SLI yes ore no :(

[loopdk 92]

Hey ...

let me start by saying that I have searched the forum and have not found nothing that is newer 200 days ....

I just today received a 780TI MSI Gaming more. SO I now have 2 cards of this model.

I ran the test with this

http://forums.bistudio.com/showthread.php?151794-ArmA3Mark-Benchmark-your-ArmA-3

Altis edition

THis is my settings:

http://peecee.dk/uploads/022015/Display1.jpg (163 kB)

http://peecee.dk/uploads/022015/AA1.jpg (165 kB)

http://peecee.dk/uploads/022015/Video_Options.jpg (170 kB)

And my system is:

Windows 7 Ultimate 64 bit

Mobo:

ASUS rog hero

Chipset:

z97

CPU:

i7 4970k (runs 5000 )

GPU:

2x 780TI

RAM:

Kingston 16gb 2666 MHZ

Hard drive:

Intel 335 SSD 180GB/ Samsung evo 240Gb/ 3Tb greenline

PSU:

Corsair TX850 V2

Soundcard:

ARGON DA2V2

Speakers:

KRK RP5 G3/ KRK SUB

Monitors:

3x2411T Benqt

With 1x 780TI i get:

and with 2 i get only 2 more.....

Supports ARMA 3 non SLI? or do I do something wrong ??

BTW I have the latest NVIDIA DRIVER

and I have read of people in here who have more FPS on ULTRA with less short ...

Help: D

Thanks in advance.

[forteh 11]

At 1080p I think the single 780 is plenty powerful enough which is why you're only getting minor increases. Try increasing the sampling to 150% and see if there is still only a small difference between the two.

I believe that sli in arma only starts to become useful when running multiple monitors, higher resolution or higher sampling :)

[loopdk 92]

i will try doing the test in 5760x1080....

[dayglow 2]

Arma is very much limited by CPU setting up draw calls. I run 290x crossfire at 5670x1080. Crossfire works in the sense that I can run at 120% render with all settings at Ultra except object detail and terrain set to High.

basically view distance, object and terrain detail have the biggest effect on fps as they effect CPU rendering speed the most. SLI will let you max out the other GPU related settings with no impact on FPS.

[loopdk 92]

so it is worthless :( MUST HA MORE FPS

---------- Post added at 18:29 ---------- Previous post was at 18:29 ----------

i will try doing the test in 5760x1080....

1x780TI gives 33 FPS

2x780TI gives 44 FPS

In 5760x1080 in same settings as above

[dayglow 2]

Not worthless, you can crank up the visual fidelity without a loss of fps

[St. Jimmy 272]

Well you could increase your resolution and sampling likely without any drop. Also you can drop Object view distance to 1700m at least unless you're doing something where you need to see that far any enemies/objects.

View/object distances are the biggest bottleneck. I run usually 3000m/1700m or less unless there's some real need for more. Object Detail can also free some fps.

One 780Ti is likely just well enough for Arma 3 @ 1080p. But only thing to see that is do you ever hit 100% GPU usage with only one 780Ti? Everytime you hit full GPU usage you can increase your fps with better GPU or lowering GPU heavy settings but all the time you don't see full usage you have just room to breath.

Example what you experience in theory:

With 1 780Ti you can run @ 1080p with those settings and if you would try 4K your fps drops because GPU is bottlenecking. When you SLI you could go for 4K with the same fps but your fps @ 1080p doesn't increase.

[loopdk 92]

But only thing to see that is do you ever hit 100% GPU usage with only one 780Ti

With 1 card i use 100% i game :/

[spamurai 3]

I am suspicious of your result as you effectively seeing what amounts to no gain at all.

Please don't interpret this the wrong way,

But you did link the cards together with the SLI cable and then enabled SLI in the Nvidia control panel?

I only ask because your benchmark appears as though only 1 card is actually doing the work.

You should see an improvement with SLI.

My SLI GTX670 was a marked improvement over a single GTX670

[loopdk 92]

Yes as you can se... With heaven benchmark i get 82% more in test with sli enablet

[I give up 152]

I am going to give you a tip.

