OFFICIAL: BF4 Mantle Cat 14.1 Performance and Feedback Thread

BF4 ran a lot better for me when I upgraded from 6 to 8GB, so you definitely need to upgrade. Shouldn't be too expensive either.

It is a dilemma for me as an AM3 user :) I prolly would have a hard time finding ram which would work on my system and then the investment is going to be a waste because the 1600mhz ram is not that great to move over to another platform

The ram that I have is pretty good it can run at 1T but sadly it is no longer being offered anywhere and 4 modules in AM3 is a pain to get working.
 
Last edited:
I doubt 4GB crashes anymore than 8GB. Run slower on higher settings?---certainly. I've heard the game has a lower ram footprint in 32 bit mode, though I haven't tracked it myself. All comparison screens I have made, show zero difference in image quality. The ram footprint differences are probably just a bigger cache for map assets, or something. I often use 32-bit mode, so that I can record with Afterburner.

This game crashes plenty, on 8+gigs of ram. In both 32 bit and 64 bit modes.

My particular system hasn't been too crashy, as far as the BF4 process. but it does happen. I get dropped from servers more often.

Two patches ago, I started blue-screening about every 40 minutes, in BF4. I had to relax my CPU and northbridge overclocks, to get it to stop. My videocard overclock had no effect on the BSOD. So it is still overclocked just as much as it has been.


*also, I just found out that Mantle only works in 64-bit, for BF4. If you have "Mantle" selected in the video options, but start the game in x86/32bit mode, it will run in D3D/DirectX and list the render option as "1" in the video options.
 
Last edited:
There might be something to extra ram and crashing. I have 16gb and have never had the game crash.

Now if I could only get the game to load with mantle enabled without crashing...
 
14.1 caused major system instability for me and I couldn't boot into Windows 8.1 without having to do a repair install. :rolleyes:

Back to 13.12 for me until AMD releases a stable driver.
 
Horrible experiences with 14.1 CFX, disabled. Single card (second removed) stock clocks, stock clocks volts increased 100% fan, etc, etc)

Will NOT run mantle in CFX (know there are bugs) crashes within 15 seconds.

14.1 in DX mode CFX lasts 3-9 minutes in game (tried ultra, auto, low)

heart breaking for me
 
I wish they would tell us when they expect to fully support GCN 1.0 and CFX. They're probably going to stay tight lipped about it. Pretty lame in my opinion, they never mentioned anything about it until last week.
 
Didn't AMD mention that CrossfireX support in games using Mantle is now dependent on the game developer to add it instead of the Mantle driver itself?

That means games would have to be developed to automatically detect two or more video cards, and take advantage of the extra GPU power provided by another video card.
 
I'm getting intermittent artifacts on my desktop. Reinstalled 14.1 drivers and the artifacts are back. I properly cleaned the drivers before install.

BUT, the actual BUG that I have is as follows: Regardless of any video settings in BF4, I get a MASSIVE fps drop when zooming in on certain maps. For example, on Zavod with DX11 I may average 100fps, and about 80 when zoomed into the forest. With Mantle, I'm getting 120fps and about 40 when zoomed in on the forest or buildings. I can put my graphics at low, peg my fps at 200, and it still drops to ~45-50 when zoomed. It's better on other maps, but still borderline unplayable for sniping.
 
Didn't AMD mention that CrossfireX support in games using Mantle is now dependent on the game developer to add it instead of the Mantle driver itself?

That means games would have to be developed to automatically detect two or more video cards, and take advantage of the extra GPU power provided by another video card.

