Battlefield 4 Beta Performance: 16 Graphics Cards, Benchmarked

[JackNaylorPE]

Would love to have seen 2Gb vs 4GB addressed given BF4's recommendations for 3GB

 

[Nuno Silva]

this one is a very nice beta, since it came out I and most of people that are trying to play it (just look on how the servers are NOT populated) cant play EVEN after secon patch came out. Seriously, why is tom's hardware doing an article on a such broken beta game ... you just lost all credibility for me

 

[cleeve]

This is bull. It says the titan's avg fps is 75. My non overclocked 780 gets 100 fps average.

There's *nothing* more futile than calling out someone else's benchmark results when you don't benchmark it in the exact same way. 😉

No, wait. Using a leaf blower might be more futile. That's about it.

 

[BigMack70]

OK, so this article got my interest piqued in some more CPU testing, which I have now done. I'm definitely curious to see how things shape up at final release, because according to the beta, this game is to CPUs what Crysis 3 is to GPUs.

My conclusion... at 1440p maximum settings, 780 SLI is CPU bottlenecked by a 5 GHz 2600k. That's insane, frankly.

Playing on a 64 man server each time, I played normally for 15 minutes and used FRAPS to benchmark the playtime. Here are the results (again, all on 780 SLI @ 1100/1500, Windows 7 x64, latest 330.41 beta driver). I am only reporting avg fps because minimum FPS seems to be all over the place in my benchmarking (anywhere from 20-40fps)... I think FRAPS is picking up some sort of framerate hitch that can occur when you die and go to menu.

2600k @ stock - 49.6
2600k @ 4.8 GHz - 70.0 FPS
2600k @ 5 GHz - 77.0 FPS
2600k @ 5 GHz, Hyperthreading disabled - 67.3 FPS

Only the bold results were fully playable. There was distracting sporadic stuttering at stock and at 5 GHz when HT was disabled that were annoying during gameplay. I'd consider the stock results unplayable, the 5 GHz w/o HT results were borderline playable. The thing that shocked me most was 100% CPU usage at 5 GHz without HT (when HT is enabled, games never load a CPU to 100% - it always loads to like 75% max).

 

[Dagstar]

FWIW with 2 4gb gtx 770s (factory overclocked only) and an i7-4770k@4ghz the lowest fps I got across all matches was 45 and the highest was 89. Avg. fps was 71. 1440p with all settings at ultra/max using 331.40 beta drivers. This was measured with fraps running each entire match. Average fps, lowest and highest marks were taken over 5 64 man conquest matches.

 

[benjamin pearcy]

Hm fx8350 looks good. I think amd fx continue to show its magic on newer games. lucky me.

 

[nukemaster]

OK, so this article got my interest piqued in some more CPU testing, which I have now done. I'm definitely curious to see how things shape up at final release, because according to the beta, this game is to CPUs what Crysis 3 is to GPUs.

My conclusion... at 1440p maximum settings, 780 SLI is CPU bottlenecked by a 5 GHz 2600k. That's insane, frankly.

Playing on a 64 man server each time, I played normally for 15 minutes and used FRAPS to benchmark the playtime. Here are the results (again, all on 780 SLI @ 1100/1500, Windows 7 x64, latest 330.41 beta driver). I am only reporting avg fps because minimum FPS seems to be all over the place in my benchmarking (anywhere from 20-40fps)... I think FRAPS is picking up some sort of framerate hitch that can occur when you die and go to menu.

2600k @ stock - 49.6
2600k @ 4.8 GHz - 70.0 FPS
2600k @ 5 GHz - 77.0 FPS
2600k @ 5 GHz, Hyperthreading disabled - 67.3 FPS

Only the bold results were fully playable. There was distracting sporadic stuttering at stock and at 5 GHz when HT was disabled that were annoying during gameplay. I'd consider the stock results unplayable, the 5 GHz w/o HT results were borderline playable. The thing that shocked me most was 100% CPU usage at 5 GHz without HT (when HT is enabled, games never load a CPU to 100% - it always loads to like 75% max).

