BF3 64 player MP CPU benchmarks!

Thanks Frito for the real MP benchmarks, your initiative is much appreciated.
There are so many threads about the CPU aspect of the game, but without a proof or full of B.S.
P.S. - you deserve a medal for this j/k :D
 
Frito said:
its not tbh. many people with i7's are reporting a studdering problem with HT enabled and are having to disable HT in their BIOS to stop it from happening. apparently DICE did not code the engine with HT in mind just multi core cpu's and having hyperthreading on causes strange performance problems. i imagine they will probably address this in a future patch but at the end of the day HT has never had any positive impact in any games so unless you want it for some non-gaming reason i would say get a 2500k
Click to expand...

i think it has more to do with how it uses the processor. if its heavy L2 cache dependent it could be causing the stuttering issue because the processor is constantly having to balance the L2 cache usage between the threads. thats all a slightly educated guess though. even BFBC2 didn't support hyperthreading though it also didn't have the stuttering issue because it forcefully ignored the extra threads.
 
Great thread; CPU dependency in BF3 is pretty much like a secret. I saw a post or two on the BF3 forums and it was night & day when I overclocked my i7-930 (especially with 3 screens).

As for Hyperthreading, as mentioned, normally it actually causes stuttering issues in BF3, also. People have reported that turning off core parking in the registry smooths out BF3 and lets you leave HT on. My rig went unstable when I tried it, and I haven't sorted it out to report first-hand. Core parking discussion: http://www.xtremesystems.org/forums/showthread.php?277129-Core-parking-on-Windows-Seven (maybe there's a discussion about it here, I haven't noticed one)
 
The fact it can be turned off, or maybe even worked through/around, makes me want to go 2600/HT for future proofing, or the chance they smooth things out with updates, not to mention if any other game(s) come out on the next few years that take better advantage of HT. The extra $ wont kill me, so it seems the best choice at the moment.


Edit: I found some numbers, of course bf3 is not in the lineup :mad:
http://forums.anandtech.com/showthread.php?t=2207167
 
Last edited:
I m curious if a 6 core i7 would help over a 4 core. I have a 6 core i7 970 and when I turn off HT I notice that all 6 cores are being used but with HT on not all 12 threads are being used (I can only tell if the 1st, middle, and last cores/threads are being used in real time) so it uses somewhere between 6 and 11 threads at a time. I would do a benchmark for you guys but I am GPU bottlenecked currently (GTX 260 at 1900x1200) but will be upgrading to a 6950 2GB shortly and then I might be able to. I do have the ability to turn off CPU cores in BIOS so I could test 8 and 10 threads as well. I personally cant tell a difference between HT on and HT off but that might be because of by GPU bottleneck.
 
3 things I like about this:

1 - You took the time to do it
2 - You're using a GPU configuration applicable most users (i.e. not 3 x 580's)
3 - Proves a point that I and others have been making, specifically that CPU power does matter

Now I'm curious to see what happens if someone does the same with other CPU configurations.
 
I sold my 2600k and bought a 2500k and didnt notice any fps difference. Although im only running 1 6970 (950/1450)

Maybe Ill test out 4vs 4.4 vs 4.8 on the same server. I can use Control Center to clock the cpu up and down.
 
Good info. Another overlooked aspect is RAM.....I see 5gb to 7gb RAM being used during 64 player maps.
 
[21CW]killerofall;1038096208 said:
I m curious if a 6 core i7 would help over a 4 core. I have a 6 core i7 970 and when I turn off HT I notice that all 6 cores are being used but with HT on not all 12 threads are being used (I can only tell if the 1st, middle, and last cores/threads are being used in real time) so it uses somewhere between 6 and 11 threads at a time. I would do a benchmark for you guys but I am GPU bottlenecked currently (GTX 260 at 1900x1200) but will be upgrading to a 6950 2GB shortly and then I might be able to. I do have the ability to turn off CPU cores in BIOS so I could test 8 and 10 threads as well. I personally cant tell a difference between HT on and HT off but that might be because of by GPU bottleneck.
Click to expand...

