BF4 - New Multiplayer Benchmarks

Joined
Dec 22, 2007
Messages
987
Some good info in this review.

If the 1440p VRAM use is accurate WOW. (Note: The two pics below are a comparison between Beta & Retail)

BF4-1920x1080-ultra-settings-gpu-ram-usage.jpg


Bf4-2560x1440-Ultra-settings-GPU-ram-usage.png
 
Last edited:
Holy shit...so I'm almost guaranteed to need 4gb or more for my surround setup
 
Doesn't leave much room for downsampling in my 1440p GTX 780 SLI setup.

That said, I'll be happy if I can just run the game without stuttering and crashing. It looks incredible even without downsampling.
 
Yes, BF4 does use that much ram for 1440p. Im hitting 2.6gb as well @ 1440p.
 
That site probably has the worst benchmark setup I have ever seen.


Doesn't leave much room for downsampling in my 1440p GTX 780 SLI setup.

That said, I'll be happy if I can just run the game without stuttering and crashing. It looks incredible even without downsampling.

Does ultra include 4xMSAA? If so bare in mind when downsampling you wont be using MSAA as well.
 
Well, considering that 1440p is using 2.6 GB VRAM @ ultra I would say you need a minimum of 4GB!

How is that? Ultra does include 4x MSAA. Why do I need a minimum of 4gb? BF4 has the highest ram usage Ive seen so far. If only hitting 2.6 on Ultra w/ 4xMSAA, whats the need of 4gb at this point in time?
 
How do you get an accurate ram usage number with BF4? Is Afterburner accurate with BF4 right now? I was thinking it wasn't because it gives me 4gb usage with my xfire 7970 setup at 1600P :\
 
The MSAA adds a lot to VRAM usage. I can't image needing 4x @ 1440p.
 
How is that? Ultra does include 4x MSAA. Why do I need a minimum of 4gb? BF4 has the highest ram usage Ive seen so far. If only hitting 2.6 on Ultra w/ 4xMSAA, whats the need of 4gb at this point in time?

His statement was in response to mine
 
I'll try GPU-Z tonight, but Afterburner last told me 4.2GB @ 1600P and zero AA usage.

64 man conquest server on the satellite map
 
How do you get an accurate ram usage number with BF4? Is Afterburner accurate with BF4 right now? I was thinking it wasn't because it gives me 4gb usage with my xfire 7970 setup at 1600P :\

On crossfire setups after ads the usage of both cards. So its actually 2gb per card instead of 4 gb its kind of stupid but how its worked for a long time.
 
Yeah, I get 4+gb usage on my xfire 7950 at 1600P. Max I hit was 4.9GB with 4xMSAA
 
Using GPU-Z, I see >1800MB usage on each GTX670 2GB card at 2560x1600. I'm able to run mostly ultra settings and 2xMSAA without feeling too much hurt, mostly locked at 60FPS outside of hickups. Game's smoother than BF3.
 
Why does the 4770K perform worse than the 4670K? I was going to buy a 4770K now I'm not sure.
 
Why does the 4770K perform worse than the 4670K? I was going to buy a 4770K now I'm not sure.

Wait for 'real' reviews, but note that Hyper-Threading is being made ever more relevant due to the presence of highly-threaded CPUs in the new consoles. Every game that runs on them will have to be very well threaded.
 
How is that? Ultra does include 4x MSAA. Why do I need a minimum of 4gb? BF4 has the highest ram usage Ive seen so far. If only hitting 2.6 on Ultra w/ 4xMSAA, whats the need of 4gb at this point in time?

The comment I was responding to referred to a SURROUND setup.

If a single 1440p display is using 2.6 I would imagine a surround setup would be significantly higher.
 
Measuring RAM like this is never accurate.

All you are measuring is RAM allocation. This says nothing about how much RAM is actually needed or actually being utilized by the application. Ask any software developer this.

Just because at some settings the RAM usage shows 2.2GB does NOT mean that a 2GB card would get any less performance. It may have allocated 2.2GB, but it may only need to use 1.8GB to avoid swapping (only the application itself is in any position to say how much RAM is needs to be using). As long as it doesn't need to actively swap much, performance will be negligibly impacted.

Most software is written to allocate more RAM if more is simply available whether or not it truely "needs" it or not. It may be able to "use" it, but it may not be able to use that extra RAM in a way that would really increase performance over a setup where it had to allocate less because less was available.
 
Measuring RAM like this is never accurate.

All you are measuring is RAM allocation. This says nothing about how much RAM is actually needed or actually being utilized by the application. Ask any software developer this.

Just because at some settings the RAM usage shows 2.2GB does NOT mean that a 2GB card would get any less performance. It may have allocated 2.2GB, but it may only need to use 1.8GB to avoid swapping (only the application itself is in any position to say how much RAM is needs to be using). As long as it doesn't need to actively swap much, performance will be negligibly impacted.

