Arma 3 Benchmarked: Graphics and CPU Performance Tested

By on September 17, 2013, 4:00 AM

Built with Real Virtuality 4, ARMA 3 builds on its predecessors' superb graphics and realism with features including an enhanced mission editor, DirectX 10 and 11 support, improved physics across the board, underwater environments, volumetric clouds, better lighting and a 20km view distance with photo-realistic terrain.

ARMA 3 will offer the largest official terrain of its franchise, with ground area covering approximately 270 km² across the Aegean islands of Altis and 20 km² on the Greek island Stratis. Between its expansive world and graphical advancements, it's no surprise that the developer's recommended specifications are set relatively high.

Read the complete article.




User Comments: 42

Got something to say? Post a comment
jeffz6 said:

I think you need to understand this game to enjoy it. Those that come from bf/cod straight to arma will most likley hate it.

The draw distance is the killer, other options give minial performace changes, but draw distance, wow.

The auto graphics function dosnt work for me? frame rate is to low for my taste, but good thing we can tinker.

2 people like this | TekGun TekGun said:

Great review, really appreciate the effort put in to compare all the different hardware. I'm quite surprised at the Sli/CF scaling, will definitely look at getting a second 7970Ghz.

This game should be the default benchmark for all gaming hardware. There is no game that compares, just have a look at the detail on everything from tree bark to roof tiles, weapons and uniforms. Not to mention the cockpits of all the vehicles, with the PIP and rear view mirrors that actually work in real time. The game is immense, it's no wonder it requires the hardware it does to run at the higher video settings. As for Altis, just beutiful.

misor misor said:

270 square kilometers to play with (on aegian islands)?. that's huge. I don't have a decent mainstream gpu yet but I'd be smiling ear to ear when NVidia gtx 760 will be locally available at a decent price and be able to test that map.

aboynamedmatt aboynamedmatt said:

Really surprised in the lack of performance by the 7970GHz. Is it finally starting to show it's age?

GhostRyder GhostRyder said:

Really surprised in the lack of performance by the 7970GHz. Is it finally starting to show it's age?

It probably has to do with immature drivers, this tends to happen in a new game though even now the drivers showed to be pretty decent at launch based off the different FPS levels. Games still new I would give it some time, really not many of the cards were able to handle the game too well and im sure on both sides of the table there will improvements over time. It seems that unless your running beyond 1080p, the need for 3gb or higher on the GPU is unnecessary as even the titan did not do well. This is straight up, good ol fashioned GPU power and drivers.

I have to say though, this is a well optimized/programmed game for CPU's and GPU's. The Fact that the CPU's from both sides keep up very well, the game can run on even lower end CPU's, and it scales to each CPU when clocking them differently shows that the developers worked real hard to make sure this game was playable. Im happy to see some good CFX/SLI profiles as well because that means we can push this game up to ultra with some ease with one of those setups.

Great review, glad to see such a wide margin of cards

Lionvibez said:

Really surprised in the lack of performance by the 7970GHz. Is it finally starting to show it's age?

I don't think its age at all.

As Ryder posted its mostly drivers and just ARMA.

This game is kicking Titan's ass also every game in this series has been this gpu demanding when you crank the settings.

Blue Falcon said:

"For the money, a pair of GTX 770s would be hard to beat if you want to crank up the settings without thinking twice, particularly since they performed better than two HD 7970 GHz Edition cards for around the same price or less."

OR

You can get 90% of the performance of $800 770 2GB SLI for $560 by getting 1Ghz 7970s in CF:

[link]

Hard to understand some of those benchmarks like 680 beating 670 by 26% on average in some of those charts.

Guest said:

At $60 a copy, it better be a good game. I put it on my wish list.

Guest said:

The graphics look incredibly realistic! Too bad my mobo is severely outdated to run anything near that.

Guest said:

I'll try it on my GTX Titans.

Guest said:

You missed making benchmarks on multiplayer, where you will get 1/2 or 1/3 performance on servers with 20 or more people. The game is made basically for multiplayer, singleplayer performance does not represent that reality. You guys should update the review.

