BF3 patch includes 7xxx series perf improvements

[UJ09]

Finally no more stuttering, and the solution is Windows Update. I didn't update my Windows Update quite some time now, so I just updated all the updates available and then stuttering is gone. Blame myself for not updating too often lol.

Although less stuttering, but both cards still run at 50-70%. I opened the video option, noticed that I haven't using any MSAA yet, so decided to try 4XMSAA. Surprisingly now I can run roughly 80% all of the time, it looks like I need to max everything in order to solve this scaling problem, at least for my case.

And I do think CPU bottlenecks my crossfire scaling. I know it because when the round starts with nothing much happens for all the players, both cards run almost 99%. But once the battle begins, it slowly scales down to 90%, then pretty much 70-80% for whole round.

Well at least I can play it without stuttering problem, I'm going to WC my system soon and hope my 2500k can get somewhere around 5Ghz, crossfire scaling problem should be fixed I believe.

 

[Rvenger]

Finally no more stuttering, and the solution is Windows Update. I didn't update my Windows Update quite some time now, so I just updated all the updates available and then stuttering is gone. Blame myself for not updating too often lol.

Although less stuttering, but both cards still run at 50-70%. I opened the video option, noticed that I haven't using any MSAA yet, so decided to try 4XMSAA. Surprisingly now I can run roughly 80% all of the time, it looks like I need to max everything in order to solve this scaling problem, at least for my case.

And I do think CPU bottlenecks my crossfire scaling. I know it because when the round starts with nothing much happens for all the players, both cards run almost 99%. But once the battle begins, it slowly scales down to 90%, then pretty much 70-80% for whole round.

Well at least I can play it without stuttering problem, I'm going to WC my system soon and hope my 2500k can get somewhere around 5Ghz, crossfire scaling problem should be fixed I believe.

More speed isn't going to completely fix the scaling issue. Battlefield 3 can use up to 6 threads, but I think its more of a bf3 core usage issue on this because I never seen all 4 of my cores pegged at 100% during the game.

 

[INGlewood78]

Any idea how to fix it? I run Heaven benchmark and both cards utilized 99% all the time, but for BF3 this's different story. I also clocked both card at 1125, it'll a little better than stock, but still suck b/c both cards are still running at 50-70%.

I would try these 3 things in order.

1. Uninstall drivers, driver cleaner, reinstall the driver that came with the cards (i.e. Driver CD). I am using Sapphire's CD drivers.
2. Turn off HT. I've read that solved low utilization problems for some people.
3. Clean reinstall of OS. This is your worse case scenario. My buddy had this issue with the 6990 (low utilization in BF3) and a clean installed cleared up whatever was causing the conflict.

 

[Rvenger]

I'm guessing your points are valid because shortly after my previous post, I joined a different 64 player server & a different map (don't recall which) that only had about 18 players (don't recall their pings, but nothing jumped off the charts).

FPS @ 100+ & both cards between 97 - 99% the entire round. (Smoothest round I've played thus far).

One thing I changed was I increased my OC from 4.5 > 4.8.

I'll do some more testing later & see how it goes.

Climb to the top of one of the hotels in Gulf of Omen and look out at the view. That usually spikes the CPU pretty good.


@ BD - Very good points. Ping can have a lot to do with the stuttering that happens from time to time. I guess it could be a fault of FPS drops as well. Never thought of that.



EDIT: I didn't think of this until now. Do you think this could have anything to do with our CPUs not being PCIE 3.0 capable and using PCIE 3.0 cards? It sounds like this could be a bandwidth restriction through the PCIE bus.

 

[wahdangun]

Finally no more stuttering, and the solution is Windows Update. I didn't update my Windows Update quite some time now, so I just updated all the updates available and then stuttering is gone. Blame myself for not updating too often lol.

Although less stuttering, but both cards still run at 50-70%. I opened the video option, noticed that I haven't using any MSAA yet, so decided to try 4XMSAA. Surprisingly now I can run roughly 80% all of the time, it looks like I need to max everything in order to solve this scaling problem, at least for my case.

And I do think CPU bottlenecks my crossfire scaling. I know it because when the round starts with nothing much happens for all the players, both cards run almost 99%. But once the battle begins, it slowly scales down to 90%, then pretty much 70-80% for whole round.

Well at least I can play it without stuttering problem, I'm going to WC my system soon and hope my 2500k can get somewhere around 5Ghz, crossfire scaling problem should be fixed I believe.

yeah, its definitely cpu bottleneck, because when you crank up the graphic option the bottleneck will shift to GPU.

 

[BD231]

The link below suggests that beyond 3 cpu cores you see very little increase in cpu usage, that's what causes the CPU bottleneck and why you're seeing an increase in frames with your clocks. Better multi-threaded optimization would have been nice, but it seems like 2 threads are doing a fair share more of the important work than any others.

http://www.techspot.com/review/458-battlefield-3-performance/page7.html

 

[Joseph F]

Damn, that's barely 50% scaling D:

His CPU is probably bottlenecking those two beastly cards.

Edit: Nevermind, I didn't know that he has a 2500k at 4.5GHz.

 

[djnsmith7]

It's running @ 4.8 for now. We'll see how things progress.

 

[Stuka87]

I have been experimenting today with my single 7950 (running @1000/1300) at 1080P, everything on Ultra except motion blur is off and using SSAO.

Typically I play with vsync on. As I provides the best experience IMHO. However, for kicks I turned it off while I had afterburner going to watch GPU usage. Turns out I very rarely hit 100% on 64 player servers. I seem to average around 75% or so. In other games I easily get 95%+ GPU usage. But at least in BF3, I am certainly CPU bound. Which I guess isnt a surprise. I do hit 100+ FPS in areas. But with vsync on, I am almost always at 60fps.

