Battlefield 3 Preview Performance Preview @ [H]

[Kor]

I'm running dual GTX 580 cards. Is anyone else having problems getting the game to start using surround gaming on three 1920x1200 monitors? When I try to start a game my monitors go blank, then they alternate on and off until they each go to sleep, then they will wake up again with a black screen until finally I get a DX error and I can see windows. I'm running win 7 64.

Yeah trying to run single screen with a display group active is completely broken, you will need to either switch off you side screens or disable the group when you want to play. I have the same issue in Eyefinity.

 

[cocheese256]

Yeah trying to run single screen with a display group active is completely broken, you will need to either switch off you side screens or disable the group when you want to play. I have the same issue in Eyefinity.

I'm not trying to run a single screen in a display group. I am trying to run the game on all three monitors (surround gaming, same as eyefinity). Sorry, I may not have made that clear. It runs fine on one display.

 

[Kor]

I'm not trying to run a single screen in a display group. I am trying to run the game on all three monitors (surround gaming, same as eyefinity). Sorry, I may not have made that clear. It runs fine on one display.

Ah I was thrown off by the 1920x1200 bit.

 

[macksomerville]

I dont know what the hell that finally did it... but I am running tri-fire 6950 toxics o/c'd to 885/1385 (for stability) on eyefinity portrait (3x 2007WFP) stutter free... NO CAP3, fresh install of Beta3's

Didnt bother to log FPS but its smooth. ultra, no msaa, no motion blur. Cat control: AA application, 16x, Quality, quality, quality (vsync on)


Looks like shit since my lian li cant hold the 3 way with a Rampage II extreme but the radial blower keeping temps low

 

[samuelmorris]

Unless your running an Eyefinity6 edition

No it's still true. The difference is not due to the extra memory, it's due to how the HD6970's core is designed.

 

[Tigerblade]

I dont know what the hell that finally did it... but I am running tri-fire 6950 toxics o/c'd to 885/1385 (for stability) on eyefinity portrait (3x 2007WFP) stutter free... NO CAP3, fresh install of Beta3's<snip>

Now that is hardcore lol

Why portrait tho? I thought the whole point of eyefinity was to get more peripheral vision, portrait with 3 monitors just seems to have the same effect as playing up close on a large screen tv or does it give more peripheral higher and lower?

 

[macksomerville]

I didnt do it for Peripheral... I did it so I can feel like I am playing on a bigger monitor than my 3 20" really are.. ZR30w or 2007wfp... 800-1300.... 3x dell 2007wfps on ebay? $320

 

[Tigerblade]

I didnt do it for Peripheral... I did it so I can feel like I am playing on a bigger monitor than my 3 20" really are.. ZR30w or 2007wfp... 800-1300.... 3x dell 2007wfps on ebay? $320

Fair do's!

 

[Caspnj]

Im still holding onto my Q6600 (OC to 3.3) and it is just about tapped out on this game. In monitoring the CPU and GPU i see that the CPU is at about 94-100% and the gpu ranges from 70-100%. I am also running a Gtx 480. My frames are good but I think there is an upgrade in my future. But for anyone wonderign how it runs on older cpus....pretty darn good.

 

[dioxholster]

so i should expect 60fps with my rig? i hope.

 

[Woodhull]

Over at rage 3d they say these are official only they are leaked as AMD wasn't going to release them yet, however I completely understand. To the capacity of writing a article for accuracy you should only use the official caps as many users may shy away from these. So far members of this forum are saying that these caps are amazing.

Official CAP 4 d/l site is up:

http://sites.amd.com/us/game/downloads/pages/crossfirex-app-profiles.aspx

 

[TheQuebecerInFrance]

Im still holding onto my Q6600 (OC to 3.3) and it is just about tapped out on this game. In monitoring the CPU and GPU i see that the CPU is at about 94-100% and the gpu ranges from 70-100%. I am also running a Gtx 480. My frames are good but I think there is an upgrade in my future. But for anyone wonderign how it runs on older cpus....pretty darn good.

What tool are you using to monitor this. My Q6600 is OC to 3ghz and that may be the first game that a actually brings it down. Still best bang for the buck I ever had. But I can feel that 2500K calling me.

 

[pandora's box]

It is, Antialiasing Deferred is standard traditional Multisample AA. I.E. 2X, 4X, 8X AA.

Antialiasing Post is NVIDIA's Shader Based FXAA method, which provides 4X AA like quality on all surfaces and objects and texture types with barely any performance hit.

You should never have both enabled at the same time, its a waste of resources, and will harm framerates.

Either use MSAA or FXAA, not both. I prefer FXAA to get great image quality with hardly any performance hit.

The options could stand to be labelled better for clarification.

