PC gamers who love great-looking first-person shooters have more than a few big-name franchises to choose from. But if you're particularly enchanted with large-scale multiplayer battles that include a wide array of player-piloted vehicles, one series stands above the rest. The Battlefield games weren't the first to focus on that style, but they certainly had a hand defining the genre.
The imminent release of Battlefield 4 is, therefore, a big deal to PC gamers. We spent some time with the public multiplayer beta to see what's new and check out pre-release performance on a wide array of graphics cards and processors.
Gone in 60 seconds
Of course, this means our analysis is limited to online play for now (though that tends to be what makes the series so long-lasting). The basics haven't changed: the four player roles are still assault, support, engineer, and recon. The beta offers the staple conquest-style game with large vehicle-populated maps, along with an infantry-focused domination option on a smaller map and void of mechanical transportation. The new obliteration and defuse games are not available yet. Neither are competition favorites like team deathmatch, rush, or squad deathmatch. There's only one map in the beta test: Seige of Shanghai. Like Battlefield 3, the game servers are selected via a Web browser.
Server browser
The sense of scale is much larger than any Battlefield game I've played before, thanks to the sprawling city and large buildings. The new Frostbite 3 game engine facilitates gorgeous visuals that remind me a lot of Crysis 3. There are other changes, like improved water dynamics, the commander mode (also available from Battlefield: 2142), and the new "levolution" feature that gives you the ability to destroy large-scale map features, such as buildings, to move the location of objectives. While levolution sounds cool, it appears to be hard-coded into specific structures. My understanding was that only one building in the Seige of Shanghai could be destroyed. It was already down in every server I joined, though, so I never saw this feature in action.
Chopper up...
I would have liked to see fully-destructible terrain, though of course that introduces logistical issues. We're hoping, then, that the final version of Battlefield 4 includes at least one map with more "levolvable" structures.
Otherwise, the beta's environment is fairly standard, perhaps a bit more interactive than prior titles in the franchise. There's a lot of glass and architectural detail that demonstrates abuse from gunfire and explosions, and much of the concrete cover is destructible. Most of the buildings are static above the first floor, though.
Chopper down.
Thus far, Battlefield 4 looks like it's shaping up to be a polished, better-looking version of its predecessor with even more interesting environments to play in.

We all know we're going to be busy playing Battlefield 4 when it comes out, so let's take a moment during the beta period to discuss DICE's Frostbite 3 engine.
Again, Battlefield 4 employs an updated Frostbite 3 engine, the newest version of Digital Illusions CE's game technology for the next-gen console and PC platforms. While it makes its commercial debut with Battlefield 4, Frostbite 3 will also power Need or Speed: Rivals later this year, in addition to the next iterations of the Dragon Age, Mass Effect, and Star Wars: Battlefront franchises.

Compared to it's predecessor, Frostbite 3 features higher-resolution textures, particle effects, and changes to tessellation, according to the company's feature video. A new networked water feature ensures that all players see the same waves in the water at the same time, allowing small naval craft to hide behind waves in rough seas.
Leavin' On A Jetski
First we tested the Low, Medium, High, and Ultra detail presets, finding that the texture detail on Low appeared dependent on the game type. We noticed higher-resolution textures on the Low setting in domination mode compared to conquest, even on the same Seige of Shanghai map. It's not clear whether this is a beta glitch, or a result of the game dynamically allocating resources based on the number and size of models in the map.


We chose to benchmark the domination map because of its higher texture resolution on the Low setting. We didn't notice a significant performance hit shifting between the Medium and High presets, so we tested Low (MSAA off, AA Deferred off, Ambient Occlusion off), High (MSAA off, AA Deferred high, HBAO enabled), and Ultra (4x MSAA, AA Deferred high, HBAO enabled) detail presets.


The challenge of benchmarking a multiplayer game is that every run is potentially different, altering the load from one test to the next. For this reason, we performed our measurements on servers with low 40 ms-or-less latencies and no other players. Populated servers are likely to exact more demanding CPU loads. However, we needed to eliminate this variable from our testing. The good news is that, when we compared our numbers to runs on servers with more people on them, frame rates didn't change noticeably.