That's for future drivers. BF4 is already programmed to take advantage of multiple GPU's but the drivers don't provide consistent performance
 
blower style coolers do not have the issue of not pushing enough air, it is the cooler itself that is constrained or lacks ability to soak the heat, blower fans have tons of pressure/flow and the noise to go with it, it is technically a good design air get sucked into its center and shot out its sides, I think maybe less fins etc, they just need a better design I hate more heat being thrown into the case :)
 
Quick guide to get Mantle Crossfire working:

- Download DDU: http://forums.guru3d.com/showthread.php?t=379505
- Reboot/Start Windows in Safe mode, and use DDU to uninstall the previous AMD drivers.
- Reboot to Windows in normal mode.
- Install 14.1 driver (Run as Administrator)
- Reboot
- Delete all the folders in %user%/documents/battlefield 4
- Enable Crossfire in CCC (It also helps to disable ULPS).
- Start BF4 in Test range, and enable Mantle.
- Exit the game, and restart again in Test range.
- At this point you can change any graphics settings. Use resolution scale at least 125%
- Make sure Vsync is disabled
- If you want to make ANY changes to your graphics settings only do them in Test range do not do it while in game/match.
- Enjoy!
 
blower style coolers do not have the issue of not pushing enough air, it is the cooler itself that is constrained or lacks ability to soak the heat, blower fans have tons of pressure/flow and the noise to go with it, it is technically a good design air get sucked into its center and shot out its sides, I think maybe less fins etc, they just need a better design I hate more heat being thrown into the case :)

I've had two blower style equipped videocards over the years. One a few years ago, one right now. Both times, the temps were higher than expected and the small turnbine fan became very audible after about 60% fan speed. Both times, removing the turbine fan and shroud and replacing it with one or two 80mm fans not only dramatically improved temps, but also dramatically lowered noise.

this is my current videocard. It has a pretty nice heatsink on it. The turbine fan/exhaust shroud setup stinks.
http://www.newegg.com/Product/Product.aspx?Item=N82E16814161404
 
I finally bothered to calculate my average frame rates


Earlier I posted raw perforoverlay frame data (CSV files) for 13.12whql, 14.1 DX11 and 14.1 Mantle. You can see my system specs and BF4 graphics settings in that post, as well.
http://hardforum.com/showpost.php?p=1040594668&postcount=29

There are two sets of data. One set is for a test server lap which was kept as similar as possible for each (I linked a video of what such a lap looks like).

The other set of data is for a roughly 6 minute play session, in 64 player Siege of Shanghai. (timed for no less than 6 minutes). I used a 24/7 Siege of Shanghai server so that I could ensure reliable testing of the same map on the same server. These multiplayer sessions also had AMD Performance Monitor dumping CPU core/s and GPU core utilization data. So the CPU load is slightly higher than what it normally would be.

Here are the average frame rates for the test server lap:

14.1 Mantle: 108
14.1 DX11: 103
13.12 DX11: 104

You can see here, the Test server probably doesn't stress CPU performance. As such, the average performance is basically the same. There are points at which Mantle performs better in the Test Server. But overall it is pretty much the same.

-----------------------------------------------------------

Average frame rates for 64 player Siege of Shanghai, during 6 minutes of play:

14.1 Mantle: 85
14.1 DX11: 58
13.12 DX11: 54


Here is where we get HUGE gains. That speaks for itself. Even despite the fact that the Mantle run was hitching every few seconds, with occasional hard stops/several frame drops, it still turns in a much higher average number.

The hitching is likely related to the fact that Mantle uses a lot more VRAM, for the same settings, on my system. which I show in this post here:
http://hardforum.com/showpost.php?p=1040596739&postcount=76

I have a 7870, so it is probably a bug with how Mantle is working on those cards/my system. However, it would be interesting to hear from other 7870 users about their experience, in specific regard to smoothness of play/hitching/stuttering. Or even 290x users that are running smooth, if they could show us VRAM comparisons.
 
Last edited:
When I play BF4 with mantle my performance was HORRIBLE and then it crashed. That was with a super clean install (*no we dont need to talk about it further).

I am running Crossfire 290X cards. My FPS was garbage and then it crashed to a black screen with no audio.

DX11 is far better for me so far. I am running an elite rig also with 3930K @ 4.7ghz so I have a gargantuan CPU to throw at the game.

The only thing I can think of is cooling as I am still on fans for the time being.
MSI afterburner works like a charm to keep my cards cooled under dx games but it wont work with mantle.

