Ultimate Battlefield 4 Guide hosted by EVGA
By MANIAC-VVV- and THEBSKI

 
            
 

This GUIDE is currently under construction
Please be patient
 
----------------------------------------------------------------------------
 
This Guide is intended to help users understand the many aspects of setting up and playing Battlefield 4 using Windows 7 -and- Windows 8.1
Covering all aspects of Battlefield 4 in both single and surround monitor setups using Nvidia graphics cards.
 
This will be my third Battlefield guide posted here at EVGA and I would like to personally thank EVGA forum member THEBSKI for partnering with me to provide the Surround configuration datasets and for his awesome graphing skills used throughout this guide.
 
I would also like to thank EVGA for hosting these Guides and providing such a great user support experience here at the EVGA forums!
*DISCLAIMER*
MANIAC-VVV- and THEBSKI are simply regular forum members helping other users
Any suggestions, settings or configuration information contained in this Guide, may or may not be applicable to your hardware, graphics card(s), system drivers or installed operating system. 
-->As such, any actions taken based on information in this guide... is at the users own risk 

Reserved

Reserved

Reserved

Alright guys, let's get this started! I'm going to be trickling performance data for 5760x1080 in over the next few days. This post will just introduce my system, my methods, as well as an outline for what I have planned as well as links to sections once they are completed.
 
Test System Setup:
 
Intel Core i7-3770K at 4.8 GHz
Asus Maximus V Extreme
8 GB G.Skill Trident X at 2400 MHz 9-11-11-31
EVGA GTX 780 Classified SLI at 1202 Core - 6200 Memory
Corsair AX860
Glossy Asus VG248QE's at 5760x1080 144 Hz
 
CPU Screenshot

 
Memory Screenshot

 
GPU Screenshot

 
Driver Version

 
Operating System

 
System Launch Condition

 
Maniac will cover this in more detail, but system launch condition appears to be very important in determining how well this game runs, even more than BF3. The screenshot above shows my processes right before I click the play button.
 
Internet Connection

 
Full Server Test Methods
 
Full server testing can obviously be a challenge as there are many variables. I've done the most that I can do to ensure runs are as comparable as possible. All servers tested in had 40 ping or lower. All test runs were completed in a server with 60 or more players in the game. Each server had to pass the "feel" test before I would record data in it. As many of you probably have noticed, there seem to be a couple different classes of servers right now. The better servers are just butter smooth and run the game quite well. The worse servers have a rubber bandish or stutterish feel to them despite low pings and great frame rates. Those servers were avoided for data collection.
 
All GPU information was recorded via Precision X 4.2.1 with the exception of FPS data. As many have noticed, RivaTuner's Statistics Server is not compatible with x64 programs, therefore no one is probably having much success with an OSD in BF4, and we certainly haven't had any success getting FPS data with it either. To gather FPS data, I turned to FRAPS. One thing that should be noted about Battlefield Online is that the frames per second produced in the spawn screen are often times quite a bit different than those produced in the actual game play. Because of this, I've done my best to not present the data where I died in the game. I wanted this data to be actual game play data.
 
To start the round, I would record all data using Precision X for the entirety of the round. For FPS data, I would start FRAPS when I spawned and stop the recording when I died. I would do this until I got a run long enough that I was satisfied with the amount of data to look at. Due to this, the FPS runs across the maps are not all of the same length. Some runs are longer than others. For the VRAM data, I present the data throughout the entirety of the round as VRAM usage is unaffected by spawn screens for the most part.
 
Empty Server Test Methods
 
Under Construction
 
Performance Comparisons
Ultra Preset - No FXAA (All Maps) FPS Data
Ultra Preset - No FXAA (All Maps) VRAM Data
High Preset (Best and Worst Performing Map)
Preferred Custom Preset (Best and Worst Performing Map)
PCI-E 2.0 vs. PCI-E 3.0 (Best and Worst Performing Map)
CPU Clock Speed Scaling (Best and Worst Performing Map)
System Memory Speed Scaling (Best and Worst Performing Map)
Hyper Threading vs. No Hyper Threading (Best and Worst Performing Map)
6 Core vs. 4 Core (Best and Worst Performing Map)
Individual In Game Settings Impacts (Empty Server Testing)
Detailed Performance Analysis of Each Map (Empty Server Testing)
 
It will take some time to compile all of this, but I hope to get a section up every day or two. As much as I wish this could be my full time job, it is not, so bear with me a little bit. 