Most software is written to allocate more RAM if more is simply available whether or not it truely "needs" it or not. It may be able to "use" it, but it may not be able to use that extra RAM in a way that would really increase performance over a setup where it had to allocate less because less was available.

Using 1.8 vs allocating 2.2 is a moot point.

2GB VRAM is not enough any more unless you like buying something that just barely has the specs to run one of today's biggest titles.
 
Using 1.8 vs allocating 2.2 is a moot point.

2GB VRAM is not enough any more unless you like buying something that just barely has the specs to run one of today's biggest titles.

I can say that in my 30 hours so far of BF4 running at ultra settings at 1080p in 64 player servers that my 2GB GTX 680 has never stuttered or dropped below 50fps yet and it rarely goes under 60.

So while the test here seems to suggest you "need" a 3GB card because it allocates more than 2GB at 1080p, my actual results show that the game does not actually need more than 2GB to run perfectly and not stutter from swapping.
 
at 1440P with 2xmsaa ultra settings, post-aa off, resolution scale to 125% I was seeing about 2.6GB vram usage.
 
Also keep in mind folks that on single GPU setups PCIe can stream/copy on-demand roughly 100-200MB per-frame (assuming 60fps target). Testing has shown that a single PCIe card can "add" about 500MB of main memory before it becomes bandwidth-starved, so even if the game is using the full 2.5GB (it may not be), it's not holding the card back.

CFX or SLI users, however... :(
 
Also keep in mind folks that on single GPU setups PCIe can stream/copy on-demand roughly 100-200MB per-frame (assuming 60fps target). Testing has shown that a single PCIe card can "add" about 500MB of main memory before it becomes bandwidth-starved, so even if the game is using the full 2.5GB (it may not be), it's not holding the card back.

CFX or SLI users, however... :(

This is interesting if true. It could explain my results. I do have my GTX 680 in a 16x PCIe 3.0 slot and have its RAM overclocked to 7GHz.
 
On my 7950GB Crossfire, I am getting about 70-80FPS but I do get dips to 40s every so often. Perhaps due to my 4670k being stock speeds? (3.4Ghz)

I was on Beta 7 while playing this. I'll see how Beta 8 works today.
 
On my 7950GB Crossfire, I am getting about 70-80FPS but I do get dips to 40s every so often. Perhaps due to my 4670k being stock speeds? (3.4Ghz)

I was on Beta 7 while playing this. I'll see how Beta 8 works today.

Crank that sucker!
 
These numbers are about spot on. I did some rudimentary benchmarking of my own, to see what I could tweak and to check in general how the game was running. Also worth noting, after this I upped my memory speed to 1866, and it increased my processor utilization to an average over 80. Here are my results:

rZlJSxb.png


This was after upping my memory speed:

A1S0BUe.png
 
I had down clocked everything before starting BF4 so I could rule out the system being the cause of any errors if they did occur. And yes, this data was gathered on 64 player official servers. I was surprised at how much the system memory speed increased the processor utilization (as you see above). I guess in theory it makes sense. Here are the settings:
3ULVyxn.jpg
 
Looks like the 100% CPU utilization has been fixed since I played on release day.

Looking forward to getting a new GPU very soon to do some benchmarking of my own.

HD3000 is just not cutting it and I'm losing my marbles waiting for the current price/performance war to settle.
 
Ah yep, I hear ya, I'm actually an Nvidia converter for the moment... I go back and forth. However, the idea behind Mantle, and the fact that it will be utilized in BF4 was my selling point. I believe I read an article today that a few more game companies hopped on the Mantle bandwagon. Only time will tell to see how it turns out, but I'm optimistic. It could be a real tide turner, especially if consoles utilize it.... Again, I mainly play BF4 so that is what I am looking towards.
 
heh and people thought that 6 and 12GB cards where crazy talk
not much now
 
I don't understand, I've never seen BF4 use more than 1700mbs of System Memory but this site and others claim they are using 7 GBs of System Ram. I play at 2560x1440 too and I had the windows 8.1 task manager running for 4 hours last night. Never even came close to 2gbs, let alone 7gbs!
 
I don't understand, I've never seen BF4 use more than 1700mbs of System Memory but this site and others claim they are using 7 GBs of System Ram. I play at 2560x1440 too and I had the windows 8.1 task manager running for 4 hours last night. Never even came close to 2gbs, let alone 7gbs!

Windows will allocate memory wildly based on usage patterns and memory capacity. I have 16GB, and I've seen over 8GB used while playing, but I also have over 20 Chrome tabs open along with other things. The game doesn't use much RAM, even in 64bit guise.

As for VRAM usage- it's hard for me to tell. It's between 1.8GB and 2GB regardless of what settings I choose, but I did notice that going to 150% resolution scaling caused regular hard pauses that killed gameplay.
 
Back
Top