We all know that graphics cards like the Radeon HD 7990 require a substantial amount of power, so XFX sent along its PRO850W 80 PLUS Bronze-certified power supply. This modular PSU employs a single +12 V rail rated for 70 A. XFX claims that this unit provides 850 W of continuous power (not peak) at 50 degrees Celsius (notably higher than the inside of most enclosures).

We've almost exclusively eliminated mechanical disks in the lab, preferring solid-state storage for eliminating I/O-related bottlenecks. Samsung sent all of our labs 256 GB 840 Pros, so we standardize on these exceptional SSDs.
As far as testing goes, we have to use Fraps in conjunction with a predefined path for 60 seconds of recording. We planned to use our FCAT tools to report frame rates for dual-GPU solutions like the Radeon HD 7990 and GeForce GTX 690, factoring out dropped and runt frames, but this turned out to be impossible. The frame overlay only works in 32-bit applications, and the Battlefield 4 multiplayer beta is 64-bit-only.
| Test System | |||||
|---|---|---|---|---|---|
| CPU | Intel Core i5-2550K (Sandy Bridge), Overclocked to 4.2 GHz @ 1.3 V | ||||
| Motherboard | Asus P8Z77-V LX, LGA 1155, Chipset: Intel Z77M | ||||
| Networking | On-Board Gigabit LAN controller | ||||
| Memory | AMD Gamer Series Memory, 2 x 4 GB, 1866 MT/s, CL 9-9-9-24-1T | ||||
| Graphics | GeForce 210 1 GB DDR3 GeForce GT 630 512 MB GDDR5 GeForce GTX 650 Ti 1 GB GDDR5 GeForce GTX 660 2 GB GDDR5 GeForce GTX 670 2 GB GDDR5 GeForce GTX 770 2 GB GDDR5 GeForce GTX Titan 6 GB GDDR5 GeForce GTX 690 4 GB GDDR5 Radeon HD 6450 512 MB GDDR5 Radeon HD 6670 512 MB DDR3 Radeon HD 7770 1 GB GDDR5 Radeon HD 7790 1 GB GDDR5 Radeon HD 7870 2 GB GDDR5 Radeon HD 7950 Boost 3 GB GDDR5 Radeon HD 7970 3 GB GDDR5 Radeon HD 7990 6 GB GDDR5 | ||||
| Hard Drive | Samsung 840 Pro, 256 GB SSD, SATA 6Gb/s | ||||
| Power | XFX PRO850W, ATX12V, EPS12V | ||||
| Software and Drivers | |||||
| Operating System | Microsoft Windows 8 Pro x64 | ||||
| DirectX | DirectX 11 | ||||
| Graphics Drivers | AMD Catalyst 13.10 Beta 2, Nvidia GeForce 331.40 Beta | ||||
| Benchmarks | ||||||||
|---|---|---|---|---|---|---|---|---|
| Battlefield 4 Multiplayer beta | Custom THG Benchmark, 60-second Fraps run Map: Siege of Shanghai, Game Type: Domination | |||||||
Our first set of tests involves the Low detail preset and 1280x720. Unless you're gaming on a smaller screen using lower-end hardware, this isn't a particularly lush way to enjoy Battlefield 4. But we want to see how budget-oriented graphics hardware handles the game.

Sub-$50 cards like the Radeon HD 6450 and GeForce 210 get booted from this competition before it even begins. You'll need a Radeon HD 6670 DDR3, Radeon HD 6750, or GeForce GT 630 GDDR5 to manage playable frame rates at this setting. That's not good news for integrated graphics engines, except for higher-end options like AMD's A8/A10 APUs and Intel's Iris Pro 5200.

Charting frame rate over time illustrates the GeForce 210 and Radeon HD 6450 struggling, while the Radeon HD 7770 and GeForce GTX 650 Ti power through these settings to maintain at least 60 FPS through the test.


When it comes to frame time variance, the Radeon HD 6450 behaves poorly. While the GeForce 210 encounters fewer spikes, low frame rates make that a moot point. The rest of the cards fare well enough, almost always staying below 10 milliseconds of variance.
Next, we stick with the Low preset and raise the resolution to 1680x1050. Can the value-oriented Radeon HD 6670 DDR3 and GeForce GT 630 GDDR5 deliver a smooth experience, or will the extra load knock them out of contention?