GhostRyder GhostRyder said:

"For the money, a pair of GTX 770s would be hard to beat if you want to crank up the settings without thinking twice, particularly since they performed better than two HD 7970 GHz Edition cards for around the same price or less."

OR

You can get 90% of the performance of $800 770 2GB SLI for $560 by getting 1Ghz 7970s in CF:

[link]

Hard to understand some of those benchmarks like 680 beating 670 by 26% on average in some of those charts.

Good points there

You missed making benchmarks on multiplayer, where you will get 1/2 or 1/3 performance on servers with 20 or more people. The game is made basically for multiplayer, singleplayer performance does not represent that reality. You guys should update the review.

Doing a multiplayer test and being accurate for all scenarios or being able to repeat scenarios is near impossible. You cannot predict or repeat the exact position, situations, or timing of scertain scenarios to give an a realistic example of what to expect in multiplayer. One card may get all these extremely hard to render events happen while another could get the easy street which would alter the results.

JC713 JC713 said:

I play BF3 at 40FPS and I deal with it. It isnt that bad. Yes, 60FPS is nice, but 30 and above is playable.

The differences between high and ultra are pretty minimal. I would rather gain the FPS than the quality.

Guest said:

Making an irrelevant review is worse than doing none at all. the other 'guest' is right to complain and the point that different hardware behaves differently is equally valid in single player.

THE PROBLEM IS:

Arma 3's performance in multiplayer is SIGNIFICANTLY worse than in singleplayer and many people claim there is a severe cpu bottle neck and some also point to the net code being 'bad' (server bandwidth and cpu performance matter too much among other things, apparently).

Therefore this whole review is 'nice' but for anyone who wants to play mostly MP it has absolutely no value.

Pro tip: there are enough well coded missions to test MP performance(CTI, FT-2, and maybe wasteland for its popularity) at least to the extent of determining whether cpu or gpu matters more or if it is even possible to play the average MP mission at 60+ fps on any hardware.

Staff
Julio Franco Julio Franco, TechSpot Editor, said:

Browsing through the ARMA forums I noticed a lot of people complaining about multiplayer performance and upcoming fixes -- meaning it's not the desired performance level, but an apparent lack of optimization.

When a new high profile game is released, we have to decide if the game is worthy of testing or not (if the graphics challenge current PC hardware, if the game is popular, etc.) In this case the decision came to:

1) Showcasing engine performance at launch with what potentially is going to be the performance you'll eventually see everywhere in the game (single and multiplayer) OR 2) Waiting a couple of weeks or months for the developer to patch things up and then test the game.

We picked #1.

INSTG8R INSTG8R said:

It would have been nice to have at least mentioned which drivers you were using for the tests. I personally have a i7 2600@4.6 and a 7970 OC: I get decent 40-60 FPS on Very High with some other settings turned down a bit but that is on the 13.6 B2's because having tried the 10.8 Betas I had absolutely horrendous FPS in some of my other games like WoT and ETS 2 so I haven't even tried the the latest 13.10 Betas because of this.

Playing this game and looking for articles and benchmarks like this using the 13.10 betas might make me give them a shot if it I saw some performance gains because of it.

I mean it was a great article and the results seem to be pretty on par but not giving an accurate test setup means they can easily be taken with a grain of salt.

Staff
Julio Franco Julio Franco, TechSpot Editor, said:

It would have been nice to have at least mentioned which drivers you were using for the tests.

Full system specs are at the bottom of the first page, as usual.

INSTG8R INSTG8R said:

Geez I apologise! I glossed right over that.. Had been playing Occupation in ARMA 3 for the last few hours and a bit out of it.

Thanks for setting me straight I may just give the 13.10s a shot tho I am pretty much getting the same results.

Guest said:

From StefenHeif:

Well, seems that ARMA 3 is hard on AMD GPU's because of the use of nVidia's PhysX.

http://en.wikipedia.org/wiki/Arma_3#Development

So I believe that if you use a lower particle setting, this should improve by a good lot.

Also, in MP, reducing particles with an AMD GPU will lower strain on CPU as well

Feel free to experiment on your own.

GhostRyder GhostRyder said:

Geez I apologise! I glossed right over that.. Had been playing Occupation in ARMA 3 for the last few hours and a bit out of it.

