Battlefield 3 official performence/fps thread

[Silverforce11]

Seems AMD drivers still need work, they suffer a big performance drop with motion blur enabled (10 fps loss vs 5 for gtx580). I myself play with it OFF so never realized how it affected performance, but interesting to see something is messed up. As per the explosions causing a fps lag spike, haven't seen it yet, though being near a tank when it explodes cause the entire screen to go into blur/flashbang mode.

Also, i think [H] may be slightly CPU bottlenecked (may affect the min fps and or lag spikes) as BF3 MP is very cpu intensive.

Same for the gtx570/6950, motion blur drops performance more on AMD cards than NV. HBAO is big fps drain in MP as suspected.

New patch: "Adjusted the “stuttering” encountered on some PC configurations"

 

[Makaveli]

I play with Motion blur off I find it distracting and its just gets me killed more than with it off. And I don't really notice that big of a hit with HBAO but then again everything is smooth so I haven't bothered looking at it with the fraps on.

I installed the Patch last night and played with 11a drivers and performance was great.

I can finally quit the game during the 1 min counter between maps instead of waiting for the map to load then quit. They also seemed to have nerfed the sniper rifle abit.

I also don't think a 920 @ 3.6 will bottleneck you at 1920x1200 in multiplayer.

 

[Termie]

I play with Motion blur off I find it distracting and its just gets me killed more than with it off. And I don't really notice that big of a hit with HBAO but then again everything is smooth so I haven't bothered looking at it with the fraps on.

I installed the Patch last night and played with 11a drivers and performance was great.

I can finally quit the game during the 1 min counter between maps instead of waiting for the map to load then quit. They also seemed to have nerfed the sniper rifle abit.

I also don't think a 920 @ 3.6 will bottleneck you at 1920x1200 in multiplayer.

It won't bottleneck any single card, but it will badly bottleneck high-end Crossfire/SLI, and HardOCP should know this. I'm surprised they are still benching on their 920 system when they have a 2600k system they built specifically due to CPU-bottlenecking they found in SLI setups.

BTW, I agree that things seemed even smoother than usual with the patch and 11a.

 

[Makaveli]

It won't bottleneck any single card, but it will badly bottleneck high-end Crossfire/SLI, and HardOCP should know this. I'm surprised they are still benching on their 920 system when they have a 2600k system they built specifically due to CPU-bottlenecking they found in SLI setups.

BTW, I agree that things seemed even smoother than usual with the patch and 11a.

I agree you will see it with a Highend SLI/Xfire setup but the images he has posted above are for a single card setup.

I do always LOL when [H] shows power consumption numbers for their c0 920 which needs ungodly level of voltage to hit 4ghz and they most likely have speedstep and other power saving features disabled yet say look at how much power its using.

I'm not sure why they haven't switched yet for the review machine.

 

[Dribble]

Impressed [H] finally did a bench at 60fps, instead of their usual 30

 

[Grooveriding]

Impressed [H] finally did a bench at 60fps, instead of their usual 30

I think it says a lot for him having the sense to know what is appropriate performance for what type of gameplay you are in.

Single-player, do you need 60FPS ? Nope, 30-40 is fine and the best possible IQ. Multiplayer, do you need 60FPS ? Yes, reduce settings till you get a playable 60FPS and give bench results. :thumbsup:

 

[imaheadcase]

Now if he would get rid of those ugly bench screens he uses that everyone complains about. But you know kyle, hard headed to a fault.

 

[m3t4lh34d]

It won't bottleneck any single card, but it will badly bottleneck high-end Crossfire/SLI, and HardOCP should know this. I'm surprised they are still benching on their 920 system when they have a 2600k system they built specifically due to CPU-bottlenecking they found in SLI setups.

BTW, I agree that things seemed even smoother than usual with the patch and 11a.