DICE disagree's:

http://forum.beyond3d.com/showpost.php?p=1586489&postcount=995

 

[Lord_Exodia]

DICE disagree's:

http://forum.beyond3d.com/showpost.php?p=1586489&postcount=995

I think your taking what he's saying too far out of context. He's not saying you need to have both he's saying they made the game that way because they can complement each other. You can see a benefit in running both types of AA methods simultaneously, however from posts I've seen from them they don't say you "need" or even "should" enable both. They have even gone as far as saying what brent is saying. Either one is fine but if you have a system that can handle both then you can enable both and they will complement each other with a very slight boost in IQ together over running either at high.

 

[Lord_Exodia]

I'll wait for the official CAP to test with it.

http://sites.amd.com/us/game/downloads/pages/crossfirex-app-profiles.aspx

Testing Time!! (does testing time dance )

 

[Silent Knight]

Looks like the final game requires less VRAM than the beta did.It could be an irrelevent comparision but i got minimum of 45-50fps,maximum up to 100fps in SP @1920x1080 with everything on Ultra,8xAF,no MSAA and no Motion Blur.While in the beta i was kicked out because it was not enough VRAM (also with lower setting),fps was ranging from 30-80fps.

But i got 6 BSODs while playing SP,not sure what went wrong.

 

[TwistedAegis]

Caspian Border 64 beta didn't feel smooth on my rig, even at 1920x1200; there was terrible stuttering. Now with retail, same drivers I had before, I can run 2560x1600 ultra (post-processing AA only) and it runs great. Really, really happy as I was a bit upset at the performance of the beta.

 

[samuelmorris]

Just launched the start of the campaign with stock 11.9 @ 2560x1600:
All Ultra: 24fps
All Ultra, Deferred AA off (postAA and HBAO still on): 41fps.

GPU1: 97% usage
GPU2: 0% usage

Time to gets me that application profile...

Edit: Still no crossfire support with CAP4 and 11.9. Assuming I need also to be running one of the 11.10 preview drivers.

Edit2: Installed 11.0 preview3 and CAP3. Crossfire now works, but the game can only run for 20 seconds before it will crash. Crossfire support for this title? Not so good...

Edit3: Can't replicate the crash. Given what I've seen so far, am attributing it to a game bug. I've seen several already, and I'm only 2 minutes in.
On the bright side, 99% GPU usage on both cards. Frame rate in the campaign at least, is reasonable, 45-55fps all Ultra.

 

[Shago]

My htpc:
- amd phenom 2 x3 720 unlocked to quad core (no oc)
- 1gb 4870 (no oc)
- 4gb 1333mhz ram

Battlefield 3:
- resolution: 1920 x 1200
- detail settings: medium

Plays very fluently, ofcourse I can't calculate an avarage fps but when you do ctrl+alt+s I usually am around 45 and don't notice any slowdowns whith explosions.

 

[-Sn1PeR-]

Edit2: Installed 11.0 preview3 and CAP3. Crossfire now works, but the game can only run for 20 seconds before it will crash. Crossfire support for this title? Not so good...

Do you mean Cap4? Also try turning down deferred AA and turning up Post Processing AA for a higher frame rate. Disabling HBAO also gives a pretty nice boost at a very minor visual cost.

 

[samuelmorris]

Yes, sorry CAP4, and yeah I'm going to look into that

 

[cold2010]

It seems that I know why you did not use MSAA in spite of striking the barrier 70 frame

When we started testing on our own it became apparent why AMD did this. With MSAA enabled the performance sharply declines with Radeon boards and previous generation Nvidia products make AMD’s current generation hardware look foolish. However disabling MSAA allows AMD to catch up to Nvidia, as you saw in our High quality 1680x1050, 1920x1200 and 2560x1600 tests.

All of our testing is at apples-to-apples settings. We set the game to 2560x1600 and set all in-game options to "Ultra" to max it out. Antialiasing Deferred was set to "Off," this is traditional MSAA and that is a huge drain on performance. Instead, we used Antialiasing Post which is FXAA, we set this to "High." We had 16X AF enabled and HBAO enabled.

Do you have reached this FAQ &#1567;

AMD Recommends Using Nvidia's FXAA in Battlefield 3

Yesterday AMD sent us a benchmarking guide for Battlefield 3 written by Antal Tungler, AMD's PR Manager. In the guide, Antal recommends using Nvidia's FXAA instead of MSAA.

"You'll notice that we recommend the 'Ultra' preset in lot of our scenarios, without the use of MSAA (Multi-Sample Anti-Aliasing). This is a traditional technique to remove aliasing from objects in the visible scene. It provides significant image quality improvements, at least in most cases. But because of its nature, working only on actual geometry, some objects will not benefit at all from this process, like foliage or fences, that are just textures, sprites etc. Luckily DICE implemented a new technique as well called FXAA (Fast Approximate Anti-Aliasing), which is a process that works pretty much on all objects. Another significant difference between the two is that while FXAA has a very small impact on performance, however using MSAA can essentially cut your framerate in half."

