Low CPU utilization & Low FPS

[xXxatrush1987 10]

Regardless of how it stacks up, me and a lot of people should be getting higher framerates.

You can see people here with even more powerful rigs like 2 4gb version 680's who are getting similar framerates with me.

that a gtx 680 with 2 gb and 4gb or hd5870 get all similar fps is due to its bandwith. just in case the cpu is not bottlenecking, the 256bit interface even on high clock speeds running gddr5 is just too small.

thats why hd79xx/titan with 384bit interface and even gtx580/480 get in most cases best fps.

and having 2 hd5870 or 2 gtx680(gtx690) doesnt double the bandwith, because in sli or crossfire gpus can´t share data good, because pcie connection then bottlenecks in a game like arma3.

mfg

[tremanarch 6]

Putting a whole lot of AI in the editor and getting low FPS is understandable. What isn't understandable is playing MP with everything on low, no particles on screen, yet getting 12 FPS. And, for the naysayers whose games happen to run ArmA3 just fine, your FPS should improve when you turn away from particles or vegetation (like, for example, looking at the ocean with your back to the island). The FPS decreases over time too, indicating poor resource management by BIS. What I mean by this is, if you are in an area with a lot of vegetation and/or particle effects, and you move to an area where there are minimal vegetation and particle effects, the game should in essence not factor that in. I understand always computing/calculating AI and stuff, but it shouldn't do the same for vegetation and particle effects. So, in essence, for the client, it should delete those from memory and then reload that when you come near those. Does what I'm saying make sense anyone?

you didnt read what i wrote. so i make it short:

solo editor 50 fps

a lot AI editor 15 fps -> after 20 mins (all AI dead) 35 fps... ::not as much as solo editor.

[tonygrunt 10]

that a gtx 680 with 2 gb and 4gb or hd5870 get all similar fps is due to its bandwith. just in case the cpu is not bottlenecking, the 256bit interface even on high clock speeds running gddr5 is just too small.

thats why hd79xx/titan with 384bit interface and even gtx580/480 get in most cases best fps.

and having 2 hd5870 or 2 gtx680(gtx690) doesnt double the bandwith, because in sli or crossfire gpus can´t share data good, because pcie connection then bottlenecks in a game like arma3.

mfg

Did you test for this or are you just assuming?

Good thing that we have overclocking software and we can rule out a thing like video ram bandwidth. My video ram at 3.1GHz and 3.6GHz gives the same frames when I am not GPU limited.

Should I swap my GTX680 4GB for my old GTX480 1.5GB? lol.

And you think ArmA3 (or any other game for the matter) can become bottle necked in a PCIEx16 3.0 connection? Don't think so.

[INVICTUS 10]

As a test last night I loaded the game on my sons computer. I found that by turning clouds to disabled I got a large performance increase. He has a AMD 5870 GPU. I have an i7 2600k 4.0Ghz, 16GB DDR3, 60GB SSD, 2x GTX 570 SLI no issues with performance 50-75 FPS in MP depending on mission. But my sons machine AMD Phenom 940 @ 3.9Ghz, 8GB DDR3, 500GB HD, HD5870 has issues with low fps. I teaked the settings to a fairly playable state 10-40 FPS MP. The biggest difference for his machine was the Clouds setting. Turning it off really helped there are still clouds just not cpu intense ones.

[white 1]

As a test last night I loaded the game on my sons computer. I found that by turning clouds to disabled I got a large performance increase. He has a AMD 5870 GPU. I have an i7 2600k 4.0Ghz, 16GB DDR3, 60GB SSD, 2x GTX 570 SLI no issues with performance 50-75 FPS in MP depending on mission. But my sons machine AMD Phenom 940 @ 3.9Ghz, 8GB DDR3, 500GB HD, HD5870 has issues with low fps. I teaked the settings to a fairly playable state 10-40 FPS MP. The biggest difference for his machine was the Clouds setting. Turning it off really helped there are still clouds just not cpu intense ones.

tried here, also tried turning off pip, changed nothing in my heli test, at least it wasnt noticiable to me. even turning on aa to 8x changes nothing, its too cpu bound.

[Minoza 11]

Check this out:

PvP scenario, tossed a lot of smoke to test what exactly will happen -> http://www4.picturepush.com/photo/a/12360102/img/Anonymous/game.jpg

Observation:

Huge FPS loss, as expected but, interesting find, check out what happens with GPU usage after tossing smoke -> http://picturepush.com/public/12360098

1. GPU usage drops below 30% (right half of the graph)

2. If you observe left side of graph you will notice some positive peaks and drops

Positive peaks happened whenever I used scoped view, zooming in, less work for GPU, usage goes up to 99%, framerate goes up. The more the GPU gets stressed the more usage drops. Whenever I exited scoped view, I got a drop to aprox 50% usage.

CPU never got over 50% on most active core, I disabled HT -> http://www3.picturepush.com/photo/a/12360096/img/Anonymous/cpu.jpg

Edited March 7, 2013 by Minoza

[white 1]

Check this out:

PvP scenario, tossed a lot of smoke to test what exactly will happen -> http://www4.picturepush.com/photo/a/12360102/img/Anonymous/game.jpg

Observation:

Huge FPS loss, as expected but, interesting find, check out what happens with GPU usage after tossing smoke -> http://picturepush.com/public/12360098

1. GPU usage drops below 30% (right half of the graph)

2. If you observe left side of graph you will notice some positive peaks and drops

Positive peaks happened whenever I used scoped view, zooming in, less work for GPU, usage goes up to 99%, framerate goes up. The more the GPU gets stressed the more usage drops. Whenever I exited scoped view, I got a drop to aprox 50% usage.