Battlefield 3 is FAR more GPU dependent than CPU dependent. Toms did a CPU scaling experiment with both AMD and Intel CPUs. Regardless of IPC or cores, the performance remained the same. Even all the way down to 2 cores @ 2ghz on an AMD platform. There is NO way that an i7 @ 3.6ghz would bottleneck a single or dual GPU setup when a dual core phenom @ 2ghz gets the same frames as a 2500k. It seems as if battlefield only requires 2 cores to get the job done without variability in avg fps.

 

[m3t4lh34d]

Now if he would get rid of those ugly bench screens he uses that everyone complains about. But you know kyle, hard headed to a fault.

I haven't seen anyone complain about them, and I actually like them to be honest. They keep everything neat without excess, and the graphs are a nice touch as well.

 

[Termie]

Battlefield 3 is FAR more GPU dependent than CPU dependent. Toms did a CPU scaling experiment with both AMD and Intel CPUs. Regardless of IPC or cores, the performance remained the same. Even all the way down to 2 cores @ 2ghz on an AMD platform. There is NO way that an i7 @ 3.6ghz would bottleneck a single or dual GPU setup when a dual core phenom @ 2ghz gets the same frames as a 2500k. It seems as if battlefield only requires 2 cores to get the job done without variability in avg fps.

Those Tom's benches have been proven to be useless in predicting multiplayer performance. It might as well be a different game.

By way of example, my i7-860@3.4 just barely keeps up with my HD5850 crossfire setup. The CPU maxes out constantly, while the GPUs almost never hit 99%. I should note that hyperthreading seems to help a lot - another thing the single-player benchmarks had people confused about. Multiplayer absolutely scales with cores and threads.

 

[destrekor]

What would you do in this situation:

If I play on a single 1080p monitor, I can play with almost-max Ultra settings and have essentially rock-steady 60fps.

If I play with triple monitors, I get to enjoy a very expansive view when running around or holding a spot. Oh, and you really get to appreciate how cramped the first-person view of the tank really is, since the two side monitors are black at that time.
Problem is, current settings are mostly medium, a few settings on low, blur off and AO might be either off or on SSAO. It doesn't look pretty in comparison to Ultra, but I can't afford the horsepower to remedy that situation entirely. Worse still, it's a rare map and rarer still is the specific actors playing their parts in the right way that produces a steady 60fps even at those settings. Typical is around 50fps, sometimes mid and low 40s with minimums around 25-30fps. Sometimes I don't see those minimums often if at all, other times they are fairly prolonged moments. On some maps, I also don't go above 45fps (no vsync, but do have a 60fps cap applied in-game).

sidenote:
I do plan on using Step-up to upgrade to a better model, I'd love the 580 3gb (surround with 1.5gb is iffy) but the $600 price tag means I'm paying a total of around $600 to upgrade my 560 Ti 2gb SLI setup.



I'm just not sure. I feel overall I can perform better with surround, at other times I feel the lower framerate might become an issue to my reaction abilities. I've only been playing with surround in BF3 since the patch, as it fixed most of the issues in that regard.

Which would y'all end up picking.

In the end, even adding some more low settings to the mix, the game still looks terrific in comparison to older engines.

 

[Silverforce11]

http://hardforum.com/showpost.php?p=1038062569&postcount=7

Excellent job by the user. 2500k @ 3.4ghz going to 4.2ghz and see a nice fps increase. That's on Metro which is one of the least intensive map.

If he had a better GPU setup, the scaling would be even more noticeable. BF3 loves more cores and faster speed in 64MP.

 

[miahallen]

 

[Termie]

I'm not sure what you're showing - seems like you're getting zero SLI scaling. Plus you're hitting zero for your min...

In other news, Tom's has just published a study of the impact of an SSD on BF3. The actual benches are near the end of the article, but the effect is dramatic - game and level loading times are cut in half, and hitching mid-game is also eliminated (they attribute that to a virus scan that hit the hard drive while they were playing): http://www.tomshardware.com/reviews/battlefield-rift-ssd,3062.html

Just another thing to upgrade for BF3!

