HardOCP does some extensive ArmA 3 GPU testing

[bangtail 0]

http://www.hardocp.com/article/2013/10/02/arma_iii_video_card_performance_iq_review/1#.Ukw7IXfD-5M

http://gamegpu.ru/action-/-fps-/-tps/arma-iii-test-gpu.html

Enjoy :)

PS: I realize the GameGPU tests are a few weeks old but I added them all the same.

Edited October 3, 2013 by BangTail
Added GameGPU tests

[froggyluv 2130]

Crysis 3 and Far Cry 3 move over, Arma III has just surpassed you on outdoor visual quality, fidelity, and performance demand. Arma III is a game that proves the PC platform’s superiority over console gaming. The PC is utilized, the graphics cards are utilized, in a most impressive way. Words cannot describe it, one must experience the game at the highest quality settings to understand the level of detail and visual quality represented in Arma III.

Arma III is what I want as a graphically forward looking game. This game should be the benchmark currently for all others to exceed in terms of visual quality. The single player campaign missions will be out in free DLC format this month. We may follow-up with performance and IQ article in those campaign missions as those may be more complex and demanding than the showcases

.

Bohemia Interactive deserves huge kudos. It provided an impressive game in ARMA III and improved upon the franchise by progressing the game’s engine to Real Virtuality 4 technology. This means better visual quality from the graphics options and tons of post processing effects, that truly showcase the artists on staff. And did we mention the innovative 20km view distance? That is a phenomenal range compared to other video games on the market

.

Wow -great reviews, congrats BI! And people wonder why we love this game :rolleyes:

And why did I now downgrade from Titan to 760 4gb *kicks self*

I can run all in Ultra at 2560 x 1600 but I guess they're maxing out view distance where I keep a conservative 3000 max. Still, great to see that there is room for improvement and that there IS a reason to upgrade video card.

Good stuff bang.

[bushlurker 45]

Kinda puts all the recent small-potatoes, lack-of-furniture, I-wanted-six-different-types-of-chicken Forum bitching into perspective, doesn't it? ;)

B

[bangtail 0]

Kinda puts all the recent small-potatoes, lack-of-furniture, I-wanted-six-different-types-of-chicken Forum bitching into perspective, doesn't it? ;)

B

Indeed it does my friend, indeed it does :)

[Rich_R 1087]

This is a great post that brings in a more objective opinion to offset some of the threads filled with posts by 'excitable' gamers who appear to have interesting expectations.

Nice find!

[Valken 621]

Good article on GPU scaling but you have to note they were using a 4.8 GHZ quad core cpu! That is not an average overclock! 4-4.5 GHZ is about where most PC gamers top out at.

So regular users will probably see less performance even with top of the line video cards.

I wish they did CPU scaling article. That would really help people configure the right system for ARMA 3.

[froggyluv 2130]

^^^ Agreed, I'd like to see CPU scale test as well -but is there really going to be noticeable difference between a 4.4 and a 4.8 OC?

[clydefrog 3]

And a lot of gamers won't overclock their CPU at all nevermind to speeds like that (especially not if they already exceed recommeded specs). By the way remember what most people here on these forums are complaining about. It's not the superficial stuff (graphics) that this review praises so much that people on these forums complain about a lot (there isn't much criticism of graphics at all), it's what you notice once you look past that. As for how "The PC is utilized, the graphics cards are utilized, in a most impressive way.", go and ask most of the people on multiplayer what they think about that.

Edited October 2, 2013 by clydefrog

[bangtail 0]

^^^ Agreed, I'd like to see CPU scale test as well -but is there really going to be noticeable difference between a 4.4 and a 4.8 OC?

It won't be a particularly noticeable difference.

[nik21 1]

Good article on GPU scaling but you have to note they were using a 4.8 GHZ quad core cpu! That is not an average overclock! 4-4.5 GHZ is about where most PC gamers top out at.

So regular users will probably see less performance even with top of the line video cards.

This Test is about GPU benchmarks so they have to make sure that the CPU is limiting as little as possible.

[Dwarden 1124]

there is error in that article, SMAA don't create blur ...

choosing

no PPAA

or

any SMAA

results into no difference of textures (see for yourself nothing is blurred)

the reason FXAA 'looks' crisper and clearer is because of auto-applied sharpen post process filter

which was used to combat FXAA overblurry issues ...

imho the sharpen post process is so good it deserve separate setting so you can use SMAA+sharpen filter too :)

(as SMAA is better than FXAA in terms of edge aliasing quality)

note: there was bug in builds prior game release (beta builds) where SMAA was blurry but that was fixed since ...

Edited October 2, 2013 by Dwarden

[kklownboy 43]

t...

imho the sharpen post process is so good it deserve separate setting so you can use SMAA+sharpen filter too :)

...

and? When?

Nice article for BIS.

[galzohar 31]

Well the way they test GPU scaling is a bit weird. When you do an average like that, you are mixing your results from different areas with different bottlenecks. Sometimes you will get an FPS drop because you had to load stuff from the HDD (I suppose more noticeable when your FPS is high), sometimes you'll get an FPS drop due to CPU needing to do more work (more noticeable when your FPS is high, maybe completely unnoticeable when your FPS is lower, which can create the illusion of lesser GPU/SLI scaling), and sometimes the scene is simply graphically-complex and your GPU is indeed the limiting factor.

For a more realistic test, you'll have to isolate the variables:

1. Make a test where you only load up the GPU - Find a place where standing still in an empty mission gives the lowest FPS, and test the effects of GPU and graphic settings there.

2. Make a test where you load up the CPU - Place lots of patrolling AI units on the map and/or make sure a lot of physics objects are moving around, while standing somewhere remote and looking at the sky, possibly with graphic settings set to minimum (though usually looking at the sky is enough to make your CPU the bottleneck in almost any mission).

