC Mode*:

Simply put, this feature disables CPU from lowering its clock speed under maximum voltage load, providing stability.

9- HPET** (High Precision Event Timer):

HPET is a modern timer used in modern OS, this regulates the timing of CPU cores. If you have an OS that doesn't support it, leave it disabled, if your OS supports it, leave it on. However, enabling this function on bios is only the half of the process. You have to enable this on OS, too. In order to do this, type "cmd" in start menu, right click and run as administrator. After command prompt is on, type:

bcdedit /set useplatformclock true

now it should say "operation is successful." Restart your pc, and congratulations, HPET is officially on :)

------

*These settings are mostly set to "disabled" by overclockers as they might cause unstability on overclocked CPUs. As I am not into overclocking, I set them enabled except for Cool n' Quiet.

**These two settings are the main variables for testing phase. I've tried them both enabled and disabled and results will be given below

Testing! Testing!

Finally, here we go! First test I did was Cool n' Quiet enabled and HPET enabled. First, I tried running Infantry Showcase, and then played on Patrol Ops servers with 2-4 players in it. Results are below:

Infantry Showcase--------------------------------Multiplayer (Patrol Ops)

FPS=40-42---------------------------------------------------------FPS: 27-31

CPU Load:---------------------------------------------------------CPU Load:

Core:0= 85-95%--------------------------------------------------70-90%

Core:1= 40-60%--------------------------------------------------35-55%

Core:2= 40-60% (Rarely spikes up to 80%)--------------------45-55%

Core:3= 20-60%--------------------------------------------------0-15%

Core:4= 45-70%--------------------------------------------------40-55%

Core:5= 0%-------------------------------------------------------35-55%

The results shown that 5 out of 6 cores are being used and only Core #0 goes above 90%. Rest stays around 40-50%. On multiplayer, however, cores except core#0 is utilized less than in Singleplayer mode. This is the main cause of FPS drop as far as I've noticed. Some of the servers I played on actually utilized 3 of 6 cores and had terrible FPS (13-15). So, while playing online, choose a server that utilizes at least 5 cores. The major reason of fps drop is the server.

Test #2

Next test is made with Cool n' Quiet disabled and HPET still on.

Infantry Showcase--------------------------------Multiplayer (Patrol Ops)

FPS=40-44---------------------------------------------------------FPS: 30-39

CPU Load:----------------------------------------------------------CPU Load:

Core:0= 88-97%--------------------------------------------------75-90%

Core:1= 30-50%(Rarely spikes down to 0%)-------------------38-50%

Core:2= 60-78%--------------------------------------------------50-55%

Core:3= 30-60%--------------------------------------------------0% (rare spikes up to 20%)

Core:4= 60-70%--------------------------------------------------50-60%

Core:5= 0%-------------------------------------------------------35-50%

Turning Cool n' Quiet off actually helped with fps a little bit. Also, in multiplayer, now Core#3 is unused rather than Core#5, I don't know if that has any impact on fps but, CPU usage spikes which randomly went down to 0% also disappeared. Instead, now the unused core randomly spikes up to 20% :)

Last Test!

The last test is performed with HPET disabled and Cool n' Quiet disabled as well. I wondered what impact HPET would have, so here we go.

Infantry Showcase--------------------------------Multiplayer (Patrol Ops)

FPS=35-44---------------------------------------------------------FPS: 29-31

CPU Load:---------------------------------------------------------CPU Load:

Core:0= 85-90%--------------------------------------------------70-85%

Core:1= 40-50%--------------------------------------------------25-40%

Core:2= 30-50% (random spikes down to 0%)-----------------40-60%

Core:3= 30-70%--------------------------------------------------0%

Core:4= 60-70%--------------------------------------------------0%

Core:5= 0%-------------------------------------------------------40-50%

Having HPET disabled, I've noticed huge spikes of FPS drop. Especially while zooming in and out, making sharp turns etc. HPET seems to reduce lag/fps spikes when enabled as far as I've seen. When HPET is disabled, 2 of my cores actually idled at 0% in multiplayer which might be the cause of FPS spikes. Plus, when pc restarts, it actually gains consciousness and becomes responsive a lot quicker.

I noticed that singleplayer utilizes cores more efficiently than multiplayer servers. This might be related to the server computer's system specs and the fact that the user missions have a lot of added scripts, add-ons etc. Another funny fact is that there is only 1-2 fps difference between highest and lowest settings which means GPU actually is not affected at all. Performance is all up to CPU as far as I've seen. It looks like our fate is sealed for now. Unless a major patch comes out and provides us with better CPU utilization, we're bound to have fps drops during Multiplayer experience.