Check your cpu usage in editor, then check it in game (Altis, preferably).

You can have even a Quad SLI, in game it has no use. It will be capped by cpu usage, exactly because the game do not allow (or it cant) extract the "juice" from cpus.

[loopdk 92]

I am runing 5000 on my i5... Using max60% of it.

Card1 is using max 33% and 29% on card2

Daamn it

[kklownboy 43]

make sure you have Core parking off

[loopdk 92]

I have.

[kklownboy 43]

I wouldnt be to wacked about the % of usage. If you hit your Vsync, your good. Is your OC on all four cores? The Game is going to use Core 0 or 1, 80-90% of the time when you push the Simulation(AI etc). Your GPU(s) can use max fillters, AA etc. And 3D settings in the Nvidia CP can be used with SLI better than a single GPU.

[loopdk 92]

Bf3 is using 67% of both cards

[St. Jimmy 272]

Bf3 is using 67% of both cards

Because CPU is less bottlenecked there so GPU can work more. Simple.

[loopdk 92]

So i just accecpt i only can use 2*35% of my gpu's

[St. Jimmy 272]

So i just accecpt i only can use 2*35% of my gpu's

Yes or lower view and object distances and detail to get more fps and naturally the usage also rises bit more on that benchmark.

[galzohar 31]

Ignore % usage, it is almost completely meaningless if you don't have access to the internals of the game code.

With SLI you shouldn't try to see if you get more FPS, but instead try see if you can increase the settings more without losing FPS. That is, find the max settings you can use without losing FPS, then try do the same with SLI.

If you just look at your FPS with the same settings you might not see an increase due to CPU.

Another way to greatly reduce the effect of CPU is to make an empty single player mission and test FPS on that - That way you have very little CPU bottle-necking and your FPS will depend almost entirely on your GPU/SLI power.

[dayglow 2]

you can increase most of your visual options to ultra with no loss in fps. Set the render scale to 120%-133% for better edge AA. I find 120% with SSMA Ultra smooths pretty much everything.

Arma manis a large game with the ability to run a ton of units all with independent AI, which is very CPU intensive.

also just try to enjoy the game. You aren't playing a fps/usage counter. I enjoy games a lot more when I turn off the fps counter then adjust settings based on perception.

[loopdk 92]

So I'm back after lots of testing: D

I have nvidia inspector changed slightly and it has brought me up to 52% with both cards: D

I have at the same time raised my settings by 19% without losing fps. So it is certainly the right way ...

BUT so today I should just testBcombat on our server and damn since !!!!

I made a very simple little mission on bystrica .. a patrol coming up the road .. just wanted to see if @Bcombat worked and how the throwing smoke etc ..

My settings are something about to everything that can is ULTRA and my view distance is 5500 ..

On this map, I use 94% of each card = 133FPS: D

I realize that there was so much to stress the CPU and gpuerne etc. but anyway ... I was using 94% of both cards ...

I went back to BORNHOLM and bang I wnt down to approximately 50% of each card again :/

Some can explain this ??

[KrazyBee 8]

what is nvidia inspector and wat does it do?is it some kind of setting like nvidia cp??if so can u give me ur settings so i can copy it..im running 2 770 2gb and no luck so far thanks

[St. Jimmy 272]

I went back to BORNHOLM and bang I wnt down to approximately 50% of each card again :/

Some can explain this ??

Maybe more objects everywhere that stress CPU more. Also Bystrica is much smaller map. 5500m VD is likely overkill for Bornholm.

[loopdk 92]

what is nvidia inspector and wat does it do?is it some kind of setting like nvidia cp??if so can u give me ur settings so i can copy it..im running 2 770 2gb and no luck so far thanks

Downloadet it here

http://www.guru3d.com/files-details/nvidia-inspector-download.html

And download my settings here

http://loopdk.synology.me:5000/fbsharing/qrMgIfid

Plz share if yoy get it better :D

---------- Post added at 14:17 ---------- Previous post was at 14:16 ----------

Maybe more objects everywhere that stress CPU more. Also Bystrica is much smaller map. 5500m VD is likely overkill for Bornholm.

On both maps i only use 55% of my CPU :/