 

[Makaveli]

Those Tom's benches have been proven to be useless in predicting multiplayer performance. It might as well be a different game.

By way of example, my i7-860@3.4 just barely keeps up with my HD5850 crossfire setup. The CPU maxes out constantly, while the GPUs almost never hit 99%. I should note that hyperthreading seems to help a lot - another thing the single-player benchmarks had people confused about. Multiplayer absolutely scales with cores and threads.

I agree with this single player benchmarks in BF3 are a complete waste of time don't look at them.

In Multiplayer you need a Quad core cpu period!

 

[Makaveli]

I'm not sure what you're showing - seems like you're getting zero SLI scaling. Plus you're hitting zero for your min...

In other news, Tom's has just published a study of the impact of an SSD on BF3. The actual benches are near the end of the article, but the effect is dramatic - game and level loading times are cut in half, and hitching mid-game is also eliminated (they attribute that to a virus scan that hit the hard drive while they were playing): http://www.tomshardware.com/reviews/battlefield-rift-ssd,3062.html

Just another thing to upgrade for BF3!

I have BF3 on a separate HD and not my SSD so I may have to test this. But So far with 12GB of ram in my system I only find the first load of the first map I play abit slow after that load time isn't really a problem I tend to be first in the map. And I get no in game hitching at all which I think is do to the amount of system ram and the 2GB vram on my gpu.

 

[Ozark_86]

Hey all - trying to understand how server ping / lag affect decision to upgrade video or not. I'm playing 1080 with gtx460, 4gb ram, i5-750. What would be the best thing to upgrade? (ssd is on it's way... thanks black friday)

 

[skipsneeky2]

Hey all - trying to understand how server ping / lag affect decision to upgrade video or not. I'm playing 1080 with gtx460, 4gb ram, i5-750. What would be the best thing to upgrade? (ssd is on it's way... thanks black friday)

gtx460 first the recommended is gtx560 anything with more then 1gb would be solid.

4gb for me when i game is cutting it close...

 

[Majcric]

Run BF3 on the default high settings, if i try ultra I run out of vram with my 1g GPU. Framerate is generally 60 and above at 1680x1050 res.

 

[Termie]

Ok, all. I've been trying to put together some CPU and GPU benchmarks to give you some idea on scaling, but it's proven to be incredibly difficult. Even when on the same server and benching an identical location on the map, the results are entirely inconclusive. You can see them below, with both the CPU overclock and the GPU overclock listed on the left hand side.

What does this tell me? This game is incredibly hard to bench (hence the fact that only HardOCP has tried to do multiplayer benches), and the results have much more to do with what is going on in your multiplayer round than what your settings are. If you were to try to conclude something from the benches, it would be that I am CPU-limited at both 2.93GHz and 3.4GHz, and that I'm averaging around 75fps on Caspian and 80fps on Seine regardless of GPU clocks. The big difference is in the minimums - they are much higher with an overclocked CPU.

Keep in mind that on singleplayer, I increase linearly with GPU speed - I can get a 16% increase in fps with a 17% increase in clockspeed. Therefore, it's completely GPU-limited and therefore a completely different game to bench. The other thing to keep in mind is that these multiplayer benches are clearly benefitting from crossfire. I was getting around 43fps on high with a single HD5850.

 

[Silverforce11]

Nice work, i think you've hit the limit of 5850 CF so you cant really see an increase in avg, but only in minimum.

Minimum reflects nicely on CPU speed, good scaling.