Is there something that will work with mantle? Or some recommendation I should look forward to using to spool my fans up to full at a much lower temperature before letting it get out of control?
 
When I play BF4 with mantle my performance was HORRIBLE and then it crashed. That was with a super clean install (*no we dont need to talk about it further).

I am running Crossfire 290X cards. My FPS was garbage and then it crashed to a black screen with no audio.

DX11 is far better for me so far. I am running an elite rig also with 3930K @ 4.7ghz so I have a gargantuan CPU to throw at the game.

The only thing I can think of is cooling as I am still on fans for the time being.
MSI afterburner works like a charm to keep my cards cooled under dx games but it wont work with mantle.

Is there something that will work with mantle? Or some recommendation I should look forward to using to spool my fans up to full at a much lower temperature before letting it get out of control?

Afterburner's overclocking and fan settings should work, even though you don't get an overlay in game. I mean, it automatically applies the overclocking and fan curve upon system startup, without even actively running afterburner. Those features don't depend upon the graphics API, to work.

But, CCC has manual fan controls, as well. You can create a custom curve, but you can at least crank the fans up to 80% or something, for testing purposes.
 
When I play BF4 with mantle my performance was HORRIBLE and then it crashed. That was with a super clean install (*no we dont need to talk about it further).

I am running Crossfire 290X cards. My FPS was garbage and then it crashed to a black screen with no audio.

DX11 is far better for me so far. I am running an elite rig also with 3930K @ 4.7ghz so I have a gargantuan CPU to throw at the game.

The only thing I can think of is cooling as I am still on fans for the time being.
MSI afterburner works like a charm to keep my cards cooled under dx games but it wont work with mantle.

Is there something that will work with mantle? Or some recommendation I should look forward to using to spool my fans up to full at a much lower temperature before letting it get out of control?

I've had many issues with amd cards combined with afterburner and dice games. Hard locks bsod etc. Very first thing you should try is closing afterburner before launching bf4. Sounds like what it could be. Known issue in bfbc2 for a long time. To this day.
 
I've had many issues with amd cards combined with afterburner and dice games. Hard locks bsod etc. Very first thing you should try is closing afterburner before launching bf4. Sounds like what it could be. Known issue in bfbc2 for a long time. To this day.

I don't think this is a problem in BF4.

I have run Afterburner as well as Trixx no problem with my 7970 CFX setup. Normally use AB, and no problems.
 
@A_N_D_Y

I'm getting artifacts too using a Diamond R9 290X (reference). I get them on my Windows taskbar, mainly when I'm browsing in Chrome while tabbed out of a game.

I also experience the known issue of stuttering using Mantle in BF4, no surprise here I bet.

I have a custom fan profile setup in MSI Afterburner and my card never exceeds 72 C despite sounding like a leaf blower at times. I doubt it is the heat causing artifacts, it must be the new driver.

I'm probably going back to 13.12 until things get sorted out, hopefully.
 
I finally rid myself of the old AMD drivers and reinstalled the 14.1's. BF4 is smooth as butter for me. Maybe 4 times a round I get a random hitch, but I have a TON of background software running at all times. Game is so smooth. Like glass. My old eyes can notice player movement and react a lot better since I'm seeing much more consistent frame times. CrossfireX seems to be working very well also.

New bug I found. Sometimes I'll watch an embedded flash video and it crashes my PC and makes it restart on it's own. My monitors start flashing white screens while it is crashing. Seems identical to an old bug from last year or 2012 that AMD fixed with flash player. I do have the enhancements in CCC on defaults which means they are enabled. It only did it once about 8 hours ago before I fired up BF4. 8 hours of seamless BF4 game play is pretty darn fun!
 
My experience so far (i5-2500K @ 4.3GHz, MSI HD7950, 8G DDR3-1866,1080P, mostly Ultra settings, Win7pro x64): I ran a 10 minute run on Siege of Shanghai under both Mantle and DX11, logging the runs with the built in performance monitor. DX11 was ~66fps and Mantle gave me ~73fps. An increase of ~10.6%. HOWEVER, in the latter few minutes of the run, Mantle starts to hiccup and stutter annoyingly bad.