With that said... I'd like to see something similar with BD....
 
My 6950 2GB just shipped so I should be able to do a benchmark sometime in the next two weeks, depending on shipping times, I don't have a tracking number yet (may not get one). I will try at low settings and go all the way up to ultra to see if it makes a difference. BTW I have 12 GB of 1600 ram in triple channel so that shouldn't be a bottleneck in my benchmarks. I also have the game installed on an Intel 320 120GB SSD so that also shouldn't be a bottleneck. I can also try overclocking to ~4GHz to see if that makes a difference.
 
sirmonkey1985 said:
i think it has more to do with how it uses the processor. if its heavy L2 cache dependent it could be causing the stuttering issue because the processor is constantly having to balance the L2 cache usage between the threads. thats all a slightly educated guess though. even BFBC2 didn't support hyperthreading though it also didn't have the stuttering issue because it forcefully ignored the extra threads.
Click to expand...

I have an i7-970 and i can tell you BFBC2/BF3 both support HT.
This screen is from aida and BFBC2.

bfbc2aida.jpg


jebo_4jc said:
Good info. Another overlooked aspect is RAM.....I see 5gb to 7gb RAM being used during 64 player maps.
Click to expand...

You are right. I had problems with texture pop-in and upgrading my ram solved.

@Frito Thanks for the results.
 
Frito, awesome job.

I think, to run in my 2 cents, however, that there is a point where you have enough cpu. For the past couple days I switched to a 6990 that I have from my 590 (and endured the noise). The 6990 is solidly 20% faster or so at 2560 everywhere I go just from eyeballing my frames in fraps, which I do not think would be the case if I was cpu limited as you've also discovered by running higher overclocks. I've noticed my cpu at 4.8ghz is typically 80% or so across all cores.

It would be interesting, if someone with a 6 core i7 or phenom, would be kind enough to try say 2.5-3ghz 4 core vs 6 core.


coolpurplefan said:
I wouldn't mind seeing if people can play offline with 1680x1050 at medium settings with cheaper hardware like a Core 2 Duo 8400 and GTX 550 Ti.
Click to expand...

The best player in my clan, and MVP in most rounds we play, plays on a e5200 overclocked to 2.6ghz and a 9600gs overclocked by about 20% at lowest settings and resolution...so you should be more than good to get in the game.
 
Last edited:
The issue that users were reporting with 2600ks and HyperThreading enabled have already been taken care of by dice in the last 2 patches and the nVidia Drivers as wells have fixed all stuttering issues. That is unless you hit a Vram wall, in which case briniging Anti Aliasing down to 2x will fix all that up for you.
 
great info i was worried my i5 at only 3.4Ghz may not be up to the game but seems it is my 6850 that will get eaten alive.
 
QuiteSufficient said:
Frito, awesome job.

I think, to run in my 2 cents, however, that there is a point where you have enough cpu. For the past couple days I switched to a 6990 that I have from my 590 (and endured the noise). The 6990 is solidly 20% faster or so at 2560 everywhere I go just from eyeballing my frames in fraps, which I do not think would be the case if I was cpu limited as you've also discovered by running higher overclocks. I've noticed my cpu at 4.8ghz is typically 80% or so across all cores.

It would be interesting, if someone with a 6 core i7 or phenom, would be kind enough to try say 2.5-3ghz 4 core vs 6 core.


The best player in my clan, and MVP in most rounds we play, plays on a e5200 overclocked to 2.6ghz and a 9600gs overclocked by about 20% at lowest settings and resolution...so you should be more than good to get in the game.
Click to expand...

oh yeah there is a point where you have enough CPU but that point will vary from person to person based on personal preferences on how high they like/want their frame rates to be and what resolution they are running, higher resolution will require more CPU power to reach the same frame rates ofc.

for me and in my opinion with my rig at 1920 x 1080 my i5 at stock clocks is close to fast enough but its not quite enough to keep multiplayer over 60 fps 100% of the time, an easy overclock to 4.2ghz will make that diff and anything beyond that is just added fps headroom.
 
