Does ARMA II support SLI?

[bangtail 0]

That's weird, the only thing I can suggest is to play around with the detail settings until you get it where you want it and upping your OC at some point.

[Louisville15 10]

That's weird, the only thing I can suggest is to play around with the detail settings until you get it where you want it and upping your OC at some point.

I wonder what is causing it though? Even on that GPU usage pic I showed, the % used is very close for all 3 cards (matching peaks and valleys), and it never goes above 70%, even on GPU1. It's like the game is spreading the load evenly to all 3 cards, yet only giving the performance that would be seen out of one card.

---------- Post added at 03:32 AM ---------- Previous post was at 03:27 AM ----------

I found this link while messing around on google, it appears he has the same problem as me but with an ATI card:

http://forums.amd.com/game/messageview.cfm?catid=260&threadid=125624

[bangtail 0]

CF is known to be problematic with A2. I've never had a problem with SLI although I've never used 3 (or more) Fermi based cards for ArmA 2 so I can't speak from experience there.

My average GPU usage across both cards moves around but it rarely pins 99% and averages about 60-75%.

Edited April 29, 2011 by BangTail

[Louisville15 10]

CF is known to be problematic with A2. I've never had a problem with SLI although I've never used 3 (or more) Fermi based cards for ArmA 2 so I can't speak from experience there.

My average GPU usage across both cards moves around but it rarely pins 99% and averages about 60-75%.

What cards do you have? And what is your average FPS in Benchmark 01 with my settings?

[bangtail 0]

I have dual 580s and I run at 2560 x 1600, all maxed, VD 3500 with no Post (as I dont like my game looking like Fable ;) )

I tested last night while we were talking and got 47 FPS.

[Louisville15 10]

What if you disable SLI?

[bangtail 0]

What if you disable SLI?

32 with SLI off, just ran it now.

[Louisville15 10]

We are talking about the same benchmark, right? The one with the lumberjack in the forest?

[Leon86 13]

I have an idea to test sli, pop a couple of smoke grenades and zoom in on it. That'll get your gpu use up.

[Louisville15 10]

This is really frustrating. Here is a pic of the GPU usage with SLI disabled (compare it to the earlier picture) while running the benchmark. I got 36fps.

---------- Post added at 01:28 PM ---------- Previous post was at 01:25 PM ----------

Here's another thread where a guy tested it with SLI 580s and look at how his charts compare to mine:

http://www.evga.com/forums/tm.aspx?high=&m=986122&mpage=1#986698

His look very similar when in SLI or with SLI disabled, while mine look similar to his with SLI disabled, yet look much different with SLI enabled.

---------- Post added at 01:37 PM ---------- Previous post was at 01:28 PM ----------

I have dual 580s and I run at 2560 x 1600, all maxed, VD 3500 with no Post (as I dont like my game looking like Fable ;) )

I tested last night while we were talking and got 47 FPS.

Is there any way that you could drop your resolution to mine temporarily and put everything on Very High and test again? Visibility at the default. I just want something to compare to.

[Leon86 13]

It's all down to settings, in most cases arma is pretty much cpu limited but with the right settings you can stress the gpu's as well.

It's nice to have some gpu power in reserve for when you pop smoke grenades.

but how about trying with low terrain and low model detail, 2000 viewdistance shadows very high, post very high, aa very high, maybe beyond 100% 3d res (with my dual 260's I run at 115% 3d res with low aa, a personal preference) vsync off.

[Louisville15 10]

It's all down to settings, in most cases arma is pretty much cpu limited but with the right settings you can stress the gpu's as well.

It's nice to have some gpu power in reserve for when you pop smoke grenades.

but how about trying with low terrain and low model detail, 2000 viewdistance shadows very high, post very high, aa very high, maybe beyond 100% 3d res (with my dual 260's I run at 115% 3d res with low aa, a personal preference) vsync off.

Run the benchmark with those settings?

[Leon86 13]

that'll probably stress the gpu's more. Benchmark 1 is a nice "overall" benchmark, stressing both the cpu and gpu. Running at these settings loads the cpu less and the gpu's more, increasing the difference between sli and nonsli.

edit: and if you want to see a greater difference, fire up the editor on utes, play as a team leader, drop a couple of smoke grenades and watch the fps by running fraps in the background.

[Louisville15 10]

I just ran Benchmark 02 with Tri-SLI and got 16fps... this is awful.

Here is a pic of my CPU usage, it looks a bit odd:

[bangtail 0]

@ Louis

Yeah, we are talking about the same benchmark. I've run it at 19x12 before and the differences are about the same.

Do you have EIST etc enabled? I seem to remember that causing a problem with Arma 2 and another user.

Also, just for the sake of being thourough, change your Power management setting from 'Adaptive' to ' Maximum Performance' in the NVCP ArmA 2 profile.

