ArmA3Mark - Benchmark your ArmA 3

[konjad 10]

Core 2 Duo 2.10 GHz

4 GB RAM DDR2 800 MHz

ATI Mobility Radeon HD 4650 (1024 MB VRAM)

5400 RPM HDD

Windows 7 64 bit

ArmA3 version: 0.77

Resolution: 1366x768, sampling 100%

Low: 30 FPS

Low with High textures: 30 FPS

Didn't bother with higher settings.

[jumpinghubert 49]

with current dev build mega stutter is completely gone (second run).

sampling: 100%

display resolution: 1680x1050

texture: very high

objects: very high

terrain: very high

shadow: ultra

particles: high

clouds: ultra

pip: low

HDR: standard

dyn. lights: ultra

visibility

overall: 3000

object: 3000

shadow: 200

vsync. disabled

bloom: 100

radial blur: 0

rot. blur: 200

depth of f.: 50

SSAO: disabled

caustics: enabled

FSAA: 8x

ATOC: all trees + grass

PPAA: smaa standard

AF: ultra + 16x forced in nvidia control panel

result: 44 fps (second run)

nvidia driver: 320.14

specs in sig

Edited September 2, 2013 by JumpingHubert

[tonygrunt 10]

.........

result: 44 fps (second run)

nvidia driver: 320.14

specs in sig

Tried your settings and all I got was 41 fps. I would expect a bit more or worst case scenario the same. :(

PC specs: i7 3930k HT @ 4,8GHz, 16GB 2133 MHz, Win7 64bit on an OCZ Vector 128GB SSD, Arma 3 on a Samsung 840 Pro 500GB SSD, 2 SLI-ed GTX 680 4GB @ 1240 MHz.

Win 7 installation is less than a week old running nvidia 326.98. Maybe the beta nvidia drivers are at fault.

Don't you have any stutters?

Cause with me it when the 2 smokes pop it's a slideshow. Also small stutters when the scenes are quickly panning.

Edited September 2, 2013 by TonyGrunt

[jumpinghubert 49]

first run I have mega stutters, second no stuttering. In first run with its stuttering I get 35 fps.

thats completely crazy your result is lower. Do you have set "performance mode" in nvidia driver panel, maybe throttling/idling cpu or gpu?. The benchmark is more cpu-heavy so I am wondering your oc´d sixcore and fast ram give these results.

edit:

My screen resolution is very low (1680x1050) compared to yours, maybe thats the answer (ssd streaming bandwith-thing?). I bet if you lower the sampling to 22"er resolution your result will be significantly higher.

Edited September 2, 2013 by JumpingHubert

[Walsh.D 0]

Figured it out :)

Edited September 2, 2013 by Munny

[tonygrunt 10]

first run I have mega stutters, second no stuttering. In first run with its stuttering I get 35 fps.

thats completely crazy your result is lower. Do you have set "performance mode" in nvidia driver panel, maybe throttling/idling cpu or gpu?. The benchmark is more cpu-heavy so I am wondering your oc´d sixcore and fast ram give these results.

edit:

My screen resolution is very low (1680x1050) compared to yours, maybe thats the answer (ssd streaming bandwith-thing?). I bet if you lower the sampling to 22"er resolution your result will be significantly higher.

I set Arma at the same resolution and scaling is done by the monitor so it shouldn't matter. I tried the nvidia settings but nothing changed, anyway as you said test is mostly CPU as both gfx cards are mostly idling at 50% usage without going at their maximum (Hz) speed.

Things to check now, disable HT, disable part of memory with msconfig, change gfx drivers, disable SLI etc.

The fun never stops when you are playing with Arma 3. :) :icon_neutral: :( :banghead:

[jumpinghubert 49]

the only idea i have is to log your cpu-clock ingame. Ensure all cores are @ max clock. I am using HWinfo64 to log the clocks of all cpu-cores........dis is armaaaa :p

edit: the 44er run without stuttering was luck.....averaged I have 43 fps in second run with minimal stutter in the green/red smoke scene.

edit²: after first run ended don´t press "space" to continue, press "esc" and then "restart" because with "space" I get the same heavy stutter like in first run.

Edited September 2, 2013 by JumpingHubert
:p

[Walsh.D 0]

i5 2500 (not K - stock speed)

16 gigs ddr3

rADEON 7990

samsung 256 ssd

win 7 64 bit

1900x1080 100%

Stratis

26 fps on Ultra

43 fps on high

Disappointing -

on the altis map it gave me a couple errors at the beginning and then there was no landmass at all, everything just occurred over water, any fix?