"If we view the same data in another way we can also see that once we add 4xMSAA we are lowering our frame rate so much that we are moving from a playable to non-playable gaming scenario on some high end graphics solutions if we use the standard 30fps average for playable. Many of the extremely competitive Battlefield 3 players will want as many frames as they can get making MSAA a questionable option! We can also see from both charts provided that FXAA has a relatively low performance cost for the great level of Image Quality it provides."

"The idea behind the following pages comparisons is not to discredit MSAA or the implementation of MSAA in Battlefield 3. MSAA is a wonderful technique that can be used to improve the image quality of your gaming. Users should be wary however that MSAA will not improve all objects in the game and comes at a high performance cost to the user that may not be necessary for someone looking to maximize their performance and gaming experience. It is highly recommended that users take time to toggle settings themselves and determine if the Image Quality improvement that MSAA provides is worth the performance hit their frames per second will take while gaming

 

[HohnyF]

For the best performance and image quality, use "Antialiasing Post" which is FXAA in-game.

FXAA is a nice addition to an AA techniques list, but it's hardly a substitute for a traditional AA. Not if you want the best possible image quality.

And MSAA (+ FXAA) is not a huge drain on performance.

It's just AMD that takes huge hit, but you already knew this.
Yet you managed to write the whole performance analysis article without even mentioning it

 

[Forceman]

FXAA is a nice addition to an AA techniques list, but it's hardly a substitute for a traditional AA. Not if you want the best possible image quality.

And MSAA (+ FXAA) is not a huge drain on performance.

It's just AMD that takes huge hit, but you already knew this.
Yet you managed to write the whole performance analysis article without even mentioning it

MSAA + FXAA isn't a huge hit, but FXAA + MSAA is. And the difference between FXAA and FXAA + MSAA is small enough that it (normally) isn't worth the performance hit.

 

[HohnyF]

Not trying to be a dick, but you're not making any sense.
You can turn Post and Deferred AA ON or OFF. Particular order does not matter.
And rendering wise, Post is allways done... well post - on a fully rendered frame.

Image quality wise...
Of all the games I tried only Tropico 4 (exterior FXAA) looks decent with only FXAA.
(FEAR3 and Dead Island 2 to an extent)

Every other game I've tried, at 1080p looks like crap without a proper AA.
And that especially goes for two most recent ones, BF3 and RO2, and especialy while on move.
Come to think of it, post processing AA on DX:HR is not too bad either, but that game only features long straight lines to being with

Anyway...if [H] was among those few reviewers that run the benchmarks according to AMD benchmrking guide for BF3... that's fine.

Lets just not call that ULTRA, and how about mentioning it, if the card performance goes tumbling down in a proper ULTRA preset, aaight

 

[Forceman]

Not trying to be a dick, but you're not making any sense.
You can turn Post and Deferred AA ON or OFF. Particular order does not matter.
And rendering wise, Post is allways done... well post - on a fully rendered frame.

The point was - adding FXAA to an image already using MSAA (MSAA + FXAA) doesn't cause much of a performance hit - but adding MSAA to an image already using FXAA (FXAA + MSAA) causes a significant performance hit.

Running both MSAA and FXAA isn't enough of an image quality gain to justify running it over just FXAA, because adding the MSAA incurs such a significant performance hit.

 

[HohnyF]

I already said what I think of IQ running exclusively FXAA, so that aside...

How do you add

MSAA to an image already using FXAA (FXAA + MSAA)

You mean forcing MSAA through AMD/NV control panel on top of using Deferred in the game?
Who would want to do that?

 

[Forceman]

I already said what I think of IQ running exclusively FXAA, so that aside...

How do you add

You mean forcing MSAA through AMD/NV control panel on top of using Deferred in the game?
Who would want to do that?

No, I mean running both deferred AA (MSAA) and Post AA (FXAA) in BF3 specifically (the topic of this thread), which is what happens when you select Ultra presets.

 

[CSC Tiscali]

I havn't seen any reviews on 3-way SLI performance and scaling for BF3, so I slapped some numbers from Fraps together to make a simple chart. It will at least give us an idea on 3-way SLI scaling in BF3...I'm glad to see very good scaling results as of yet, albeit at 1920x1080.

I used the same canned benchmark as tomshardware, using the same sequence as seen in the following video...
http://www.youtube.com/watch?v=wAA-NJj0Z0k&feature=player_embedded
2-way SLI scaling numbers were consistent with Tom's. 5760x1080 benchmarks to come at some point in the future.
Test setup in sig

 

[Roadhog]

Here are my results using 3 6970's and copying the tomshardware bench sequence exactly.