Thanks for that info.

 

[Duckhunt]

,Quote :
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.

Ha ha ha. Everyone wants to me be maniac and blow up the big features. ha ha. BF4 needs more of this feature. We need more destruction. I can see BF4 is moving in the right direction. I appreciate the difficulty in putting in the resources for this kind of feature. BUT PLEASE make more use of it.

 

[smeezekitty]

The thing about having freely destroyable/modifiable worlds is each and every object needs to be modeled in detail. This is very CPU intensive.
That's why usually only a few items can be damaged/modified in world.

 

[kato128]

I'm a little disappointed to see such a poor selection of 6000 series radeon's. I would expect a lot of the people (myself included) who are using amd and reading this article would have either a 6950 or 6970 and are wondering if an upgrade is worth it. Having the low end cards is fine but what about those who had the top of the line of a gen or 2 back?

 

[ojas]

OK, so this article got my interest piqued in some more CPU testing, which I have now done. I'm definitely curious to see how things shape up at final release, because according to the beta, this game is to CPUs what Crysis 3 is to GPUs.

My conclusion... at 1440p maximum settings, 780 SLI is CPU bottlenecked by a 5 GHz 2600k. That's insane, frankly.

Playing on a 64 man server each time, I played normally for 15 minutes and used FRAPS to benchmark the playtime. Here are the results (again, all on 780 SLI @ 1100/1500, Windows 7 x64, latest 330.41 beta driver). I am only reporting avg fps because minimum FPS seems to be all over the place in my benchmarking (anywhere from 20-40fps)... I think FRAPS is picking up some sort of framerate hitch that can occur when you die and go to menu.

2600k @ stock - 49.6
2600k @ 4.8 GHz - 70.0 FPS
2600k @ 5 GHz - 77.0 FPS
2600k @ 5 GHz, Hyperthreading disabled - 67.3 FPS

Only the bold results were fully playable. There was distracting sporadic stuttering at stock and at 5 GHz when HT was disabled that were annoying during gameplay. I'd consider the stock results unplayable, the 5 GHz w/o HT results were borderline playable. The thing that shocked me most was 100% CPU usage at 5 GHz without HT (when HT is enabled, games never load a CPU to 100% - it always loads to like 75% max).

Wow that's interesting...i don't have access to any of that hardware but i really enjoy tests like these, so thanks for posting that!

BTW: You think SLI issues (artificial overhead?) are contributing? Don didn't report CPU utilization on his rigs so i'm not sure we can rule that out completely...(but mostly, i guess).

 

[ojas]

I'm a little disappointed to see such a poor selection of 6000 series radeon's. I would expect a lot of the people (myself included) who are using amd and reading this article would have either a 6950 or 6970 and are wondering if an upgrade is worth it. Having the low end cards is fine but what about those who had the top of the line of a gen or 2 back?

Wait. For. The. Game's. Release. For. The. Full. Article.

 

[JustinTHFC1984]

Is there any way you can describe how you benched the game so I can try and replicate your results? I understand what you are saying however I was not trying to criticise your methods by the way just trying to understand how you got them kind of numbers

No offense taken, just letting you know.

We outline our method in the article: domination map, a server with about 40 ms latency. No other players on the server to keep the results consistent. Run along the sidewalk around the outskirts of the map for 60 seconds.

My apologies I missed that entirely in the article, when doing the same I am not dropping below 60 fps either so your results are entirely correct.

Apologies once again.

 

[damian86]

It's like some of the cards just made it. I am surprised they didn't test the 780.

 

[mauller07]

The thing that shocked me most was 100% CPU usage at 5 GHz without HT (when HT is enabled, games never load a CPU to 100% - it always loads to like 75% max).