[Louisville15 10]

that'll probably stress the gpu's more. Benchmark 1 is a nice "overall" benchmark, stressing both the cpu and gpu. Running at these settings loads the cpu less and the gpu's more, increasing the difference between sli and nonsli.

edit: and if you want to see a greater difference, fire up the editor on utes, play as a team leader, drop a couple of smoke grenades and watch the fps by running fraps in the background.

Ok, I just switched the settings Terrain Detail : Low, ... Objects Detail: Low

3D resolution: 120% , Visibility: 2036

Benchmark 01 result: 61fps avg

Everything else was on Very High.

[bangtail 0]

Object detail is the biggest hit FPS wise between those two settings so I'd experiment with those.

[Louisville15 10]

@ Louis

Yeah, we are talking about the same benchmark. I've run it at 19x12 before and the differences are about the same.

Do you have EIST etc enabled? I seem to remember that causing a problem with Arma 2 and another user.

Also, just for the sake of being thourough, change your Power management setting from 'Adaptive' to ' Maximum Performance' in the NVCP ArmA 2 profile.

I just did that... I'll try again now.

---------- Post added at 02:38 PM ---------- Previous post was at 02:27 PM ----------

No difference, again.

I know messing with the settings will give me better performance, but I'm trying to figure out why SLI is bugged. Even if tweaking settings results in better fps, it doesn't explain why disabling SLI entirely gives me 2 extra fps with identical settings. That alone says that something definitely isn't right.

[bangtail 0]

Yah, something is definitely wrong but it's through elimination that we get to the root cause.

I would try removing the drivers and then use driver sweeper - reinstall and see if that helps. Ideally, you want to make sure that you get rid of any trace of Nvidia's drivers before reinstalling.

---------- Post added at 10:44 ---------- Previous post was at 10:42 ----------

Ok, I just switched the settings Terrain Detail : Low, ... Objects Detail: Low

3D resolution: 120% , Visibility: 2036

Benchmark 01 result: 61fps avg

Everything else was on Very High.

Did you run this with and without SLI enabled or just with it enabled?

[Louisville15 10]

Yah, something is definitely wrong but it's through elimination that we get to the root cause.

I would try removing the drivers and then use driver sweeper - reinstall and see if that helps. Ideally, you want to make sure that you get rid of any trace of Nvidia's drivers before reinstalling.

---------- Post added at 10:44 ---------- Previous post was at 10:42 ----------

Did you run this with and without SLI enabled or just with it enabled?

Just with it enabled. I'll disable and run it again.

---------- Post added at 02:52 PM ---------- Previous post was at 02:45 PM ----------

Ok, with SLI disabled that time I got 41fps with the same settings (2080ish visibility [couldn't select 2036 again, lol], low object and terrain detail, everything else very high, 120% 3D).

[Leon86 13]

see, sli works :)

[Louisville15 10]

see, sli works :)

lol. it is a step in the right direction!

However, does this mean that my CPU is the bottleneck when I crank the settings back up? What else would explain the lack of difference between 1/2/3 GPUs?

Further, if that is the case, why did I only gain 3fps when bumping the CPU from 2.9 to 3.3? If I was that starved for CPU power, shouldn't a slight OC result in significant frames? For example, the same OC gained me an additional 8fps in the Just Cause 2 benchmark (concrete jungle), but the game itself already runs at absurdly high fps.

[-DirTyDeeDs--Ziggy- 0]

The game isn't good at utilizing multiple cards.

this blanket statement is totally FALSE, Celery ! Please, dont make comments like this without experience with single and multiple cards, AMD and Nvidia.

SLI performance depends on the specific hardware, in my case I have two gtx 275s in SLI and it scales near 100% at 1600x900 (22")

that means I get twice the performance, which isn't shabby to begin with.

this is the most current benchmark I documented, but its a bit out of date using Arma2OAv1.57 :

http://forums.bistudio.com/showpost.php?p=1857288&postcount=6615

[bangtail 0]

lol. it is a step in the right direction!

However, does this mean that my CPU is the bottleneck when I crank the settings back up? What else would explain the lack of difference between 1/2/3 GPUs?

Further, if that is the case, why did I only gain 3fps when bumping the CPU from 2.9 to 3.3? If I was that starved for CPU power, shouldn't a slight OC result in significant frames? For example, the same OC gained me an additional 8fps in the Just Cause 2 benchmark (concrete jungle), but the game itself already runs at absurdly high fps.

Yep, it's what I had thought originally, your OC is not enough to feed those cards. I'd try to get it into 4+ territory and see what happens.

[Louisville15 10]

Yep, it's what I had thought originally, your OC is not enough to feed those cards. I'd try to get it into 4+ territory and see what happens.

Will pushing it to 3.5 real quick by an indicator of anything? I'm wary of pushing it to the 4ghz range on air, but I should be getting a WC setup in a few months or so. I wouldn't expect too much though, as 2.9->3.3 only gave me 3fps.