 

[mattdallastx]

2 x EVGA SuperClocked 015-P3-1582-TR GeForce GTX 580 (Fermi) 1536MB in SLI maxed out settings @ 61fps constant. May dip to 60

COOLER MASTER HAF X RC-942-KKN1 Full Tower Computer Case
CORSAIR H70 Core High Performance Liquid CPU Cooler
Intel Core i7-980X Extreme Edition Gulftown 3.46GHz LGA 1366 Six-Core Desktop Processor
CORSAIR DOMINATOR GT 24GB (6 x 4GB) 240-Pin DDR3 SDRAM DDR3 2000 (PC3 16000) memory
CORSAIR Pro Series Gold AX1200 watt power supply SLI Certified 80 PLUS GOLD Certified
ASUS Rampage III Extreme LGA 1366 Intel X58 SATA 6Gb/s USB 3.0 ATX Intel Motherboard
LG Black Blu-ray Drive
(2)EVGA SuperClocked 015-P3-1582-TR GeForce GTX 580 (Fermi) 1536MB in SLI
Crucial 256gb SSD sata III 6gb/sec
Western Digital Caviar Black 1.5 gb sata III 6gb/sec
(2) Western Digital Caviar Blue 500gb - striped RAID
Microsoft Windows 7 Pro

 

[Termie]

Nice work, i think you've hit the limit of 5850 CF so you cant really see an increase in avg, but only in minimum.

Minimum reflects nicely on CPU speed, good scaling.

Yeah, the bottlenecking is really quite complicated in multiplayer. There's just so much going on, but at least I feel my system is pretty well balanced. The GPUs were close to maxing out (95-99&#37, and I'm not getting huge dips (especially with the CPU overclocked). So I'm happy - just wish I could show the impact of GPU overclocking a bit better, because it scales so incredibly well in single-player. The funny thing is that if I stand in one spot on a multiplayer map and don't move, I can see the scaling perfectly as I adjust GPU clocks on the fly, but once I start getting into the action, the GPUs really don't make that much of a difference - I think the CPU becomes the bottleneck for me at that point.

Actually, if you look at HardOCP's 1920x1200 SLI/Crossfire benches, it looks very similar to mine: http://www.hardocp.com/article/2011/11/22/battlefield_3_multiplayer_performance_iq_review/4

They have way more GPU power, and they're running with textures at ultra and motion blur on, so it's a bit more load than I was running, and they're getting 15fps more. But I'm pretty sure they're partially CPU-limited. They benched on an i7-920@3.6, which is basically on par with my 860@3.4, and way underpowered for a 580SLI rig.

2 x EVGA SuperClocked 015-P3-1582-TR GeForce GTX 580 (Fermi) 1536MB in SLI maxed out settings @ 61fps constant. May dip to 60

What resolution? I assume this is with vsync enabled. What do you get with it off? It should be close to 95fps based on HardOCP's testing.

 

[mattdallastx]

Yeah, the bottlenecking is really quite complicated in multiplayer. There's just so much going on, but at least I feel my system is pretty well balanced. The GPUs were close to maxing out (95-99%), and I'm not getting huge dips (especially with the CPU overclocked). So I'm happy - just wish I could show the impact of GPU overclocking a bit better, because it scales so incredibly well in single-player. The funny thing is that if I stand in one spot on a multiplayer map and don't move, I can see the scaling perfectly as I adjust GPU clocks on the fly, but once I start getting into the action, the GPUs really don't make that much of a difference - I think the CPU becomes the bottleneck for me at that point.

Actually, if you look at HardOCP's 1920x1200 SLI/Crossfire benches, it looks very similar to mine: http://www.hardocp.com/article/2011/11/22/battlefield_3_multiplayer_performance_iq_review/4

They have way more GPU power, and they're running with textures at ultra and motion blur on, so it's a bit more load than I was running, and they're getting 15fps more. But I'm pretty sure they're partially CPU-limited. They benched on an i7-920@3.6, which is basically on par with my 860@3.4, and way underpowered for a 580SLI rig.



What resolution? I assume this is with vsync enabled. What do you get with it off? It should be close to 95fps based on HardOCP's testing.

Yes this is with vsync enabled. Im running @ 1980 x 1080 on my 42" led. I'll have to check the fps with it turned off.

 

[ColtMaverick]

How will a 32" 1080P (assume 1920X1080) 120 Hz tv stand up against a 1920X1080 monitor? I get conflicting answers on the internet.