jebo_4jc said:
Good info. Another overlooked aspect is RAM.....I see 5gb to 7gb RAM being used during 64 player maps.
Click to expand...

what resolution are you running?

i did pay attention to system RAM use on my rig in 64p multiplayer and it seemed to stay around 1.8-2 gb (total system ram used was 4-4.5 gb but thats got little to do with what the game is taking)
 
I'm currently running 4 gb of ocz platinum @1066 in dual channel mode. I have 2 extra 1gb sticks of corsair 800 ddr2 value ram laying around. Do you guys think I would benefit from 6gb of ram not in dual channel over 4 in dual channel? I'm currently ocd to 3ghz from my understanding adding more ram to open dimms puts a bigger stress on my northbridge, yes? Would you guys recommend it?
654.jpg

Here's my ram usage on 64 player strike at karkand on all high settings.
 
If you pit those two 1gb sticks in, you will still be in dual channel.

Give it a try
 
Frito said:
for those who are wondering why i have so much interest in this its because when i first got the game my system was a C2Q Q8200 overclocked to 2.8 ghz with 4gb of ram and 1 560 Ti, due to reviews on the net stating this game is not CPU depentant i bought another 560 Ti only to find out it was my CPU holding my GPU back even though it was a quad core. although i can no longer benchmark that CPU because i gave it to my brother i can tell you without a doubt that even my i5-2500k crippled down to basically what amounts to an i3 running at 3.4 ghz without HT the performance i experienced running that test above is still well above what my core 2 QUAD was giving me with the same 2 video cards in SLI.
Click to expand...

First off, good job!

Now the question is, do you get better performance increase from a GPU upgrade or a CPU upgrade? Say you still had that C2Q 8200 and you had an older 400 series GPU, would it be better to upgrade the CPU and MOBO, or the GPU?

I am also curious how much performance increase there is going from just the one 560Ti to 2 560Tis?

As for others that have commented on why test BF3 in single player and not multiplayer, that is easier to answer: It is much harder to produce similar play throughs with multiplayer. When you bench, you want as much as possible to be the same between tests. Since multiplayer is much more fluid and has more variables than single player, it is not the best way to benchmark. But I have seen many many posts saying BF3 multiplayer is completely different than BF3 single player, so I don't know that it is a "secret".
 
NoOther said:
First off, good job!

Now the question is, do you get better performance increase from a GPU upgrade or a CPU upgrade? Say you still had that C2Q 8200 and you had an older 400 series GPU, would it be better to upgrade the CPU and MOBO, or the GPU?

I am also curious how much performance increase there is going from just the one 560Ti to 2 560Tis?

As for others that have commented on why test BF3 in single player and not multiplayer, that is easier to answer: It is much harder to produce similar play throughs with multiplayer. When you bench, you want as much as possible to be the same between tests. Since multiplayer is much more fluid and has more variables than single player, it is not the best way to benchmark. But I have seen many many posts saying BF3 multiplayer is completely different than BF3 single player, so I don't know that it is a "secret".
Click to expand...

I think in my case a CPU/mobo/ram upgrade would be key. So far this card has been a champ. I can play this game on high settings with 40+ fps on anything 32 player or less. Put it this way. If I run around on strike at karkand by myself my fps doesn't drop below 0. Throw 32 more players and I'm.hovering 20's. My CPU is the obvious culprit here. I think the real question is how much upgrading do I need to do to Max out my gpu load. Hmm...
 
Let me start off by saying: good work.

Frito said:
for those who are wondering why i have so much interest in this its because when i first got the game my system was a C2Q Q8200 overclocked to 2.8 ghz with 4gb of ram and 1 560 Ti, due to reviews on the net stating this game is not CPU depentant i bought another 560 Ti only to find out it was my CPU holding my GPU back even though it was a quad core. although i can no longer benchmark that CPU because i gave it to my brother i can tell you without a doubt that even my i5-2500k crippled down to basically what amounts to an i3 running at 3.4 ghz without HT the performance i experienced running that test above is still well above what my core 2 QUAD was giving me with the same 2 video cards in SLI.
Click to expand...

