The limited time open beta for Battlefield 3 started last week, and we've spent some time with it. Performance is quite a wild ride in this beta, and it doesn't even have the full graphics the retail version will, but we'll examine it closely to show you how your video cards might stack up when Battlefield 3 finally launches in a few weeks.
Introduction
DICE's Battlefield franchise is one of the best know shooter series in PC gaming. Beginning in 2002 with the World War II shooter Battlefield 1942, the series has had at least six full-length games, several smaller downloadable games, and several expansion packs. In about three weeks, the next big title in the series, Battlefield 3, will be released. In lieu of a pre-release demo, EA has made access to a beta build of the game public. The beta began on September 29th and is set to end on October 10th.
Article Image
Battlefield 3 Public Beta
Though the code appears to be of early beta (or even alpha) quality, and it is missing some graphics features, the game is surprisingly demanding of processing horsepower, including CPU and GPU. We know the performance situation is very likely to change once the retail game is released, but we wanted to get some data and screenshots for our readers for the public beta.
The Technology
Battlefield 3 is powered by EA/DICE's Frostbite 2 engine. The PC version of the Frostbite 2 engine supports 64-bit processors and Microsoft's DirectX 11 technology. In fact, DirectX 9 is not supported at all, and so the game will not run in Windows XP. Battlefield 3, and its beta, require either Windows Vista or Windows 7.
Frostbite 2 and Battlefield 3 support such graphics features as tile-based deferred shading via DirectCompute, Morphological AA (MLAA) also via DirectCompute, radiosity lighting, bokeh depth of field (DoF), and ambient occlusion in both SSAO and HBAO formats. One of the most anticipated graphical features of the game is its supposedly more realistically destructible environments.
We will have more detailed information when the retail game launches. For now, what is important is that (1.) Battlefield 3's premium platform is the PC, (2.) Frostbite 2 is native to DirectX 11 on the PC, and (3.) this public beta is one of the most graphically demanding DX11 games we've seen in a while.
Test Setup
For our test system platform we are using an ASUS P6T6 WS Revolution motherboard with an Intel Core i7 920 overclocked to 3.6GHz, and 6GB of Corsair DDR3-1600. For the power supply, we will be using a CoolerMaster Real Power Pro 1250W.
Article Image
For all three NVIDIA-based video cards in this evaluation, we are using NVIDIA's GeForce/ION Driver 285.38 Beta package, dated 26 September 2011. For the three AMD-based video cards in this evaluation, we are using AMD's Catalyst 11.10 Preview driver package, dated 26 September August 2011. These driver versions were specifically launched to improve performance in the BF3 open beta.
Performance Testing
Performance in the BF3 open beta varies to a great extent. The two available maps are large, and performance can vary significantly between one part of either map and any other part. Thus, testing the BF3 open beta was a tremendously frustrating experience. Due to the nature of this being a multiplayer only map without the full-game's features, we were unable to reproduce a testing run with anything remotely resembling consistency. In the end, we were forced to just play the game for a while on each video card while adjusting settings to make sure it was playable everywhere we tested to find the highest playable settings in the maps offered.
For this process, we chose to focus on the "Operation Métro" map. Taking place in Paris, France, this map focuses on Paris' famed Métropolitain rapid transit system. While it focuses on the mostly-underground trains, it also takes us to wide open parks and narrower urban corridors. To record performance, we played for about 7 minutes, recording framerates with FRAPS, looking for an average framerate between 38 and 42. This envelope will likely change for the full version of the game, since it will have more graphics features. For now, within a few FPS of 40 is where we were looking to be.
In general, the subterranean areas seem to perform better than the above-ground areas. It's not consistent though, as there are sometimes large firefights underground which can drag performance down. The game seems to also enjoy randomly spawning players underground regardless of whether or not that is where the action is. All of these things make performance testing in the beta a bit frustrating to get the most consistent results.