Thanks for setting me straight I may just give the 13.10s a shot tho I am pretty much getting the same results.

I have tried both 13.8s and 13.10 and im running Dual HD 6990's, so far ive only had positive results for the most part that ive noticed. Only game that really did not change for me was BF3 at least.

TekGun TekGun said:

From StefenHeif:

Well, seems that ARMA 3 is hard on AMD GPU's because of the use of nVidia's PhysX.

http://en.wikipedia.org/wiki/Arma_3#Development

.

Actually the PhysX is run only on the CPU no matter which type of GPU you are running.

As for the multiplayer performance, most of the problems seem to be badly created maps and players thinking they can host 20+ players and play on the same machine. A decent dedicated server is needed.

j05hh j05hh said:

Being a BF3 and COD player, this was my first time playing the ARMA series... boy was I bored!!! Thank goodness for not purchasing this game.... it was quickly uninstalled.

GhostRyder GhostRyder said:

Being a BF3 and COD player, this was my first time playing the ARMA series... boy was I bored!!! Thank goodness for not purchasing this game.... it was quickly uninstalled.

Personally, I cant understand how you can call CoD interesting and Arma Boring, but then again everyone is different and likes different things though Battlefield is definitely one of the best Shooters out there.

Actually the PhysX is run only on the CPU no matter which type of GPU you are running.

As for the multiplayer performance, most of the problems seem to be badly created maps and players thinking they can host 20+ players and play on the same machine. A decent dedicated server is needed.

That's not true, the PhysX on an Nvidia GPU system will automatically if you don't play with the system default to the GPU's PhysX processor. You can of course force it to the CPU, but what would be the point if you have an Nvidia GPU. On an AMD system, if you install the PhysX, it will run on the CPU unless you do a hybrid PhysX system which is a little bit difficult to setup.

Advanced PhysX requires and Nvidia GPU to run, but if there is not one it will be off and just running normal PhysX on the CPU which will give you most of the benefits so long as your CPU can keep up (Almost any CPU from recent Generations on both sides can handle regular PhysX pretty well).

TekGun TekGun said:

No that's not right, what you are talking about with the GPU acceleration is an optional part of PhysX called "Apex", this enhances special effects etc. The part of PhysX that deals with collisions and other calculations is done on the CPU.

ArmA 3 does not include the use of Apex PhysX and so it is impossible to accelerate it on any GPU.

St1ckM4n St1ckM4n said:

No that's not right, what you are talking about with the GPU acceleration is an optional part of PhysX called "Apex", this enhances special effects etc. The part of PhysX that deals with collisions and other calculations is done on the CPU.

So what is fluid dynamics and 'cloth' effects? These are collisions, are they not?

TekGun TekGun said:

By collisions I meant the type you get when a vehicle crashes etc. From what I understand rigid body physics calculations ie a car hitting a wall are all done on the CPU. Fluids, cloth and particle physics are the kind of effects that can be accelerated by a GPU, however these enhancements are not included in ArmA 3, and this is the reason why PhysX in A3 cannot be accelerated by gpu.

List of GPU accelerated effects.

Blog post on PhysX

GhostRyder GhostRyder said:

Interesting, the game does say PhysX enabled so I would assume that it means it will use the GPU PhysX as well. If a game is PhysX enabled and advertises it, they are using the GPU PhysX from at least what ive seen otherwise they don't really mention it. I was under the impression this game also used PhysX acceleration for effects and such, but I could be mistaken as I have not been following this game.

However, I would like to see some more information on this theory for this game as I had understood the PhysX was actually being used on the GPU if available on Arma 3.

2 people like this | TekGun TekGun said:

ArmA 3 is not the only game to advertise the fact it uses PhysX, but does not include GPU acceleration.

Take Bioshock Infinite, on the back of the box it has both the NVidia PhysX logo and the Gaming Evolved AMD logo... which is odd no? this is because they have used the PhysX sdk for rigid body physics, done solely on the CPU. So basically just because a game has the PhysX logo, it does not always mean that GPU acceleration is possible.