Checking the log file shows that the Max CPU Time keeps spiking, as high as 776ms (~1.3fps minimum), frequently (40+ instances over 10min) saw spikes above 100ms (10fps) whereas the Max GPU Time only got as high as 47ms (~21fps minimum). At first i thought it may be another background program using the cpu to do misc work, so i closed programs and minimized the number of necessary startup programs and retried. Still get the awful stutter. I set BF4 to High Priority in the task manager to prioritize it over other cpu usage and still get the stutter. The Render.DrawScreenInfo option shows 3 job threads on my 4core processor. Not sure what the issue is.

My DX11 runs, while having a lower average fps, gave me much more consistent frame rates. Max CPU Time only spiked up to 54ms (18.5fps min), Max Frame Time up to 44ms (22.7fps min), and Max GPU Time up to 27.9ms(35.8fps min)
Perhaps this is part of the 7000 series optimizations they mention, but it seems like it isn't the GPU that is having issues because GPU Time under each API is low without spiking. I think if i can fix the CPU Time spikes, i'll see a bigger performance increase under Mantle. I'll see if i can force it to use a 4th cpu thread and retry also. As it stands, i think i'm going to wait for another patch or two before i really switch over to Mantle.

Maybe its an issue that manifests itself over time? Others are speculating that its a PunkBuster issue, where its trying to take a system screenshot. Some also suspecting GPU clocks changing due to power requirements or heat.
 
Getting artifacts on the desktop during mining. No artifacts on Cat 13.12

7970 CFX.

BF4 has been fine in Crossfire with no hitching.
 
@Buyers. Make sure you follow the directions in the first post and see if removing the old drivers with that program makes a difference.
 
I rolled back my drivers. These mantle ones would crash on the BF4 loading screen.
 
Is anyone able to test Tri-Fire or Quad-fire with Mantle yet? Or is simply crossfire with 2 cards enough of a headache?
 
My experience so far (i5-2500K @ 4.3GHz, MSI HD7950, 8G DDR3-1866,1080P, mostly Ultra settings, Win7pro x64): I ran a 10 minute run on Siege of Shanghai under both Mantle and DX11, logging the runs with the built in performance monitor. DX11 was ~66fps and Mantle gave me ~73fps. An increase of ~10.6%. HOWEVER, in the latter few minutes of the run, Mantle starts to hiccup and stutter annoyingly bad.

Sounds exactly like what's happening to me (7850 here). It can perform better, but it's inconsistent and it gets worse after a few minutes. Mantle drivers for 7000 series are simply not finished
 
I7 3930k @ 4.5
Rampage formula iv
4 X 4GB corsair vengeance 1600mhz
2 X 290x, XFX stock blowers
3x27 1080p in landscape. 5780x1080p v sync off 60hz.
win 8.1

Installed 14.1 using the prescribed proceedure.
Under Mantle bf4 crashed to a white screen at the first startup. Seemed to happen around the time that the game grabs focus from the desktop.
I got past this by not waiting for the game to grab focus, instead maximizing it to fullscreen myself.
no crashes after that point.

Poor crossfire results under mantle and dx 11.1 with the beta driver. Spawning infantry on test range I lost at least 40+ fps and cpu frame times virtually tripled. Gpu frame times weren't drastically different. Tried the test range, and multi player with similar results.

I didn't disable ulps, and the standby light on the second 290x was off.

Rolled back to 13.12 and everything is back to normal.


Render info showed 2 gpus, 12 cpu cores and 4 job threads under any combination of api or driver.

Game runs great under 13.12 and dx11.1 so everything is cool. I'm optimistic over what the future might provide and happy for the people who made out like fat cats.
 