Highest Playable Settings
Article Image
The BF3 open beta was playable on the NVIDIA GeForce GTX 580 at 2560x1600 with 16X AF, no AA, and mostly Ultra settings. We did have to disable motion blur and reduce the shadows setting to medium, but other detail settings were left at Ultra. In disabling MSAA, we also disabled "Post AA", which is MLAA.
The Radeon HD 6970 performed similarly. At 2560x1600, we disabled MSAA and MLAA, as well as motion blur in order to bring framerates up to an average of about 40 FPS and make the game playable. With the HD 6970, we did not have to lower the shadow quality setting. Both of these video cards gave us very high framerates at 1920x1200 with 4X MSAA and maximum in-game settings selected.
The GeForce GTX 570 and the AMD Radeon HD 6950 were both playable with 4X MSAA, Medium "Post AA" (MLAA), 16X AF and Ultra settings at 1920x1200. We didn't have to disable motion blur or change the shadow quality setting for either of these two video cards at this resolution.
Running at 1680x1050 was not a challenge for either the NVIDIA GeForce GTX 560 Ti or the AMD Radeon HD 6870. They were both playable at that resolution with maximum in-game settings at a very high level of performance. For our highest playable settings, we chose to increase the resolution to 1920x1200 and tweak settings from there. Neither of these two inexpensive video cards was playable in the entire test level with maximum settings at 1920x1200, so we had to disable MSAA and MLAA on both. With the GeForce GTX 560 Ti installed, we also had to disable motion blur. With the Radeon HD 6870, disabling AA was all we had to do to make the game playable at 1920x1200 with Ultra in-game settings selected.
Individual Performance
Because we were not able to reproduce the same (or even similar) testing procedure for each video card, performance graphs will be presented individually for each of our six video cards. All of these graphs represent data gathered using each video card's respective highest playable settings, as described in the table above.
NVIDIA GeForce GTX 580
Article Image
With mostly Ultra settings selected and AA disabled at 2560x1600, performance was relatively smooth with the GeForce GTX 580. We saw an average FPS very close to our 40 FPS target, and it only dipped below 30 slightly and infrequently.
AMD Radeon HD 6970
Article Image
Like with the GeForce GTX 580, performance dipped below 30 FPS infrequently with the Radeon HD 6970. Our average FPS was higher here, but that doesn't mean much with performance as unpredictable as it is in this beta.
NVIDIA GeForce GTX 570
Article Image
The GeForce GTX 570 ran beautifully at 1920x1200 with Ultra settings selected. Choosing Ultra automatically enables 4X MSAA and Medium Post AA (but you have to restart the game to see those changes). Framerates rarely dipped below 30 FPS, and when they did, it wasn't far. Our average was within 1 FPS of where we wanted to be, as we spent most of this test above ground.
AMD Radeon HD 6950
Article Image
The Radeon HD 6950 performed a little better than the GeForce GTX 570, though it is hard to be conclusive about it given the state of this beta. We didn't see a single drop below 30 FPS during testing with Ultra settings @ 1920x1200 using the HD 6950.
NVIDIA GeForce GTX 560 Ti
Article Image
The GeForce GTX 560 Ti was the most challenged video card in this test. At 1920x1200, we had to disable AA and Motion Blur, but other settings were left at Ultra. We saw occasional dips below 30 FPS. Our average framerate was acceptable and the gameplay experience was smooth enough, so we were not compelled to reduce settings any further.
AMD Radeon HD 6870
Article Image
Again, the AMD video card in this price segment had slightly better performance than the NVIDIA counterpart. The Radeon HD 6870 was playable with Ultra settings at 1920x1200, but we had to disable AA, including both MSAA and MLAA. Framerate dips below 30 FPS were rare, but they did happen. Our average framerate was right in target, and the resulting gameplay was very smooth and enjoyable.