This was also news to me, until Bohemia Interactive announced they would be using PhysX and I started looking into it. I just presumed any title with the PhysX logo would have the ability to be GPU accelerated but this is not always the case.

Here's a list of games using the PhysX SDK and which ones actually have GPU support.

4 people like this | Borivoj Hlava Borivoj Hlava said:

Interesting, the game does say PhysX enabled so I would assume that it means it will use the GPU PhysX as well. If a game is PhysX enabled and advertises it, they are using the GPU PhysX from at least what ive seen otherwise they don't really mention it. I was under the impression this game also used PhysX acceleration for effects and such, but I could be mistaken as I have not been following this game.

However, I would like to see some more information on this theory for this game as I had understood the PhysX was actually being used on the GPU if available on Arma 3.

TekGun is right. I am working on Arma 3 and I can confirm that all PhysX calculations in Arma 3 are done on CPU. We are working on PhysX acceleration for effects, but it is not in the game yet. Current particle effects don't use PhysX.

TekGun TekGun said:

Thanks for the confirmation, wow PhysX acceleration coming to ArmA3 that's some big news...

Guest said:

Placing a high score on this game based on the singleplayer when it doesnÂ't even have a single player campaign is criminal, it makes people assume they will have decent performance on multiplayer which they wont. Please visit the official game forum and the steam discussion, everyday you will see several topics being created about the terrible unplayable multiplayer performance (less than 20fps with i7 4770 and titans) on servers populated by as low as 20 people, when the official webpage says people can play it with 60 VS 60 players. Also, not only they actively delete most complains but they have a few people (always the same 3) working towards trying to discredit anyone who do.

Reviews on singleplayer are a disservice to the gaming community and excuses the broken product they launched. Also, they actively delete most complains.

Guest said:

Re: PhysX - I think you have to view the logo on the box and the use of PhysX as just as you would view the "Havoc" physics logo. It's essentially a pre-packaged plugin they code into their engine. People get confused and assume it's GPU accelerated automatically because it says Nvidia. At one point, they were trying to bring out standalone PhysX cards, iirc, but now it can be CPU based (or else, no-one would use it in their engines in lieu of a more universal plugin).

As for the game itself. I disagree with some of the testing methods for this game. For one, it looks like you're testing on Stratis (all of the showcase missions). The majority of players having performance issues are with the Altis map. Which was released to the dev-branch a few weeks before the official launch, so it's horribly unoptimized (and huge). I was running SP Stratis comfortably with very high settings and a decent mission with around 40+ FPS. On Altis, that same setup was pushing 19-24 FPS.

You failed to minimize variances in your testing methodology. A) You shouldn't turn off AI, as you're most likely lessening the load on the CPU (what are they doing? nothing, there's not a lot of calculations if they don't have to react). B) You should have used a preprogrammed mission benchmark. There's a few out there, but you could also have someone set something up for you. The showcases aren't nearly as complex as a real mission would be.

GhostRyder GhostRyder said:

TekGun is right. I am working on Arma 3 and I can confirm that all PhysX calculations in Arma 3 are done on CPU. We are working on PhysX acceleration for effects, but it is not in the game yet. Current particle effects don't use PhysX.

Interesting, ok well that solves that issue then doesn't it.

NeurotechHD NeurotechHD said:

I don't quite understand this performance article. They clearly showed that the game was CPU intensive (40% increase in speed equaled 40% increase in performance) - then went on to say that no, it's GPU-intensive. I didn't see a lot of evidence that ARMA is a GPU-intensive game. There was a 30% increase in performance if you buy the card that is 30% more powerful (and a lot more expensive).

"Before starting the test, we plan to turn the difficulty down to the easiest level ("Recruit" and then further handicap the AI by setting its skill level to 0."

Without the AI there is no point to the game. ARMA is a simulator - if you can't simulate then what's the point? It's not an Environment/ Horizon Simulator - the purpose of the game is to win a fight.

.

That's like "performance benchmarking" a chess game where they disable the pieces and wander around the chess board. What's the point? Teaching us which graphics cards can play a broken/ disabled game the fastest doesn't mean anything to anyone in any meaningful way, not really.

.

Know what I mean?