Last edited:
My experience so far (i5-2500K @ 4.3GHz, MSI HD7950, 8G DDR3-1866,1080P, mostly Ultra settings, Win7pro x64): I ran a 10 minute run on Siege of Shanghai under both Mantle and DX11, logging the runs with the built in performance monitor. DX11 was ~66fps and Mantle gave me ~73fps. An increase of ~10.6%. HOWEVER, in the latter few minutes of the run, Mantle starts to hiccup and stutter annoyingly bad.

Checking the log file shows that the Max CPU Time keeps spiking, as high as 776ms (~1.3fps minimum), frequently (40+ instances over 10min) saw spikes above 100ms (10fps) whereas the Max GPU Time only got as high as 47ms (~21fps minimum). At first i thought it may be another background program using the cpu to do misc work, so i closed programs and minimized the number of necessary startup programs and retried. Still get the awful stutter. I set BF4 to High Priority in the task manager to prioritize it over other cpu usage and still get the stutter. The Render.DrawScreenInfo option shows 3 job threads on my 4core processor. Not sure what the issue is.

My DX11 runs, while having a lower average fps, gave me much more consistent frame rates. Max CPU Time only spiked up to 54ms (18.5fps min), Max Frame Time up to 44ms (22.7fps min), and Max GPU Time up to 27.9ms(35.8fps min)
Perhaps this is part of the 7000 series optimizations they mention, but it seems like it isn't the GPU that is having issues because GPU Time under each API is low without spiking. I think if i can fix the CPU Time spikes, i'll see a bigger performance increase under Mantle. I'll see if i can force it to use a 4th cpu thread and retry also. As it stands, i think i'm going to wait for another patch or two before i really switch over to Mantle.

Maybe its an issue that manifests itself over time? Others are speculating that its a PunkBuster issue, where its trying to take a system screenshot. Some also suspecting GPU clocks changing due to power requirements or heat.

Can you run GPU-Z and compare your VRAM usage between DX11 and Mantle? Even in the test server, I get hitching in Mantle and I have found that my VRAM usage is way higher, on my 7870.

*here is the post where I show it:
http://hardforum.com/showpost.php?p=1040596739&postcount=76
 
@Cageymaru, i used the DDU tool to clean out old drivers, rebooted, and then installed Cat14.1 before my testing.

@chameleoneel, i'll update GPU-Z and give VRAM a check.

Also going to see if i can force a 4th cpu job thread or try setting BF4's core affinity to the first 3 cores and punkbuster to the 4th core to try and see if its the 'punkbuster screenshot' theory.
 
In my experience, restricting BF3 and BF4 to certain core affinity kills performance.
 
From what I read on a single screen single GPU config it seems ALMOST that the 7xxx series cards are behaving better than the Hawaii cards. Just from what I've read on people's responses. Even though those were the cards that were supposedly not optimized. lol
 
There appears to be a mantle vram leak that is causing the system eat up massive amounts of memory and to swap out more frequently in certain cases.

This causing the prolonged lags after a certain abount of time.

Supposedly, dice is aware of this and working on it.
 
There appears to be a mantle vram leak that is causing the system eat up massive amounts of memory and to swap out more frequently in certain cases.

This causing the prolonged lags after a certain abount of time.

Supposedly, dice is aware of this and working on it.


Thats good to hear. I just ran a match with GPU-Z logging and it does appear i'm maxing out the VRAM (bouncing between 2844MB and 2912MB on my 3Gig 7950) on Siege of Shanghai. I noticed my GPU clock dropping from the overclocked 1100MHz setting down to 1007MHz, and a few times down to 500MHz and 300MHz (idle speed). GPU temps never went over 68c and VRM temp never over 70c, so i dont believe there is any actual throttling going on.
 
Thought I'd throw in my 2 cents worth since I'm trying to figure out how I'm seeing people with similar systems post gains with Mantle, and yet I'm seeing a loss.

