Solution for low FPS in 1.14

[shoaibz 0]

I found this on the SimHQ forums, I used it myself, and noticed a remarkable change, maybe it can work for you too.

Basically just delete two files in your My Documents\ARMA folder which are "AMRA.cfg" and "YourName.ARMAProfile". Start up ARMA which will launch with all settings to default. Setup your settings and see the performance boost. Details are available on SimHQ

[maddogx 13]

I found this on the SimHQ forums, I used it myself, and noticed a remarkable change, maybe it can work for your too.

Basically just delete two files in your My Documents\ARMA folder which are "AMRA.cfg" and "YourName.ARMAProfile". Start up ARMA which will launch with all settings to default. Setup your settings and see the performance boost. Details are available on SimHQ

I just tried it out by backing up my old config files and then deleting them. When you start ArmA, it basically just creates new ones.

The only real difference I can see between them (apart from the fact that all the settings are reset to default) is that the 3D_Performance parameter in the ArmA.cfg is about 10% lower than before. That's all.

I'm guessing they've lowered the value of the game's internal benchmark for 1.14 - I'm not sure if or how exactly that number affects performance, but it's worth lowering it to see what happens. No need to delete your config files.

EDIT:

Just tried playing around with the values a bit. First I set it to just under half of what it was before - result was LOD dropping at very short distance, resulting in low visual quality but overall higher framerates.

Then I tried setting it to 3 times the old value. Result: hardly any visible LOD drops, good visual quality at further distances and also smooother, albeit lower FPS.

Overall I surmise that the 3D_Performance setting has an effect on LOD fall-off. The lower the value, the lower the overall level of detail, but also more switching between LODs. I would recommend playing around with the value, to see what suits your system best.

I'm not on my own computer right now, but I'm definitely gonna try this out at home.

[malick 0]

Interesting find... In OFP, the 3D performance was determined by the launch program and I don't remember having access to it. In ArmA, I thought it was only informative. If it has an influence and we can tweak it : let's tweak it to death !

I believe I have already tweaked/changed every single aspect of this game, from visual quality, sounds, textures, units, islands, effects, performances... That's what you call an open game engine !

Malick

[maddogx 13]

This setting definitely has a large impact. After some further testing I can say one thing: don't set it too high. You'll regret it in forests.

[SaBrE_UK 0]

I thought "model quality" (or whatever it's called) did this in-game anyway; trees will switch to lower LODs at shorter distances on lower settings. Or am I misunderstanding this topic?

[BraTTy 0]

Thanks for testing that MadDogX

I noticed less lod switching after installing 1.14, I think I will leave it alone, lol

[suma 8]

The only real difference I can see between them (apart from the fact that all the settings are reset to default) is that the 3D_Performance parameter in the ArmA.cfg is about 10% lower than before. That's all.

I'm guessing they've lowered the value of the game's internal benchmark for 1.14 - I'm not sure if or how exactly that number affects performance, but it's worth lowering it to see what happens. No need to delete your config files.

...

Just tried playing around with the values a bit. First I set it to just under half of what it was before - result was LOD dropping at very short distance, resulting in low visual quality but overall higher framerates.

Then I tried setting it to 3 times the old value. Result: hardly any visible LOD drops, good visual quality at further distances and also smoother, albeit lower FPS.

Overall I surmise that the 3D_Performance setting has an effect on LOD fall-off. The lower the value, the lower the overall level of detail, but also more switching between LODs. I would recommend playing around with the value, to see what suits your system best.

This is really weird. Reading the above, I inspected the source code to see what this value is used for, and I can see only two things:

1) your Terrain Detail settings are limited by the benchmark value. When you have the 3D_Performance low, you are more limited in selecting higher terrain detail levels when playing with larger view distance.

2) scripting function "benchmark"

I cannot see how any of above could explain what you see.

Moreover, we definitely did not change any "3D_Performance" handling in 1.14 at all compared to previous versions - none of the code handling it was touched since 1.08 (and most likely even before, but I did not check for that now).

I have also tried to reproduce if I can observe any effect when changing the value, and I did not see any.

[maddogx 13]

I'll make a video to show what I mean. The effect is easily visible for me.

EDIT: I created a demo camera fly-over of Paraiso on South Sahrani from ArmA 1.14. Tried with 3D_Benchmark values 1000, 2000, 10000, 20000 and 50000.

There is a definite difference of about 8-9 FPS between the lower and higher benchmark settings, but I can't find the break point. I am also still sure I'm seeing less LOD swapping with higher benchmark settings.

Problem is: as soon as I try it with Fraps, my framerate drops to the 20s and the LOD popping appears to increase. So basically it's the same for all videos I made. Either it is directly related to the FPS, or I am just completely deluding myself.

Though if that is the case, I wonder why other people have been seeing a performance increase? Could it just be a placebo effect?

[cross 1]

I can confirm Suma in limiting settings according to benchmark value.

When I lower the settings my "texture level" is set & limited to normal for example (used to be very high) with the same viewdistance. This is probably the reason to your increasing FPS.

MadDogX, can you confirm your video settings (texture level) do not change when you do you were performing your test.

Also, if the developer says i does not effect the game other than limiting your video level options we better believe him

[gL33k 0]

I'll make a video to show what I mean. The effect is easily visible for me.

EDIT: I created a demo camera fly-over of Paraiso on South Sahrani from ArmA 1.14. Tried with 3D_Benchmark values 1000, 2000, 10000, 20000 and 50000.