Sadly, a lot of review sites don't understand computing hardware.

For example, lets say the review runs with a single GTX 580 (top of the line). There may very well be a minimal amount of difference at 1080p Ultra settings between the various quad-core CPU's. Then the reviewer calls the game "non-cpu" dependent. But this would be false because if you have something more powerful than a GTX 580 (like GTX 560 SLI or Radeon 6950 Xfire), then you could be back to cpu-bound again. That's another reason why low resolution performance is important. Even though X and Y CPU perform identically with GPU A at the game settings/resolution you play on, making GPU A into SLI/Xfire might totally change that picture.

It also doesn't help that gamers have been fed the whole games are not CPU dependent bit for years now. This was mainly because (imo) we've had nothing but console ports for years.
 
dexvx said:
It also doesn't help that gamers have been fed the whole games are not CPU dependent bit for years now. This was mainly because (imo) we've had nothing but console ports for years.
Click to expand...

I think this is often misconstrued. It isn't that games are not CPU dependent, it is that most games are more GPU dependent than CPU dependent. Performance of games generally scales better with GPU upgrades rather than CPU upgrades. And upgrading a GPU is typically cheaper than doing a CPU/Mobo upgrade.
 
rushaoz said:
I think in my case a CPU/mobo/ram upgrade would be key. So far this card has been a champ. I can play this game on high settings with 40+ fps on anything 32 player or less. Put it this way. If I run around on strike at karkand by myself my fps doesn't drop below 0. Throw 32 more players and I'm.hovering 20's. My CPU is the obvious culprit here. I think the real question is how much upgrading do I need to do to Max out my gpu load. Hmm...
Click to expand...

I think you are GPU limited there as well with a 460 GTX. I actually played with my 460GTX OC'd on my i7 920, and then with my i7 2600k OC'd. I was still getting in the mid 40s FPS. When I got my 6950, my FPS jumped quite a bit.
 
NoOther said:
I think you are GPU limited there as well with a 460 GTX. I actually played with my 460GTX OC'd on my i7 920, and then with my i7 2600k OC'd. I was still getting in the mid 40s FPS. When I got my 6950, my FPS jumped quite a bit.
Click to expand...

Sorry my post was screwed up and I didn't realize it. I meant to say that if I run around karkand bymyself my fps doesn't drop below 40. Nor on any other map. And karkand is actually the worst for my rig. Like I said I can play on high on any 32 player map and not drop below 40, except Karkand. So basically I need a new CPU to keep my gtx 460 above 40 fps. This game cripples this q6600. Whoever said games aren't CPU dependant were WRONG!! LOL
 
rushaoz said:
Sorry my post was screwed up and I didn't realize it. I meant to say that if I run around karkand bymyself my fps doesn't drop below 40. Nor on any other map. And karkand is actually the worst for my rig. Like I said I can play on high on any 32 player map and not drop below 40, except Karkand. So basically I need a new CPU to keep my gtx 460 above 40 fps. This game cripples this q6600. Whoever said games aren't CPU dependant were WRONG!! LOL
Click to expand...

what's your clock speed? Sorry if I missed it, I'm on my phone
 
Hi Guys,

I wanted to share my experience with you regarding this debate of GPU vs CPU.

Old specs:
Core 2 duo 3.0ghz 8400
4 gb ram
9800 GTX+

On those specs I was able to run BF3 on medium without a problem. But of course I wanted to get more out of the game so I began by upgrading my video card. I got a 6950.

What I noticed was that I was getting 30 fps on ultra and 35 to 40 on high but overall the game wasn't really playable. Too much lagging, stuttering. My CPU was running 100% and my ram was going into virtual ram. So I felt strongly that my CPU was chocking my system up if I ever wanted to play this game on ultra.

So, yes I upgraded to a i5 2500k with 8 gigs of ram and voila!! My god what a difference! I am now playing the game at 50 to 60 fps. No hiccups and smooth on ULTRA.

I did spend a good amount of money on the upgrades but I am another proof that CPU matters. My case was easy because I moved from a duo to a quattro but still it matters.