PC Stats:
Core i7 2700k stock speed
16gb Kingston PC3-10700
Asus Maximus IV Extreme-Z
Sapphire HD 7770 1GB stock speeds

The data: (apologies for the largeness of the images)

BF4 set to Medium preset

Average FPS
Avg.FPS.Chart.PNG


FPS Linechart
FPS.Linechart.PNG


Frame Times Linechart
Frame.Times.Linechart.PNG


And if anyone wants to see all the data collected, here's my spreadsheet.
https://dl.dropboxusercontent.com/u/37751736/Mantle/Zavod.311-Comparison.Data.xlsx

Any input is welcome.
 
Thought I'd throw in my 2 cents worth since I'm trying to figure out how I'm seeing people with similar systems post gains with Mantle, and yet I'm seeing a loss.

PC Stats:
Core i7 2700k stock speed
16gb Kingston PC3-10700
Asus Maximus IV Extreme-Z
Sapphire HD 7770 1GB stock speeds

Any input is welcome.

According to what I've read, the only three cards that are fully supported (right now) are the 260, 290, and 290x, in a single card configuration. I think anyone else reporting gains hasn't really tested their card in a large multiplayer match, or they run at really low settings that hardly use any VRAM.

Like thebufenator a few posts up. If he's getting no hitching, he's either not using mantle, or he's not running at a high resolution. With two of those cards he should be able to run 1440p.

EDIT:And I forgot to mention, even when we (7xxx owners) get full support, you're still not going to get mindblowing gains. Mantle doesn't help a lot with GPU-limited scenarios, with BF4 at least. The mantle benchmarks with the 260 showed a less than 5% gain in one scenario. The 260 is the new budget card and is pretty similar to your 7770
 
Last edited:
According to what I've read, the only three cards that are fully supported (right now) are the 260, 290, and 290x, in a single card configuration. I think anyone else reporting gains hasn't really tested their card in a large multiplayer match, or they run at really low settings that hardly use any VRAM.

Like thebufenator a few posts up. If he's getting no hitching, he's either not using mantle, or he's not running at a high resolution. With two of those cards he should be able to run 1440p.

My system is:
FX8320
2x 7970
4x4GB
2560x1440 96hz display

Using mantle at 1440P Ultra (No AA) I had zero hitching.

The first time I went to Cat 14.1, I had to remove them due to hitching every few seconds. went back to 13.12

Then I rebooted to safe mode, DDU'd the drivers, booted normal, installed 14.1, test range, set mantle, quit, then jumped into a server.

No more hitching, HOWEVER, I was getting artifacts on the desktop when mining scrypt. If no mining was going on, artifacts were not present. Blocky artifacts similar to memory artifacting while overclocking. Not present while gaming in BF4 with Mantle.
 
Hmmm, well I guess your 6GB total VRAM is enough to overcome any ineffeciencies on that end. I imagine you'll be able to downsample or add AA once they optimize the drivers.
 
Okay, graphed my data logs.
System Info: i5-2500K @ 4.3GHz, 8G DDR3-1866 RAM, MSI HD7950 @1100/1250MHz, Win7Pro-x64, 1080p.
Recorded on a full 64 man Siege of Shanghai server.
Game Settings:
ScreenshotWin32-0002.png

=====================================================================

VRAM usage as logged by GPU-Z:
D3D11 (only using ~2100MB)
GPUZ%20D3D%20GPU%20VRAM.png


Mantle (maxing out 3Gig card)
GPUZ%20Mantle%20VRAM.png

=====================================================================

CPU Time using BF4 built-in tool (Lower is better):
D3D11 - 5/65.28/19.98 min/max/avg
D3D%20CPU%20Time1.png


Mantle - 5/1716.62/19.62 min/max/avg (Scale different due to such high spikes)
Mantle%20CPU%20Time1.png

=====================================================================

Frame Time using BF4 built-in tool (Lower is better):
D3D11 - 8.84/46.74/19.98 min/max/avg --> ~50fps avg
D3D%20Frame%20Time1.png


Mantle - 8.53/2105.5/19.56 min/max/avg --> ~51fps avg
Mantle%20Frame%20Time1.png
 
Last edited:
Back
Top