NeurotechHD NeurotechHD said:

I don't quite understand this performance article. They clearly showed that the game was CPU intensive (40% increase in speed equaled 40% increase in performance) - then went on to say that no, it's GPU-intensive. I didn't see a lot of evidence that ARMA is a GPU-intensive game. There was a 30% increase in performance if you buy the card that is 30% more powerful (and a lot more expensive).

"Before starting the test, we plan to turn the difficulty down to the easiest level and then further handicap the AI by setting its skill level to 0."

Without the AI there is no point to the game. ARMA is a simulator - if you can't simulate then what's the point? It's not an Environment/ Horizon Simulator - the purpose of the game is to win a fight.

.

That's like "performance benchmarking" a chess game where they disable the pieces and wander around the chess board. What's the point? Teaching us which graphics cards can play a broken/ disabled game the fastest doesn't mean anything to anyone in any meaningful way, not really.

.

Know what I mean?

Was my comment cut off? I'm a little new to this commenting system.

mailpup mailpup said:

What exactly was cut off?

NeurotechHD NeurotechHD said:

What exactly was cut off?

It seems to have resolved itself - when I had posted the comment initially it ended abruptly just after "("Recruit" " in the sentence.

GhostRyder GhostRyder said:

I don't quite understand this performance article. They clearly showed that the game was CPU intensive (40% increase in speed equaled 40% increase in performance) - then went on to say that no, it's GPU-intensive. I didn't see a lot of evidence that ARMA is a GPU-intensive game. There was a 30% increase in performance if you buy the card that is 30% more powerful (and a lot more expensive).

"Before starting the test, we plan to turn the difficulty down to the easiest level ("Recruit" and then further handicap the AI by setting its skill level to 0."

Without the AI there is no point to the game. ARMA is a simulator - if you can't simulate then what's the point? It's not an Environment/ Horizon Simulator - the purpose of the game is to win a fight.

.

That's like "performance benchmarking" a chess game where they disable the pieces and wander around the chess board. What's the point? Teaching us which graphics cards can play a broken/ disabled game the fastest doesn't mean anything to anyone in any meaningful way, not really.

.

Know what I mean?

The ARMA series does scale with CPU's as was shown with the overclocking section of this game. You can overclocked the CPU more and more to up the performance, but that does not define it eas a CPU intensive game, as stock speeds for the CPU's was enough to do well in benchmarks (Meaning overclocking will help get the max, but it was not required of the game). The GPU section showed you needed some pretty powerful GPU's to put the game on ultra (Not as much as a game like Metro 2033, but it needs some umph to play).

The fact the game scaled with overclocking CPU's and different GPU setups shows that the game was just programmed well for all types of setups to help people play.

As for the AI part, they turned it off to make it easier to repeat scenarios. When benchmarking a game, you have to make each scenario as similar as possible to make each test more legitimate and fair. Would it be fair to test the GTX 760 with no guys on the screen moving and record results while testing an GTX 780 with 50 guys running around in front of you? That would change the results and might show the 760 as being better just because the FPS recorded in the situation was easier to obtain without much stuff on screen vs needing lots of power with tons of AI on screen.

ziiip ziiip said:

Theres a nasty bug that may or may not happen to you and it causes to FPS to be worse than what it should it be unless you alt-tab after launching the game or changing the graphics to a preset setting.

Btw, mouse smoothing also makes the game feel unresponsive, did you have it turned off for the test? Just because you said the game felt bad under 40 FPS.

Guest said:

Judging by the CPU performance charts; I see an Intel I7-4770k gets 53 fps at 3.5Ghz and/or 4.0Ghz. So which is it? So is a 4770k slower than an I5? What am I missing? Is the data fictional? Biased?

Guest said:

Did not read all posts so may have been asked...WHY no SLI TITAN benchmarks???

St1ckM4n St1ckM4n said:

There are no SLI TITAN benchmarks because we know it will increase the community participation of Guests'.

Load all comments...

Add New Comment

TechSpot Members
Login or sign up for free,
it takes about 30 seconds.
You may also...
Get complete access to the TechSpot community. Join thousands of technology enthusiasts that contribute and share knowledge in our forum. Get a private inbox, upload your own photo gallery and more.