EDIT: Hah, just realized I had vsync on when I ended up launching Afterburner, as the FPS count is a solid 60. So disregard the above, will retest it.

 

[Rvenger]

His CPU is probably bottlenecking those two beastly cards.

Edit: Nevermind, I didn't know that he has a 2500k at 4.5GHz.

It still could be bottlenecking. BF3 loves multithreaded performance and the 2500k only has 4 cores without HT. 2600k is proven to run better in BF3 because it has 8 threads and BF3 can use up to 6 threads. Right now BF3 is literally the only game that the 2600k benefits from. We still need more solid proof of this though.

 

[Rvenger]

Kharg Island 32 player map CPU usage. 2500k @ 4.4ghz

 

[Stuka87]

I didn't graph mine, but did just play a bunch with vsync off. I averaged roughly 85-90% CPU. I am certainly CPU bound. GPU was around 80-85% average. It would hit 95-100% in certain areas. This was on 64 player Caspian Border. Which IMHO is a worst case scenario as far as BF3 maps go.

 

[Rvenger]

Operation Metro 64 Player
http://i749.photobucket.com/albums/xx132/rpisano01/OperationMetro64player.png

Sharqi Peninsula 64 player
http://i749.photobucket.com/albums/xx132/rpisano01/SharqiPeninsula.png

Damavand Peak 64 Player
http://i749.photobucket.com/albums/xx132/rpisano01/DamavandPeak64Player.png


Caspian Border 64 Player
http://i749.photobucket.com/albums/xx132/rpisano01/caspianborder64player.png

 

[UJ09]

4Ghz 2500K on Operation Firestorm 64 players map. Almost fully loaded on CPU all of time, no wonder my crossfire scaling is not fully utilized because BF3 itself needs much more CPU power on 64 players map.

 

[Rvenger]

Yes, and I think its even bottlenecking my GPU somewhat because I run 1920x1080 on all low with Vsync off and AA/AF maxed. FPS range from 53 - 180. Its seems to be dipping in the high traffic areas. I noticed it the most on Sharqi Peninsula. GPU usage according to afterburner is dipping in the low 90s. There is no reason why I should be dipping below 60fps on all low and everytime my FPS dip, I see a drop in GPU usage.


Either this game is poorly coded or the 2500k has poor multithreaded performance (at least for this game).

UJ09 - your pic isn't showing.

 

[Rvenger]

4Ghz 2500K on Operation Firestorm 64 players map. Almost fully loaded on CPU all of time, no wonder my crossfire scaling is not fully utilized because BF3 itself needs much more CPU power on 64 players map.

Noticed you are using the same program I found I think I might make a program tonight that monitors and logs CPU usage more precise. I would like to see my exact numbers on all 4 cores. I want to see if this actually hits 100% at some point. I am betting yours is for sure but how many cores is hitting 100% is the question.

 

[Rvenger]

FX6100 @ 4ghz 64 player map Gulf of Omen (all 6 cores)

FX6100 with 4 cores enabled (Grand Bazaar 64 player)

6 cores enabled (Grand Bazaar 64 player)

 

[Stuka87]

So here are mine. This was taken on Strike at Karkland:

VSYNC ON:

VSYNC OFF:

 

[Rvenger]

So here are mine. This was taken on Strike at Karkland:

VSYNC ON:

VSYNC OFF:

This just about proves it that 4 cores choke on BF3. HT would help and I find it interesting that bulldozer didn't even touch 80% on 6 cores. I am kinda upset that this 2500k is bottlenecking in BF3.

 

[Stuka87]

This just about proves it that 4 cores choke on BF3. HT would help and I find it interesting that bulldozer didn't even touch 80% on 6 cores. I am kinda upset that this 2500k is bottlenecking in BF3.

Are you still using the 6750 in the FX rig? If so, this may be why its so much lower. When I had a 5750 before the 7950, my CPU was down around 50% all the time. Switching cards made a huge difference in CPU usage.

 

[Rvenger]

Are you still using the 6750 in the FX rig? If so, this may be why its so much lower. When I had a 5750 before the 7950, my CPU was down around 50% all the time. Switching cards made a huge difference in CPU usage.

Take a look above. I tested 6 cores and 4 cores on the FX rig. 4 cores showed similar results as the 2500k. If you like I can test with the GTX560 but I am pretty sure I will get the same results.

 

[Grooveriding]

That is a full 64 player match on Gulf of Oman on the system in my sig.

edit: I actually use the frame rate limiter in BF3 to cap at 65fps though, so that probably affected things.

 

[Rvenger]

That is a full 64 player match on Gulf of Oman on the system in my sig.

edit: I actually use the frame rate limiter in BF3 to cap at 65fps though, so that probably affected things.

Can you turn off vsync to see if anything changes?

 

[Thor86]

Not sure if anyone mentioned this in this thread, but Win7 Power settings for Core Parking should help the micro-stuttering.

I haven't tried this myself, but may give it a shot.

http://ultimatecomputers.net/forum/viewtopic.php?f=7&t=3644

 

[Silverforce11]

Not sure if anyone mentioned this in this thread, but Win7 Power settings for Core Parking should help the micro-stuttering.

I haven't tried this myself, but may give it a shot.

http://ultimatecomputers.net/forum/viewtopic.php?f=7&t=3644

Wow thanks for that, a few games i play sometimes there's a tiny microstutter, i always assume its for loading resources etc.