[tonygrunt 10]

on the altis map it gave me a couple errors at the beginning and then there was no landmass at all, everything just occurred over water, any fix?

Do you have the development build selected?

Do you see Altis in the editor?

If yes try to "Verify Integrity Of Game Cache".

edit:@JumpingHubert, I started playing with all the command line parameters and sometime in between runs it reached 45 fps.

Reverted all changes and still get 45 fps, can't find any reason why it didn't reach those fps before. :confused:

edit2:@JumpingHubert, disregard above, it was the -maxmem=2047 command line parameter. With it I get 41 fps, without 45 fps. Didn't catch it 1st time due to a mistype.

Edited September 2, 2013 by TonyGrunt

[Walsh.D 0]

Yep on the dev build, I havd made mission and played missions on altis, see it in my editor and Verify came back fine, with all files intact

^^ Screencaps of ht errors

Weird

[griffz 1]

i5 2500 (not K - stock speed)

16 gigs ddr3

rADEON 7990

samsung 256 ssd

win 7 64 bit

1900x1080 100%

why are you using 100% sampling with a such monster GPU ? it is pointless. use a 7790 instead .

crank up sampling to 200% , remove or reduce FSAA. a such card is MADE for that

[tonygrunt 10]

why are you using 100% sampling with a such monster GPU ? it is pointless. use a 7790 instead .

crank up sampling to 200% , remove or reduce FSAA. a such card is MADE for that

From my tests sampling is cpu dependent and the image quality is worse with no FSAA applied.

Using my previous tests where I reached 45 FPS, when I set sampling to 200% and disable FSAA my score goes to 34 FPS. Both of my cards start to run at lower speed due to the lowered GPU usage. Image quality is worse with visible flickering edges in fences.

[Walsh.D 0]

why are you using 100% sampling with a such monster GPU ? it is pointless. use a 7790 instead .

crank up sampling to 200% , remove or reduce FSAA. a such card is MADE for that

tbh I don't even know what sampling does... That was how autodetect set up my sliders

[griffz 1]

image quality will not be worst. as sampling 200% is equal to supersampling 2x. the picture is rendered in 2X resolution, then is scaled down to "display" resolution. it's similar to enbserie antialiasing hack for GTA4. very demanding, but visually great and flawless. (not FSAA artefact in some lighting condition etc ...). you don't even need AtoC anymore...

you can't reach this quality with FSAA alone . let alone PPAA and its pixel "blinkering". everything is smooth , even in motion. trooper are harder to spot too ;)

200% plus FSAA 2x

the reason you feel that cpu dependency , is either SSAO (the more pixel , the slower), and / or geometry LOD wich tend to increase with pixel number too.

Edited September 3, 2013 by griffz

[jumpinghubert 49]

Both of my cards start to run at lower speed due to the lowered GPU usage.

Is it possible to fix the gpu-clock? My old GTX570 hasn´t the dynamic overclocking feature.

[tonygrunt 10]

Is it possible to fix the gpu-clock? My old GTX570 hasn´t the dynamic overclocking feature.

I think you might be able with nvidia inspector and definitely with bios mod but why would you?

There would be minimal performance increase at best as the cards at the time have not much to render, reading the gpu usage.

Can you try running the benchmark with the -maxmem=2047 parameter and see if you have lowered performance?

[jumpinghubert 49]

I think you might be able with nvidia inspector and definitely with bios mod but why would you?

There would be minimal performance increase at best as the cards at the time have not much to render, reading the gpu usage.

Can you try running the benchmark with the -maxmem=2047 parameter and see if you have lowered performance?

yes yes i know: to increase gpu clock will lower arma-gpu-usage so it will not result in better fps. It was only a suggestion to test it :). I will bench with maxmem when I am at home :)

please remember to end the first run with pressing escape and then repeat, not space to continue because then will appear a loading screen before you can start new run. It will eat the loaded data i think.

edit: no difference with maxmem parameter, second run no stutter and 43-44 fps.

Edited September 3, 2013 by JumpingHubert

[Walsh.D 0]

Stratis

26 fps on Ultra

43 fps on high

Disappointing -

Tried using 200% sampling - got 21 fps with fsaa and ppaa both turned on. got 22fps with 200% and fsaa and ppaa both turned off

[msilva 2]

Every time i run this benchmark in Altis, for the first time i get low FPS(20-22). From the second run to next i get a proper FPS(31-32).