 

[CSC Tiscali]

Finally some 2d surround goodness. 4xMSAA was unplayable due to most likely hitting the VRAM wall, where the menu/cursor instantly becomes sluggish, which carried over into gameplay as indicated in the unplayable framerates.

I played on Caspian Border with 64 slots and the only setting I could find consistently smooth was 0xMSAA. I am a nitpick when it comes to framerates not dipping below 60fps, so I was pretty pleased with the peformance. IMO, the sequence benchmarked stresses the GPU's about 80% as much as a 64 player map with a lot of action going on. Nothing beats playing the actual game and getting a good feel for playable settings.

 

[SaViOr-]

Just launched the start of the campaign with stock 11.9 @ 2560x1600:
All Ultra: 24fps
All Ultra, Deferred AA off (postAA and HBAO still on): 41fps.

GPU1: 97% usage
GPU2: 0% usage

Time to gets me that application profile...

Edit: Still no crossfire support with CAP4 and 11.9. Assuming I need also to be running one of the 11.10 preview drivers.

Edit2: Installed 11.0 preview3 and CAP3. Crossfire now works, but the game can only run for 20 seconds before it will crash. Crossfire support for this title? Not so good...

Edit3: Can't replicate the crash. Given what I've seen so far, am attributing it to a game bug. I've seen several already, and I'm only 2 minutes in.
On the bright side, 99% GPU usage on both cards. Frame rate in the campaign at least, is reasonable, 45-55fps all Ultra.

Try running the 11.10 Preview 3 Drivers with CAP4. 99% Usage on All 4 GPUs at 5760x1080 and 50-60% at 1920x1080 (Lower Usage since the game is capped at 200 so no more usage than that is required per gpu at 1080p, fps 140-200 MP)

 

[samuelmorris]

Finally some 2d surround goodness. I would throw out the 4xMSAA results as there seems to be a bug from the very moment I switch it to the setting, where the menu/cursor instantly becomes sluggish, which carries over into gameplay as indicated in the unplayable framerates.

I played on Caspian Border with 64 slots and the only setting I could find consistently smooth was 0xMSAA. I am a nitpick when it comes to framerates not dipping below 60fps, so I was pretty pleased with the peformance. IMO, the sequence benchmarked stresses the GPU's about 80% as much as a 64 player map with a lot of action going on. Nothing beats playing the actual game and getting a good feel for playable settings.

That's not a bug. You're out of video memory.

 

[CSC Tiscali]

That's not a bug. You're out of video memory.

Point taken. 1400MB was my VRAM utilization when switching over to 4xMSAA. My computer was never cripplied when running at over 1500MB, even in multiplayer. What's the explanation for that?

 

[samuelmorris]

It's only going as far as 1400MB because video memory does not fit neatly right up until the maximum. At only 2560x1600 even without MSAA I'm at 1.6GB per card. At your res with 4xAA, you'd probably find yourself using in excess of 2.5GB per card if you actually had it. The 1.4GB just represents what will actually fir on your card. At reduced MSAA/detail the game can dynamically adjust detail to prevent frame rate drops, but with that much MSAA you're so far over the limit it can't compensate for it, so your performance drops off a cliff.

 

[CSC Tiscali]

It's only going as far as 1400MB because video memory does not fit neatly right up until the maximum. At only 2560x1600 even without MSAA I'm at 1.6GB per card. At your res with 4xAA, you'd probably find yourself using in excess of 2.5GB per card if you actually had it. The 1.4GB just represents what will actually fir on your card. At reduced MSAA/detail the game can dynamically adjust detail to prevent frame rate drops, but with that much MSAA you're so far over the limit it can't compensate for it, so your performance drops off a cliff.

Thanks. "Hitting the wall" is quite an understatement then. It's more like "hitting a freight train head on" when switching over to 4xMSAA. 64FPS to 10FPS average with a click of a button.

 

[xebo]

is HardOCP going to follow up this preview with a larger review? More cards, more settings, more detailed info?

 

[chuch]

sweet.

 

[HohnyF]

It's only going as far as 1400MB because video memory does not fit neatly right up until the maximum. At only 2560x1600 even without MSAA I'm at 1.6GB per card.

Ideally coded it would use whole 2GB. Caching, precaching call it what you want.
That does not mean it NEEDS 2GB not to fall of the cliff.

See that 1.5GB framebuffer on GTX 580...
It's perfectly capable of running 5760x1080 with 2xMSAA
It needs 4xMSAA to hit the wall.

 

[blazarcher]

Anyone know if the 5970 will play this game maxed settings at 1920x1200? And when I mean play, I mean smoothly =)