Also, the 6950 is a great card (mine is overclocked to 900mhz) and it can play ULTRA unlike what other people are saying. I dont have it maxed completely however but perfect enough to get ultra quality.

So Frito, thanks for sharing this important thread because this debate has to stop and it has to be known that CPU matters. For that case, i would say RAM too because my RAM during game play it jumped to over 4 gbs.

Good luck to all.
 
NoOther said:
I think this is often misconstrued. It isn't that games are not CPU dependent, it is that most games are more GPU dependent than CPU dependent. Performance of games generally scales better with GPU upgrades rather than CPU upgrades. And upgrading a GPU is typically cheaper than doing a CPU/Mobo upgrade.
Click to expand...

This. If you're after a consistent 60fps then you'll be fine with a relatively old quad at stock in BF3. My i7 920 is still at stock speed and I haven't needed to overlclock it. FRAPS says 60 for me. I have 580SLI and I'm waiting for faster GPU's to increase my resolution from 4040x720 to 6060x1080. There is no game so far that is forcing me to upgrade my mobo/ram/cpu.

I only play 32 player though but that's more a limitation of the very small maps.
 
nice to see someone doing CPU benchmarks but OP Metro is the worst possible map to test it on. Caspian Border x64 is pretty much the standard CPU benchmark now.
 
Finkus said:
This. If you're after a consistent 60fps then you'll be fine with a relatively old quad at stock in BF3. My i7 920 is still at stock speed and I haven't needed to overlclock it. FRAPS says 60 for me. I have 580SLI and I'm waiting for faster GPU's to increase my resolution from 4040x720 to 6060x1080. There is no game so far that is forcing me to upgrade my mobo/ram/cpu.

I only play 32 player though but that's more a limitation of the very small maps.
Click to expand...

This is true. I can play single player on ultra. I have to run half my shit on medium for playable fps
 
I did a couple run-through's of 64 player servers (Metro/Karkand) to show you guys some CPU/GPU characteristics of the FX8150 with 6970 CFX. As you can see I get great scaling/usage with my 6970's.
Before the FX chip I was using a Phenom 2 970BE @ 3.9Ghz and I saw anywhere from 45-55% GPU usage in those same scenarios and felt extremely sluggish (literally about half the frame-rate), more cores certainly help in a game such as BF3.
Single-player plays at 97-99% GPU usage at all times and never drops with the FX8150. With the Phenom 2 970, even in single-player the GPU usage would drop into the Mid-80% range at parts where there is a lot of action.
Even at times in single-player where the GPU's were running at 99% with the Phenom, I still had a 5-10FPS increase when switching to the FX chip.


Example:

The "Kaffarov" mission where you do a H.A.L.O jump out of the airplane. When the mission begins you are in the airplane checking each others equipment before the door opens.

Phenom II: ~100-104fps @99% GPU Usage
FX8150: ~108-114fps @99% GPU Usage
(Same cards, same drivers, same GPU usage, same in-game settings. I did this multiple times with both CPU's with very consistent results.)


PC Specs:

- FX8150 @4.515Ghz 210Mhz bus
- 2x Powercolor 6970 2GB PCS+ @ 950/1500
- 8GB DDR3 1680Mhz
- 2x WD3000HLFS 300GB Velociraptor HDD
- Corsair AX1200 Power Supply


Game Settings:

- 1920x1080 @ 120Hz
- All Ultra
- Motion Blur on @ 50%
- 4x AA


Average FPS:

Metro: 85-120
Karkand: 55-95
Kaffarov: 70-110


Operation Metro 64 Player:






Strike at Karkand 64 Player:






Kaffarov Mission First 5 Minutes:

 
Last edited:
Wow, i need to do some extensive testing on my system. No idea what im pulling for GPU and CPU usage. Settings are Medium with a 5750 card at a small OC. But it will be interesting to run these tests on my little i3.
 