 
Ultra Preset - No FXAA FPS Data
 
 
This section will cover performance for Battlefield 4 64-player online multiplayer for all 10 maps on the Ultra Preset with FXAA turned off at a resolution of 5760x1080. The test setup and method details can be found here.
 
Here is an in-game screenshot of the exact settings used to collect all of this data.

 
Notice I have outlined Antialiasing Post (FXAA) in red as it is the only setting changed from the Ultra Preset. I have changed it from High to off simply because the Ultra Preset runs 4x MSAA anyways and running FXAA over MSAA degrades the image quality.
 
I am now going to show each time series of FPS data for the different maps and then have a little discussion in the end.
 
Dawnbreaker

 
Flood Zone

 
Golmud Railway

 
Hainan Resort

 
Lancang Dam

 
Operation Locker

 
Paracel Storm (around Base A, running to and ending at Base B)
.
 
Paracel Storm (around Base D)

 
Rogue Transmission

 
Siege of Shanghai

 
Zavod 311

 
The first thing I'd like to comment on in general is the consistency of performance at 5760x1080 that the game engine can provide. This is quite different from what we saw in Battlefield 3. In Battlefield 3, for example, it was not uncommon to see frame rates in the 50's at one part of the map and in the 120's on another part of the map. Those extremes, for the most part, seem to be gone in Battlefield 4, and it really makes for a great gaming experience.
 
I think this is due to primarily two reasons. Number one, I don't have near the graphics overhead relative to BF4 performance compared to BF3. This doesn't allow the CPU bound areas to come into play as much. Number two, thread balancing, anyone? Holy smokes, did they nail that with this game, or what? I have not yet tested in Windows 7, but in my 8.1 testing I see extremely even usage across all 8 threads on the 3770K. I have seen sustained CPU usages in the 90's and actually saw it hit 99% usage at one instant. I will cover CPU stuff in more detail in another section, but the bottom line is that Battlefield 4 seems to make much better use of the CPU which likely allows for more consistent performance.
 
As far as analyzing the graphs, the min, max, and average are nice, but given that this is online multiplayer, those particular numbers would likely be difficult to reproduce exactly. My intent with these graphs was to present as long of a time series as I could provide to show as much performance data as possible. I also tried to look many different directions and cover multiple bases to include as much variation as possible. What I'm looking for is problem spots. How does the performance maintain as the game plays? There are a couple spots in about 3 different maps where the data show what I would call a real performance drop. I'm talking dropping from 65 FPS to 40 FPS. There are 5-10 FPS fluctuations all over, but that is to be expected. Overall, performance is just very consistent and the maps perform very similarly.
 
Ultra Settings would not be my preferred settings with my setup because I like frames to be a little higher than 60ish for online gaming, but I was pretty amazed at how well these settings performed for a game that just launched.
 
Intro/Test Setup
Ultra Preset - No FXAA VRAM Data

 
Ultra Preset - No FXAA VRAM Data
 
 
This section will cover performance for Battlefield 4 64-player online multiplayer for all 10 maps on the Ultra Preset with FXAA turned off at a resolution of 5760x1080. The test setup and method details can be found here.
 
Here is an in-game screenshot of the exact settings used to collect all of this data.

 
Notice I have outlined Antialiasing Post (FXAA) in red as it is the only setting changed from the Ultra Preset. I have changed it from High to off simply because the Ultra Preset runs 4x MSAA anyways and running FXAA over MSAA degrades the image quality.
 
I am now going to show each time series of VRAM data for the different maps and then have a little discussion in the end.
 
Dawnbreaker

 
Flood Zone

 
Golmud Railway

 
Hainan Resort

 
Lancang Dam

 
Operation Locker

 
Paracel Storm

 
Rogue Transmission

 
Siege of Shanghai

 
Zavod 311

 
Dice did a nice job here making the game fit perfectly within a 3GB envelope with really no issues. The max memory used was exactly 3000 MB, and I have seen the cards use higher so I know they can go a little beyond 3000. At this point I have no VRAM issues to report on full Ultra settings with 3 GB GPU's! Excellent news!
 
Intro/Test Setup
Ultra Preset - No FXAA FPS Data

Reserved

Reserved

Reserved

Reserved

Reserved

Reserved

Reserved