This will be because the game does not utilise hyperthreading efficiently if at all.

since windows shows each hyperthreaded physical core as 2 logical cores, since it does not fully utilise hyperthreading it will not fully load both logical threads but can easily fully load a physical core with hyperthreading disabled.

 

[Alen Kirm]

expecting 80fps with r9 290x ...

 

[logainofhades]

I'm a little disappointed to see such a poor selection of 6000 series radeon's. I would expect a lot of the people (myself included) who are using amd and reading this article would have either a 6950 or 6970 and are wondering if an upgrade is worth it. Having the low end cards is fine but what about those who had the top of the line of a gen or 2 back?

6950 would probably trade blows with a 7790 and the 6970 with a 7850. Roughly the idea I managed to get using anandtech's bench. All it takes is a little time to see what your generation cards performance is relative to current gen.

 

[gsxrme]

Hyper threading is Junka for gaming, I agree with i5 CPUs. I only disabled my hyper threading on my 2600k to get 5.1Ghz at a much much lower temp and voltage. Now games love the extra 300Mhz over 4 hyper threaded threads. *puke*

 

[gsxrme]

Hyper threading is Junka for gaming, I agree with i5 CPUs. I only disabled my hyper threading on my 2600k to get 5.1Ghz at a much much lower temp and voltage. Now games love the extra 300Mhz over 4 hyper threaded threads. *puke*

 

[logainofhades]

Hyper threading is Junka for gaming, I agree with i5 CPUs. I only disabled my hyper threading on my 2600k to get 5.1Ghz at a much much lower temp and voltage. Now games love the extra 300Mhz over 4 hyper threaded threads. *puke*

Hyperthreading is actually good for BF4, though. Crysis 3 also makes use of those extra threads.

 

[BigMack70]

Hyperthreading appears to have tangible benefit in this game, to a degree I've not seen before in a game.

I don't think SLI overhead is to blame on my setup. Over on the BF4 forums, some people were suggesting it may be an issue with Windows 7 (as people with similar rigs to mine were claiming higher-than-60fps minimum framerates with Windows 8). So it's possible that Windows 8 would fix the CPU bottleneck. I'm not sure... I never know how much to trust anecdotal claims like that, and I don't feel like setting up a dual boot with Windows 8 or 8.1 to find out myself.

 

[logainofhades]

Given that is still in Beta, I wouldn't bother with dual booting either.

 

[ninong]

i need a new graphics card before bf4 starts..should i buy the 7970/280x now or wait and hope nvidia drops their prices? my budget is only $350 (2560x1440)

 

[logainofhades]

i need a new graphics card before bf4 starts..should i buy the 7970/280x now or wait and hope nvidia drops their prices? my budget is only $350 (2560x1440)

Is your power supply capable of handling such a GPU?

 

[ojas]

The thing that shocked me most was 100% CPU usage at 5 GHz without HT (when HT is enabled, games never load a CPU to 100% - it always loads to like 75% max).

This will be because the game does not utilise hyperthreading efficiently if at all.

since windows shows each hyperthreaded physical core as 2 logical cores, since it does not fully utilise hyperthreading it will not fully load both logical threads but can easily fully load a physical core with hyperthreading disabled.

It does utilise HTT, since BF4 / Frostbite 3 can use up to 8 hardware threads.

HOWEVER, interesting thing i found in Chis' review of BF3:

We’ve also read about folks complaining about stuttering issues caused by Hyper-Threading; disabling the feature seems to smooth things out for them. At no point did a perceptible stuttering (aside from the jerkiness attributable to a too-slow GPU at a too-high setting) afflict our platform. However, we can confirm that turning off Hyper-Threading on the Core i7-2600K, going from eight logical processors to four physical ones, doesn’t hurt performance in any way, and in fact may slightly increase it. The rest of our tests were run with Hyper-Threading enabled, but feel free to shut it off if it benefits your experience!

But then BF3's single player at least didn't scale above 4 threads anyway so that may have been the scheduler jungling stuff around too much.