CPU never got over 50% on most active core, I disabled HT -> http://www3.picturepush.com/photo/a/12360096/img/Anonymous/cpu.jpg

perhaps smoke is being calculated by the cpu and it being more busy made the fps drop althought the gpu was free? and when u focus away from cpu intensive things the gpu roams free for high fps? just my hipothesis.

when u have 4 cores like its showing there, windows switches the load between them but the game never uses more than 2 cores, thats why its 50% for you and 30-35% for me (six cores). it shows 35% on all of them, which is exactly the same as 2 full cores (screenshot on my post on the previous page).

[Sephisto 1]

Without launch parameters. Using RAMDisk.

With launch parameters. Using RAMDisk.

Edited March 7, 2013 by Sephisto
edit

[shankly1985 1]

Before:

(FPS here was 38)

http://i.imgur.com/vao0Ha0.jpg

After:

(Added more launch parameters here, currently -malloc=system -nobenchmark -maxVRAM=2047 -nosplash -noPause -skipIntro -world=empty -noFilePatching -maxMem -cpuCount=4)

http://i.imgur.com/7goXuHL.jpg

Thing with that, you have moved from the place image 1 and 2. To test you will need to be in the same place for both changes and see if FPS change.

[Sephisto 1]

Thing with that, you have moved from the place image 1 and 2. To test you will need to be in the same place for both changes and see if FPS change.

Good point ;o edited last post

Edited March 7, 2013 by Sephisto

[tremanarch 6]

and the fps is going up and down. the point in time when u press Screen shot save might not be the same in both pics - might be more or less AI involved, different AI activities - not comparable. dont u have an AVG FPS Multimeter?

[Sephisto 1]

Go in-game and do it yourself then :P if that's not comparable then I don't know what is.

[MavericK96 0]

about 50% load, could be worse.

Actually, considering it seems to be a CPU-bound game in a lot of ways, that's pretty terrible.

[ephant 10]

i5 2500k @4GHz, 6950, 8gb ram 1333mhz 7CL, SSD

1920x1080

no vsync

Visibility

Overall: 1700m

Object: 1100m

AA, PPAA, Atoc, Post Proc. disabled

HDR: Standard (didn't notice a loss in fps so it's personal preference)

Anistropic Filtering: High

PIP: Low (is enough imo)

Dynamic Lights: Standard (didn't test them at night yet)

Texture: High

Object: Standard

Terrain: Standard

Cloud: Disabled (not a huge fps difference between of and ultra)

Shadow: Disabled

Particle: Standard

Infantry Showcase beginning: 83-90fps

Infantry Showcase 1st firefight: 72-80fps

Infantry Showcase 2nd firefight: 68-7xfps

My whole squad died so an ran into the village to see the fps there. Dropped to 58-7xfps

Helicopter Showcase beginning: 41-42fps

Shadows, AA, Visibility, Terrain Quality and Post Processing kills your fps. Also AI...

[shankly1985 1]

Good point ;o edited last post

very little change.

[Sephisto 1]

very little change.

Yeah, and for some reason it decided to use different CPU threads after adding launch parameters. Don't know what to make of that, probably no big deal.

[RadicalAtHeart 11]

Put Terrain Quality on low for a huge CPU improvement. I also noticed the amount of objects drawn makes a difference. Not object quality.

Looks very promising. I am going to try to delay my laptop to Haswell.

[Sephisto 1]

Disabled Hyperthreading. With Launch Parameters. Using RAMDisk

No difference really. CPU thread 1 usage high, others below 50%. Fail.

[Leon86 13]

Actually, considering it seems to be a CPU-bound game in a lot of ways, that's pretty terrible.

Multithreading complex stuff is hard. I get 50-70% cpu use on my quad, pretty decent imo.

and sephisto, your gpu is at 99% load.

[Sephisto 1]

Multithreading complex stuff is hard. I get 50-70% cpu use on my quad, pretty decent imo.

and sephisto, your gpu is at 99% load.

What about it?

[Leon86 13]

What about it?

if your gpu is at 99% load that means the cpu is not a bottleneck.

[kazenokizu 10]

what is command to disable hyper threading ? and can any body help me with my frames. im on a amd 8120 8 gigs of ddr3 ram a ssd with a 7970

[stk2008 14]

what is command to disable hyper threading ? and can any body help me with my frames. im on a amd 8120 8 gigs of ddr3 ram a ssd with a 7970

I think you disable it VIA the BIOS but im not to sure

[white 1]

what is command to disable hyper threading ? and can any body help me with my frames. im on a amd 8120 8 gigs of ddr3 ram a ssd with a 7970

afaik only intel has hyperthreading. the only thing you can do is overclock as high as you can, since you have "8 cores" and this games uses only 2.

[SavageCDN 231]

you didnt read what i wrote. so i make it short:

solo editor 50 fps

a lot AI editor 15 fps -> after 20 mins (all AI dead) 35 fps... ::not as much as solo editor.

Ah but that is not a good test unless you had a script delete all the bodies and other ground objects (blood, casings, dropped weapons, etc)