Unfortunately, neither card averages more than 25 FPS, dipping as low as 20 FPS at times.

Looking at the frame rates over time shows that the Radeon HD 6670 DDR3 and GeForce GT 630 GDDR5 spend most of their time in unplayable territory. On the other hand, the Radeon HD 7770 and GeForce GTX 650 Ti have no problem keeping their noses above 45 FPS, usually exceeding 50.


The frame time variance displayed by all of the cards, aside from the 7770 and 650 Ti, climbs to a level we don't particularly approve of. Interestingly, the Radeon board is more consistent than the GeForce, though. Perhaps this has something to do with AMD's involvement with DICE leading up to this introduction, while Nvidia got its hands on the title to optimize its drivers more recently.
Now we shift from Low to High detail at 1680x1050. As mentioned, we're skipping the Medium preset entirely because it didn't give us any perceptible performance advantage compared to High.
Aside from the texture and model detail improvements we get in this transition, the High preset adds deferred anti-aliasing and HBAO ambient occlusion.

All of these cards maintain more than 30 FPS, but AMD's Radeon HD 7770 struggles more than the rest to stay above that boundary.

Frame rate over time charting shows that the Radeon HD 7770 shifts between 35 and 45 FPS over the course of the test. This is acceptable, even for a fast-paced shooter, though we naturally prefer more powerful solutions that don't drop below 40 FPS. The 7770's advantage is that it's the cheapest card on our chart by a significant margin.


Frame time variance hurts the Radeon HD 7770's story. It regularly pops above 15 ms using the High preset, which we've seen as noticeable latency in blind testing. You don't want numbers that high if you're sensitive to stutter. On the other hand, the rest of the solutions fare quite well.
Now we're pushing the resolution up to 1920x1080 and still using the High preset. We'd venture to say that most enthusiasts are using panels that natively support Full HD, so this becomes one of the best benchmarks in today's story. It tells you what you'll need in a gaming desktop to play Battlefield 4 using reasonably taxing settings.

Surprisingly, even the Radeon HD 7770 manages to keep its performance above 30 FPS. The Radeon HD 7790 doesn't push that minimum number much higher, but it does give us an average that exceeds 40 FPS. Nvidia's GeForce GTX 650 Ti demonstrates an even higher minimum frame rate. The Radeon HD 7870 and GeForce GTX 660 keep north of 50 FPS throughout the benchmark. So, those cards become a solid baseline for playing Battlefield 4 at the High preset and 1920x1080.

The frame rate over time chart shows us what happens at each step of the way during our test run, and, for the most part, everything checks out. There is some disturbing fluctuation from the Radeon HD 7790, though. We'll have to see if that's reflected in our frame time variance measurements.


Indeed, the Radeon HD 7790 suffers, along with AMD's Radeon HD 7770. We tried two different Bonaire-powered cards, both of which yield the same strange result. We don't know what is causing the high variance, but it's consistent and specific to the 7790.
The rest of the field keeps below 10 ms most of the time.
Finally, we dial the detail preset up to Ultra for a better idea of how high-end graphics hardware handles Battlefield 4. In addition to bumping up detail, this setting enables 4x MSAA, deferred AA, and HBAO ambient occlusion.

The Ultra setting incurs a significant performance penalty, and while the Radeon HD 7870 and GeForce GTX 660 manage to stay above 30 FPS, their performance is pretty marginal at this preset. We suspect that our frame time variance chart may not be particularly friendly to either card. For now, let's have a look at frame rate over time.

Our contenders land in a pretty tight grouping. The GeForce GTX Titan, 690, and Radeon HD 7990 don't dip below 60 FPS, though again, we'd really like to look at these numbers from an FCAT-based perspective.


Clearly, AMD's Radeon HD 7870 struggles to deliver frames at a consistent pace, and the Radeon HD 7950 Boost has a bit of an issue, too. It's interesting that the Radeon HD 7990 appears to fare better than Nvidia's GeForce GTX 690. We cannot test this theory using FCAT due to the tool's 32-bit limitation, however, the Fraps-based data does appear consistent with our real-world experience. Nvidia's dual-GPU board felt choppier, despite its high frame rate. This could be attributable to its 2 GB of memory per GK104 GPU. But even the GeForce GTX Titan encounters some spikes.
If you're lucky enough to be gaming on a QHD display, we're also including results from a 27" panel at 2560x1440.

