NVIDIA profile setting change = FPS boost

[SeaVee 10]

I just discovered an inexplicable but welcome increase in FPS when changing NVIDIA's "threaded optimization" option in the NVIDIA control panel for the Arma2-specific settings.

When I set it to AUTO, I got 14 FPS in Benchmark 2. When set to either ON or OFF as opposed to AUTO, my FPS was 16 with the same benchmark... Tried it back and forth several times and the FPS gain/loss is repeatable each time. Going from 14 to 16 FPS is a 15% improvement.

I have an I-7 920 overclocked to 3.8GHZ so for users without multi-core CPUs this may not apply or be an option. I am running NVIDIA's 197.13 drivers and also twin GTX260 cards with SLI and Physx enabled. Also running XP32.

I've read that Arma2 is NOT multi-threaded so I have no idea why this option would have an effect on FPS. Maybe the NVIDIA setting is something specific to the GPUs rather than the game .exe? At any rate why would AUTO drop FPS while ON or OFF boost it?

The FPS results above were run with ALOT of Arma2 add-on mods which are certainly more demanding on the system than plain vanilla Arma2, specifically:

"C:\Program Files\Bohemia Interactive\ArmA 2\arma2.exe" -mod=@CBA;@sthud;@ACE;@ACEX;@sakusun;@HiFi_ACE;@HiFi_Air;@HiFi_SFX;@HiFi_Vehicle_Weapons;@HiFi_Weapons_Long;@WarFX_Particles_Beta_1.41;@GL4 -nosplash -world=empty

The mods make a big difference in FPS in an intensive, system straining benchmark like this. In regular gameplay where there is rarely, if ever so much going on the FPS hit is not nearly as bad. And for me the improvements are more than worth it. If I run the same graphic settins but no mods on a bone-stock plain vanilla Arma2 my FPS rises to 23 in Benchmark 2 - thats nearly a 40% difference from 16FPS. Dropping GL4 alone from the mods above makes a 4 FPS difference for me for example in that benchmark 2. Ace2 is about a 2-3 FPS hit.

My in-game video options settings:

Also in NVIDIA control panel I have:

Vsync & Triple buffering = forced ON

In "regular" singleplayer gameplay versus Benchmark 2 I am getting a low of around 42FPS and the majority of the time 55-60 (with Vsync on I am maxed at 60FPS as that is my monitor's refresh rate).

PS: I had posted this in the "NVIDIA owner's threa" but it will get buried there and not sure many will see it so I made a separate thread.

Edited March 28, 2010 by SeaVee

[Alex72 1]

Nice will test. However i have never seen any difference between high/very high in post processing. What is the difference? I have changed it back and forth many many times but cant see any change.

[suma 8]

I've read that Arma2 is NOT multi-threaded

I have no idea where could you read this. ArmA 2 definitely IS multithreaded. As for driver threading settings, it is quite possible (and we have already seen it while testing) that if the drivers try to perform their own threading, they interfere with the game threads (CPUs being overused by too many threads).

[-=seany=- 5]

So disabling this may actually be a fix? e.g. it wont affect any of Arma's own multi threading?

must give it a go then

[11aTony 0]

Well Im glad you have a boost SeaVee but Threaded Optimization and Triple Buffering has no effect on my system whatsoever. Only thing that makes a difference is Forcing VSync OFF and even that is really minimal, almost nothing.

[AnimalMother92 10]

How much does Arma 2 benefit from a having quad-core over a dual-core? I've been trying tweaking various things recently and my frame rate seems similar with cpucount=2 or 4 :confused:

[Master gamawa 0]

Isn't 14 fps really really really low? Can you play like this or you just maxed out everything to test it?

[Dead3yez 0]

Running the ArmA2 Benchmark tests won't really test your GPU. It is the CPU on these tests which are most likely the bottleneck. Hence the ~2more FPS from 14.

To test your GPU and settings for performance you need to be using No AI and/or scripts.

[Azza FHI 50]

Isn't 14 fps really really really low? Can you play like this or you just maxed out everything to test it?

bench 2 has a max of 25 average for some reason, not a very good benchmark. # 1 is more realistic in the sense of what a normal battle would be, i reckon if hes gettin 14 in #2 then he would prolly be getting high 30's to low 40's in bench 1

[forteh 11]

How much does Arma 2 benefit from a having quad-core over a dual-core? I've been trying tweaking various things recently and my frame rate seems similar with cpucount=2 or 4 :confused:

Ive upgraded to an i3 530 @4ghz and its doubled my fps, its a multithreaded dual core and Im pretty sure arma runs on all 4cores (2 logical, 2 virtual); from reading on here quad cores are largely redundant in A2.

Isn't 14 fps really really really low? Can you play like this or you just maxed out everything to test it?

Thats 14fps in a benchmark test with alot of AI and lighting affects (its a flyover of a nightime battle), its not really indicative of real gameplay. I get 11-14fps in bench2 (depending on mods installed) and 45-50 in bench1.

In my experience, arma2 isnt really dependant on super fast framerates in order to be playable, anything above 25fps feels buttery smooth on my machine. On my old system I would get between 15 and 25fps on the single player campaign - it was just about playable, fortunately the fps dips occured when there was no enemy :D

[SeaVee 10]

I have no idea where could you read this. ArmA 2 definitely IS multithreaded. As for driver threading settings, it is quite possible (and we have already seen it while testing) that if the drivers try to perform their own threading, they interfere with the game threads (CPUs being overused by too many threads).

My mistake - I meant to say I've read that Arma2 does not apparently take advantage of or has potential conflicts with Intel's Hyperthreading technology.

Edited March 30, 2010 by SeaVee

[forteh 11]

It seems to make use of it on my hyperthreading dual core i3 530, I know some of the i7s had problems with HT.