Frito said:
for me and in my opinion with my rig at 1920 x 1080 my i5 at stock clocks is close to fast enough but its not quite enough to keep multiplayer over 60 fps 100% of the time, an easy overclock to 4.2ghz will make that diff and anything beyond that is just added fps headroom.
Click to expand...

I agree, and I'm only running a single GTX460 @ 822/4200.

http://www./Optimize/OPS/Battlefield-3-NVIDIA-GeForce-GTX-460-OPS

I'm running those 1080p settings except with AA deferred off. With 2x MSAA as suggested, my framerates would dip down to ~30 in some spots even OC'd. I was still probably averaging 50ish, but the dips were definitely noticeable.

Anyways, the CPU side of things. I don't run anything to track my FPS average; I just run with render.drawfps 1 and renderdevice.vsyncenable 1. At stock 2500K speeds, I am often at the 60 mark but it definitely dips down to the low-mid 50s and sometimes into the high 40s, but I never really notice it worse than that.

Now that I've gone to a Z68 board, I'm running @ 4.2GHz, and I rarely ever see it move from 60fps. Definitely no noticeable slow-downs anymore.

I was pretty happy at stock speeds, but I am definitely happy at 4.2. I'm eagerly awaiting the next gen cards though to see what they can give me in a ~9.5" card with rear-facing power connectors. I REALLY liked how it looked with even only 2x MSAA. I'd love to run with 4x. Even without any MSAA, AfterBurner is showing that I'm using the entire 1GB frame buffer. Really looking forward to a 2GB 7850/7870 or 760 Ti.
 
DuronClocker said:
I agree, and I'm only running a single GTX460 @ 822/4200.

http://www./Optimize/OPS/Battlefield-3-NVIDIA-GeForce-GTX-460-OPS

I'm running those 1080p settings except with AA deferred off. With 2x MSAA as suggested, my framerates would dip down to ~30 in some spots even OC'd. I was still probably averaging 50ish, but the dips were definitely noticeable.

Anyways, the CPU side of things. I don't run anything to track my FPS average; I just run with render.drawfps 1 and renderdevice.vsyncenable 1. At stock 2500K speeds, I am often at the 60 mark but it definitely dips down to the low-mid 50s and sometimes into the high 40s, but I never really notice it worse than that.

Now that I've gone to a Z68 board, I'm running @ 4.2GHz, and I rarely ever see it move from 60fps. Definitely no noticeable slow-downs anymore.

I was pretty happy at stock speeds, but I am definitely happy at 4.2. I'm eagerly awaiting the next gen cards though to see what they can give me in a ~9.5" card with rear-facing power connectors. I REALLY liked how it looked with even only 2x MSAA. I'd love to run with 4x. Even without any MSAA, AfterBurner is showing that I'm using the entire 1GB frame buffer. Really looking forward to a 2GB 7850/7870 or 760 Ti.
Click to expand...

So this is awesome news. Can you possibly check something for me? Run strike at karkand. With 64 players and keep your eye on fraps. See what your lowest fps is on that map. I'm in the high 20s. Basically I want to know what my gtx 460 is capable of before I upgrade. If going to am i5 2500k @ 4.2 ghz means my gtx 460 wont drop below 45 fps under any circumstance in bf3... I'm in.
 
ZimCoe said:
good work m8.

I wonder how my c2duo @ 4GHz stacks up to a sandybridge!

I know I should upgrade but can't stomach ditching my whole rig bar drives and gfx card :(

Are you using fraps to bench this? Might give it a go.
Click to expand...

Send me a copy of BF3 on steam and I'll do a bench for you. I run a C2D @ 3.7GHz
 
I was running a Q9550 at 4.1 ghz with terrible fps with my 2 6950's in crossfire. (28-55 usually on High/Ultra) Going to install my new 2500k and overclock it and post my new results later. Can't wait!
 
neokeelo said:
I was running a Q9550 at 4.1 ghz with terrible fps with my 2 6950's in crossfire. (28-55 usually on High/Ultra) Going to install my new 2500k and overclock it and post my new results later. Can't wait!
Click to expand...

Your probably going to see significant improvement, especially if you OC it.
 
You must log in or register to reply here.
Back
Top