Image Quality
We were unable to take screenshots with any degree of consistency due to the unpredictable nature of the game and its spawning system. Add to that that we have to restart the game to apply some settings, and side-by-side comparisons have proven to be practically impossible to acquire in the BF3 open beta. For this image quality section, we are going to show you a collection of screenshots at various in-game setting levels. We'll look at Ultra, High, Medium, and Low. The following linked images are JPEG compressed to save bandwidth, but we used high quality compression. We aren't after fine detail in these screenshots as much as we are after a comparison of the overall visual experience the different graphics levels provide. We will have a more in-depth look at image quality in the full version game evaluation.
Ultra
Article Image Article Image Article Image Article Image Article Image
High
Article Image Article Image Article Image Article Image Article Image
Medium
Article Image Article Image Article Image Article Image Article Image
Low
Article Image Article Image Article Image Article Image Article Image
Image Quality Analysis
The most obvious difference here is the color tone. With Ultra and High selected, the color is washed out somewhat in the bright sun. There is a lower level of contrast with Ultra and High selected than with Low and Medium. Looking at the third image in each series, we can also see that there is a difference in the Depth of Field quality. In ultra, the focused area is clear and well-defined. In high, some clarity is lost in the focused part of the image. With medium selected, it is hard to tell what is supposed to be in focus. Finally, with low selected, depth of field appears not to be enabled at all. The character has just been hit, but the resulting blurred screen did not happen as it did with medium, high, or ultra selected.
Beyond those differences, anti-aliasing is a big issue. With ultra selected, 4X MSAA and Medium MLAA is enabled. Going down to high disabled MSAA but keeps MLAA at medium. Selecting medium moves MLAA down to low, and selecting low disables MLAA. This is a game that benefits a great deal from anti-aliasing. Unlike Battlefield: Bad Company 2, it appears that anti-aliasing happens after the HDR stage of post-processing. Some gamers will recall that aliasing is nearly defeated by HDR in BFBC2. Edges that appear in front of bright light sources would appear jagged and as if AA is ignoring them. In the BF3 beta, that problem doesn't happen. Of course different people have difference levels of sensitivity to aliasing, but we found that AA was quite welcome and effective in the BF3 beta. We will, of course, have a lot more information and analysis on AA in Battlefield 3 when the full game launches.
Beta Is As Beta Does
Wow, does this game ever feel and behave like a beta. Though we played for several hours, we were hard-pressed to find results even remotely consistent, even with the same video card installed. First we'd get spawned in the underground train stations and get our settings adjusted. Then we'd spawn out in the open and have to adjust them again. It took a significant amount of going back-and-forth between different servers through several different rounds to make sure each video card was tested on all relevant parts of the Operation Metro map. That is also excluding the many crashed and random disconnects we experienced, and the fact that we had to restart the game client on purpose several times for each test so that we could adjust settings for the sake of tuning performance.
At the end of the testing day, the impression we are left with is that performance was so inconsistent, it actually seems rather hard to predict how the full version game will behave. We know that some graphics features are missing in the beta client, and we know that a great deal of content is missing as well. This game looks like it is going to be of a simply tremendous scale. We are certainly anxious to see what the retail build will look like. It almost certainly has to be better. As unstable and flaky as the beta client has been for us, it is hard to imagine having a more frustrating experience.
The Bottom Line
Performance in the Battlefield 3 public beta inconsistent. The multiplayer nature of the beta made it impractical to objectively compare performance between video cards, but we did see some relatively repetitive behavior among video card brands. In general, we had better performance using video cards equipped with AMD GPUs than we did with NVIDIA GPUs. We downloaded the newest beta drivers from both GPU makers specifically released for this game. We tested each video card thoroughly, and the results with AMD video cards were faster, but not by much in this open beta.
On the official BF3 Blog, EA/DICE staffer Joe Ellis pointed out that "The beta won’t include all the graphics features which will make it in the final version". When the final retail build is released, everything we've seen so far could be turned around completely. But for now, single-GPU gamers with AMD Radeon HD 6000 series video cards will probably see slightly better performance than gamers with NVIDIA GeForce GTX 500 series counterparts.
<Follow the original link to see all the charts and graphs which accompany the article>
Its all shot placement. Your brother proved it.