There is a definite difference of about 8-9 FPS between the lower and higher benchmark settings, but I can't find the break point. I am also still sure I'm seeing less LOD swapping with higher benchmark settings.have been seeing a performance increase? Could it just be a placebo effect?

definitly yes. the difference is huge ! no more hexagonal wheel at 30 meter , i set it to 30000 , and all mesh are a lot more detailled .

I can still see low poly truck , but it could be something related to truck mesh itself, because i didn't see same for neither static weapon , nor armored vehicle.

it could be a way to go , only if you have a good CPU i presume.

[suma 8]

I am not a type to disregard measurements - measurements are always true. Still, I have no explanation what other influence this setting could have, other than low 3D_Performance causing worse terrain detail, which I would expect to increase fps if you are using some higher view distance. If there is any real impact, I would definitely like to know what it is causing it, because if there is some hidden or unexpected influence, it might be used to tweak default settings to perform better for most users.

Could you perhaps make some experiment log here, and for each experiment to record:

- what was the 3D_Performance value

- what was the fps in some particular scene

- what was the Terrain Detail setting and View distance in Video Options

As for LOD switching, I guess this is somewhat harder to measure objectively - could you perhaps select some particular object, watch it closely and report if there is any reproducible difference?

The only other idea I have is perhaps you might have installed some script (some 3rd party addon?) which checks the benchmark value and takes different paths or performs a different functionality depending on it.

[maddogx 13]

I compared my settings while using benchmarks 10.000 (what I usually use) and then 1.000 as a reference. My settings in the advanced video options didn't change. I have terrain, object detail and textures set to normal. This was still the case with benchmark 1000.

In both cases I had a soldier standing at a specific spot in a wine field on south Sahrani, without moving. FPS while stationary were the same for both benchmark settings. While moving I didn't notice much difference either, but it's hard to tell. I'm starting to think I may be going mad   .

I guess more testing will be needed to get some more objective results, since most of the improvements I noticed before were purely subjective.

[stryker555 0]

I have gained a performance increase in the field on top of the fps I got from the 1.14 patch.

BUT no significant increase in performance in cities or towns, those stay the same.

[gL33k 0]

The only real difference I can see between them (apart from the fact that all the settings are reset to default) is that the 3D_Performance parameter in the ArmA.cfg is about 10% lower than before. That's all.

I'm guessing they've lowered the value of the game's internal benchmark for 1.14 - I'm not sure if or how exactly that number affects performance, but it's worth lowering it to see what happens. No need to delete your config files.

...

Just tried playing around with the values a bit. First I set it to just under half of what it was before - result was LOD dropping at very short distance, resulting in low visual quality but overall higher framerates.

Then I tried setting it to 3 times the old value. Result: hardly any visible LOD drops, good visual quality at further distances and also smoother, albeit lower FPS.

Overall I surmise that the 3D_Performance setting has an effect on LOD fall-off. The lower the value, the lower the overall level of detail, but also more switching between LODs. I would recommend playing around with the value, to see what suits your system best.

This is really weird. Reading the above, I inspected the source code to see what this value is used for, and I can see only two things:

1) your Terrain Detail settings are limited by the benchmark value. When you have the 3D_Performance low, you are more limited in selecting higher terrain detail levels when playing with larger view distance.

2) scripting function "benchmark"

I cannot see how any of above could explain what you see.

Moreover, we definitely did not change any "3D_Performance" handling in 1.14 at all compared to previous versions - none of the code handling it was touched since 1.08 (and most likely even before, but I did not check for that now).

I have also tried to reproduce if I can observe any effect when changing the value, and I did not see any.

so this parameter doesn't affect at all LOD swiching ?

hmm strange so .

i will make screenshot for be sure

ok, then , object detail change lod switching , i can't find any difference between 8000 & 30000. i probably change object settings at same time so ...

[maddogx 13]

Well, after some more testing I must admit defeat. Seems that the improvements I was experiencing were the result of pure chance combined with wishful thinking.

One thing I have learned from this: do not jump to conclusions too quickly, or Suma will hunt you down and destroy your world using nothing but solid facts.

[Yoma 0]

Hang on a second there. I did some testing with ArmaMark

My system is a Intel 6750,8800GTX, 2GB Ram, WinxpSP3,on latest whql NVIDIA drivers.

I did 3 Armamark runs with different 3dperformance settings.

I noticed no difference in LOD/Textureloading.

These are the results

Viewdistance 2251

Texture normal

Terrain very low (heh i don't like grass, it makes me nautious)

3dperformance: 1000

1) 4002

2) 4085

3) 4129

Average: 4072

3dperformance: 10638 (the detected value)

1) 3980

2) 4136

3) 4108

Average: 4075

3dperformance: 30000

1) 3911

2) 4358

3) 4196

Average: 4155

3dperformance: 50000

1) 3897

2) 4239

3) 4153

Average: 4096

After i did the tests i've put the 3dperformance parameter back to 30000 (as this gave the best result, be it a very slight difference) and ran the first 2 tests again, and again it was slightly better then all other used settings.

Allthough the difference in performance is minimal (i notice NO visual difference at all) it does perform a tad differently.

Of course the difference is completely negligable compared to what you get with 8800GTX and High shadows vs normal shadows.

That said ArmaMark is probably completely inaccurate.

But i wanted to post these numbers anyway.

The best thing is i'm most likely gonna install VistaX64 on my system this weekend. So now i have some benchmarks to compare with

[gL33k 0]

Quote[/b] ]One thing I have learned from this: do not jump to conclusions too quickly, or Suma will hunt you down and destroy your world using nothing but solid facts.

we saw !

[SnR 1]

i am still testing but its looking promising

x1950Pro 2mb ram 2.8 dual core Xp