3. Make a test for loading objects/textures stressing the HDD/SSD/Ramdisk - Teleport somewhere far away on the map and measure how long it takes the objects/textures to fully load.

This would be much better than just playing through a mission and recording the average performance with different hardware setups and/or graphic settings. In the end we don't care about the average nearly as much as we care about that FPS drop that can happen just as we're trying to shoot someone. Also, the single player missions are a lot less indicative than most multiplayer missions, as many have noticed already (see the many threads complaining about "performance in MP").

Another big problem with measuring something like minimum or maximum FPS is that they can spike up randomly for no apparent reason (or at least nothing to do with the GPU/CPU you're using). For example the load speed of objects/textures is rather random and much more dependent on what you're looking at at any given time, in what order etc, than how it depends on your GPU or CPU. Accidentally looking at the sky will bump up your max FPS like crazy, and done at the right time (for example right when your CPU stress is relatively less).

Keep in mind that your CPU may be stressed even if it's loaded with a measly 30%. As long as you have at least 1 core at any given time working at 100% (which only contributes 25% to the total on a 4-core CPU), your CPU is your bottleneck. Of course, we can't really measure it directly, and must do indirect tests (like the "looking at the sky" test) to really tell. For similar reasons, your GPU can be the bottleneck even if it is far lower than 100%. This isn't ArmA-specific. It is very difficult to write software to take full use of your hardware (while actually getting something useful out of it, of course). The only programs that will take you anywhere near 100% are either programs that are very easy to do that with (such as encoders/decoders or pure benchmarks). Games and more dynamic applications simply can't do that, and even applications that can do it simply won't because it requires many more man-hours to get something like this done and working correctly and efficiently. That said, there is always room for improvement in taking better advantage of your hardware. But don't expect any games to use anywhere near 100% of a quad-core CPU while also working properly.

Edited October 2, 2013 by galzohar

[gavc 28]

thanks for this, good read.

[RSF TheCapulet 59]

there is error in that article, SMAA don't create blur ...

choosing

no PPAA

or

any SMAA

results into no difference of textures (see for yourself nothing is blurred)

the reason FXAA 'looks' crisper and clearer is because of auto-applied sharpen post process filter

which was used to combat FXAA overblurry issues ...

imho the sharpen post process is so good it deserve separate setting so you can use SMAA+sharpen filter too :)

(as SMAA is better than FXAA in terms of edge aliasing quality)

note: there was bug in builds prior game release (beta builds) where SMAA was blurry but that was fixed since ...

http://feedback.arma3.com/view.php?id=15175

[calin_banc 19]

The only programs that will take you anywhere near 100% are either programs that are very easy to do that with (such as encoders/decoders or pure benchmarks). Games and more dynamic applications simply can't do that, and even applications that can do it simply won't because it requires many more man-hours to get something like this done and working correctly and efficiently.

BF 4 Beta runs at over 90% most of the time, hitting 97-98% load on the CPU - 2500k@4,5GHz and it hold back a 7950. It's all about what you throw at it, some code can scale really well.

That aside, I'm really impress of how well the game feels even at lower FPS, with very low input lag on the controls . which is the one thing that hinders the most, in a gameplay session at lower than, let's say, 60fps. Runs very well even on multidisplay and overall it's great step forward from ArmA 2. A truly "next gen".

Indeed, it doesn't have some nice effects like the other games have, however, this can be relatively easily be implemented compared to what they've manage to do this far. Bohemia deserve some sort of a medal. :)

[ric 1]

there is error in that article

there is also another error..

As we stated above, this game is solely developed on the PC. This is not a console port, and therefore this game can utilize the PC to its full potential.

is there a part in that article that shows CPU scaling?

[Cathrynn 10]

Was watching some youtube vids of Arma 2 the other day and I couldn't get over just how much better Arma 3 looks in comparison, BI really have raised the bar.

[mamasan8 11]

there is also another error..

is there a part in that article that shows CPU scaling?

That comment confused me. Full potential?

My CPU & GPU are never running 100%, both are almost idle, going by temps.

In other games under load, CPU is 40C, GPU 50C +.

In Arma 3, CPU is 2 degrees above idle temp, 32C and graphics card around 40-45C. Its like something is holding back my whole machine.

CPU cores at 70% utilization max, graphics usually 0-20% <- Only rises if I use Antialiasing 4x or 8x. But still low utilization, maybe 50%.

[calin_banc 19]

Most likely he was referring to the size corroborated with the level of detail, which means you have great detail up close, but also from the distance. Overall I'd say it looks better than BF 3 and performs better in optimized scenarios (same for BF 4 from what I've played in the Beta). If you look at the industry seen as a whole, there isn't really a game that does the same things as ArmA 3. You can have the game or engine and turn it into whatever you want: flight sim, sandbox GTA, Mafia, JC style, DM, TDM or whichever mode you prefer from any numbers of games and so on. Those (other) games cannot really do that or scale upwards to the size and complexity of ArmA. The only other game, that I'm aware of, which will come close or go over this standard, will be Star Citizen in about 2 years time.

[jumpinghubert 48]

can´t find information about details of the benchmark: with Ai or without, flyover, vehicle, inf, on altis or stratis, in town or not..

[mr_centipede 31]

It's about GPU benchmark, so probably no AI, single player unit only.

[galzohar 31]

can´t find information about details of the benchmark: with Ai or without, flyover, vehicle, inf, on altis or stratis, in town or not..

They specifically mentioned the single player showcases.

[RSF TheCapulet 59]

Yup, the command showcase, to be specific.