Is something wrong? Is supposed to be so?

[helo 10]

I have the same problem.

THe reason behind this is that by the first time the vast amount of textures and objects will be loaded into your memory slowing everything down in the first time. When you run it a second time, then all the textures/objects are ready and t

[xXxatrush1987 10]

sorry for off topic but for ultimate gfx quality you set your catalyst antialiasing mode to supersampling(which applies rgssaa) wich with 4xfsaa setting in arma menu is much more beatiful than the sampling rate which simply applies something like ogssaa.

ordered grid supersample antialiasing(ogssaa) or downsampling just heats up gpu for no real benefit.

rotated grid supersample antialiasing(rgssaa) filters every aliasing away with multiple(!) random sprayed supersample grids.

as rgssaa is officially supported by nvidia and amd drivers so the performance is much better than the sampling from arma 3/2.

for every one who don´t like blurry aliased pictures, he could resharpen the picture with fxaa, or everyone who cant crunch rgssaa on his card could use smaa to blur the aliasing effects for just little gpu load cost.

for a competitive game i chose 4xfsaa(rgssaa from driver)+ fxaa ultra, getting 40-80fps in kavala on a non oc single 7970. a super sharp but smooth picture, almost no flickering and easy to see movements.

[msilva 2]

I have the same problem.

THe reason behind this is that by the first time the vast amount of textures and objects will be loaded into your memory slowing everything down in the first time. When you run it a second time, then all the textures/objects are ready and t

Thank you. So I can assume that the second measurement is the most accurate

[jumpinghubert 49]

sorry for off topic but for ultimate gfx quality you set your catalyst antialiasing mode to supersampling(which applies rgssaa) wich with 4xfsaa setting in arma menu is much more beatiful than the sampling rate which simply applies something like ogssaa.

ordered grid supersample antialiasing(ogssaa) or downsampling just heats up gpu for no real benefit.

rotated grid supersample antialiasing(rgssaa) filters every aliasing away with multiple(!) random sprayed supersample grids.

as rgssaa is officially supported by nvidia and amd drivers so the performance is much better than the sampling from arma 3/2.

for every one who don´t like blurry aliased pictures, he could resharpen the picture with fxaa, or everyone who cant crunch rgssaa on his card could use smaa to blur the aliasing effects for just little gpu load cost.

for a competitive game i chose 4xfsaa(rgssaa from driver)+ fxaa ultra, getting 40-80fps in kavala on a non oc single 7970. a super sharp but smooth picture, almost no flickering and easy to see movements.

I want to change from my GTX570 to a HD7970. Whats your cpu, the settings and your benchmark result? Thanks in advance.

[xXxatrush1987 10]

45fps in second run, the altis edition.

all max except for:

pip standard

viewdistance 3.5km

objectdistance 2km

atoc off(the most ugly antialiasing ever...)

fsaa x4 with supersampling set in catalyst.

core i7 2700k @ 4.5ghz

cpu is bottlenecking all the time with 60-80% load all cores, faster ram over 1866 should help either. for multiplayer i set objectdetail on standard, as stutter free city fights tend to be more fun.

most benefit of the 7970 is the wide 384bit interface with 3gb ram, so with a 780gtx or titan it should be possible to get even more fps in gpu heavy scenarios.

the gtx570 off my girlfriend is outperformed by the 7970 by allmost 200%(!) with much less stutter due to the wide memory interface and more than double videomemory.

11fps on her machine on same settings^^

but just 1066 ddr3, core i 7 920 and most important, no ssd(yes i should give her pc some love too...maybe^^)

[griffz 1]

sorry for off topic but for ultimate gfx quality you set your catalyst antialiasing mode to supersampling(which applies rgssaa) wich with 4xfsaa setting in arma menu is much more beatiful than the sampling rate which simply applies something like ogssaa.

ordered grid supersample antialiasing(ogssaa) or downsampling just heats up gpu for no real benefit.

rotated grid supersample antialiasing(rgssaa) filters every aliasing away with multiple(!) random sprayed supersample grids.

as rgssaa is officially supported by nvidia and amd drivers so the performance is much better than the sampling from arma 3/2.

thanks for your input. i didn't think to test the "standard" way. didn't know you can force AA method in DX11 title...

it seems forcing SSAA doesn't work for me.

Edited September 4, 2013 by griffz