Ratcheting up to 2560x1440 exacts a more substantial performance hit on all of these cards. It takes a Radeon HD 7970 or GeForce GTX 770 just to keep frame rates above 29 FPS. Even the once-powerful GeForce GTX 670 is pushed down to 26 FPS. And keep in mind that the newer GeForce GTX 760 performs very similarly to that card.

The Radeon HD 7990 and GeForce GTX 690 are finally pushed below 60 FPS at times.


Our frame time variance data shows the GeForce GTX 690 running into trouble, while we'd confidently attribute to its on-board memory. With just 2 GB per GPU, pushing higher resolutions at more demanding detail settings puts the dual-GK104 board at a distinct disadvantage to AMD's 3 GB per processor. Our in-game experience confirmed that something was hurting the Nvidia card.
DICE's recommended hardware list for Battlefield 4 includes quad-core Intel CPUs and hexa-core AMD processors, suggesting optimizations for threading that we typically don't see from games. So, let's see how different CPUs and architectures handle the game:

While we see a clear performance drop from the FX-4170 and Core i3-3220, frame rates remain comfortably above 40 FPS. That's great new, though if you own a GeForce GTX Titan, you probably wouldn't be enthused about that 20 FPS-drop attributable to too-slow of a CPU. This beautifully illustrates the importance of balance in a gaming PC build.
There's a noticeable speed-up going from FX-4170 at 4.2 GHz to the FX-8350 at 4 GHz. Although AMD doesn't fare as well as the Core i5s and i7s, we can at least speculate that Battlefield 4 does get some utility from a six- or eight-core CPU that you don't see from a quad-core chip.
It's also interesting that Sandy Bridge-E delivers the best minimum frame rate, despite a relatively slow operating frequency. That has to be attributable to lots of L3 cache or core count. Why the bunch-up around 74 FPS on average? That's the GeForce GTX Titan bottlenecking performance. We'd need a dual-GPU configuration to push Battlefield 4's frame rate higher.

The lower-end processors certainly encounter spikes in frame time latency. But keep in mind that this is the Ultra preset, and that a less demanding setting would likely help smooth some of those out.
The Battlefield franchise is back and better-looking than ever. No matter how you slice it, this is an iterative release. But that's the way a lot of franchise fans want to see things happen. However you felt about Battlefield 3, you'll probably feel the same way about Battlefield 4.

You don't need a monster graphics subsystem to play it, but low-end cards like the Radeon HD 6450 and GeForce 210 just aren't fast enough. At least spend the money on a Radeon HD 6570/6670 DDR3 or GeForce GT 630 GDDR5 to play this game using low details at 1280x720. Higher resolutions necessitate a Radeon HD 7770 or GeForce GTX 650, and that'd only be for 1680x1050.
Doctor Jones, oh please wake up!
If you really want to step up to Battlefield 4 at its High detail preset, the GeForce GTX 650 Ti or Radeon HD 7850 should be sufficient for 1680x1050. More enthusiasts have 1080p displays though, and we're recommending a $200 Radeon HD 7870 or $180 GeForce GTX 660. For those of you who just caught Angelini's AMD Radeon R9 280X, R9 270X, And R7 260X: Old GPUs, New Names, make that an R9 270X for at least $200.
Boom.
How about the most hardcore enthusiasts? The demanding Ultra preset compels you go go with a Radeon HD 7970/R9 280X or GeForce GTX 670/760 at 1920x1080. Stepping up to QHD means you don't want anything less than a 7970 or GTX 680/770, though an even faster graphics subsystem is recommended.

The good news is that a fairly inexpensive Core i3 or FX-4000-series CPU could be fast enough to handle the High preset without capping graphics performance. If you want to play at the Ultra preset, grab a Core i5 or FX-6000-class processor at the very least. Of course, we realize the irony: nobody with a $1000 GeForce GTX Titan goes cheap on a mid-range CPU. Chances are good that if you want to enjoy Battlefield 4 in its full glory, you're doing it on a beast of a machine. More power to you!
This game doesn't need the best of the best for playable performance, but it does want a capable gaming machine. We look forward to the commercial release at the end of the month, at which point we'll be revisiting it.