Battlefield 4 Patch 2/13/2014 with Mantle improvements

[VulgarDisplay]

It is a shame the IQ difference is still high. Draw distance, fog, gamma etc. Obviously not as easy to fix, or the performance impact is too great

The only map where I saw any draw distance issues was the Test Range. There was also something wrong because it showed less detail in test range, but in the actual game the difference wasn't there.

I think something is just funky in the test range.

EDIT: I Figured out why the draw distance issue was in my Mantle screenshot in test range. I took the SS too fast and the tanks hadn't spawned yet on the left side.

Updated SS's

Direct X 11

Mantle (Post AA is off in this SS and HBAO is on)

The LOD on the weapon/hand is back to normal also. Sorry the settings are different in the new picture. I just wanted to show that the draw distance is actually fine.

 

[wand3r3r]

What's the consensus on the latest patch?

I wish the 14.1 drivers worked for mining. I believe my new MB is coming today so I'll be doing a fresh install of windows again, I'm debating testing 14.1 for a bit in BF4 before rolling back to 13.12 for mining.

 

[ShintaiDK]

Techreport did a test with the latest patch on their A8 7600 system.

http://techreport.com/review/26042/a-quick-look-at-mantle-on-amd-kaveri-apu/2

 

[el etro]

Updated SS's

LOD is equal in both!

 

[jason166]

Just wanted to chime in and say that the latest patch has mostly fixed the lag spikes on my R9 270 (Basically a 7870) and it is now playable. There are still on occation some hiccups I think are attributable to the mantle driver but this is a huge improvement.

I have not run any performance tests but my general impression is there's a reduction in input lag that I can sense when I'm in mantle API mode. Net result is I seem to do better in game.

Regards,
Jason

 

[kane15]

btw for some reason mantle looks alot brighter in screenshots if you lower in game settings it will look the same as DX

 

[f1sherman]

btw for some reason mantle looks alot brighter in screenshots if you lower in game settings it will look the same as DX

No it won't.

AA is broken, HDR is broken. Washed out DX9 look can't be fixed with brightness

 

[VulgarDisplay]

No it won't.

AA is broken, HDR is broken. Washed out DX9 look can't be fixed with brightness

AA isn't broken, it just shifts the bottleneck back to the GPU drastically cutting performance.

DX11

Mantle

When switching back and forth between the two quickly on my desktop, mantle actually shows more detail in the sand up close. The washed out look can absolutely be fixed with gamma and contrast settings. I put my screenshots into Lightroom and upped the constrast 50% and put 100% saturation on the blues and the differences are largely gone(These are not the ones I changed.)

 

[5150Joker]

It is a shame the IQ difference is still high. Draw distance, fog, gamma etc. Obviously not as easy to fix, or the performance impact is too great

I suspect if the IQ were equal, the Mantle gains would be substantially less.

 

[krumme]

I suspect if the IQ were equal, the Mantle gains would be substantially less.

Mantle is dangerous !

 

[VulgarDisplay]

I suspect if the IQ were equal, the Mantle gains would be substantially less.

The only difference is color grading. LOD is nearly identical between mantle and dx11. I spent like 2 minutes in Photoshop and upped blue saturation to like 100% on a mantle screenshot and it looks a lot more like the dx11 side.

The dx11 while more pleasing to the eye looks far less like the real world version of a blue sky than the mantle version.

 

[itsmydamnation]

I suspect if the IQ were equal, the Mantle gains would be substantially less.

i suspect most people who come in here with this comments know nothing of what they are talking about and are jsut hear to spread FUD.

People who have played battlefield will know that the colour settings are per map. in BF3 you could use
http://battlefield.realmware.co.uk/colour-tweaker/
this includes
Brightness:
R,G,B (separate values)
Contrast:
R,G,B (separate values)
Satuation:
R,G,B (separate values)
Hue
Colour grading
Sunlight intensity:
R,G,B (separate values)

the logical explanation is that mantel either doesn't interpret these values exactly the same or it uses different values. In BF3 changing these values had no performance impact.

Also on the Aliasing front, make sure you understand what Aliasing actual IS (http://www.beyond3d.com/content/articles/122/)

to me AA looks very close if not identical, what you are seeing is the difference in contrasts between the sky and objects (trees, etc) is greater on mantle creating more pixel to pixel variance compared to be more saturated DX11 render.

As to which one is better.... i like the "FAKE" DX11 more saturated colour, Hopefully DICE will get them inline.

 

[VulgarDisplay]

In my last set of ultra screenshots my 7970 was running stock clocks because afterburner was not behaving properly and my fan would stick at 50% on desktop. I believe the dx11 shot I took earlier was at 1180mhz on the core.

 

[Teizo]

it almost looks like the contrast is set a lot lower on the mantle screen shots

Looks to me from those screenshots that Mantle does not have as much detail as DX in regards to tessellation and DX looks more crisp.

May just be those two screenshots, but if that is consistent throughout, as it stands right now DX has better IQ vs Mantle which for PC gaming is the most important thing because IQ with performance is what sets PC's apart from consoles mostly. That doesn't mean that Mantle can't match DX IQ, but from those shots it looks like it isn't atm.

 

[5150Joker]

i suspect most people who come in here with this comments know nothing of what they are talking about and are jsut hear to spread FUD.

People who have played battlefield will know that the colour settings are per map. in BF3 you could use
http://battlefield.realmware.co.uk/colour-tweaker/
this includes
Brightness:
R,G,B (separate values)
Contrast:
R,G,B (separate values)
Satuation:
R,G,B (separate values)
Hue
Colour grading
Sunlight intensity:
R,G,B (separate values)

the logical explanation is that mantel either doesn't interpret these values exactly the same or it uses different values. In BF3 changing these values had no performance impact.

Also on the Aliasing front, make sure you understand what Aliasing actual IS (http://www.beyond3d.com/content/articles/122/)

to me AA looks very close if not identical, what you are seeing is the difference in contrasts between the sky and objects (trees, etc) is greater on mantle creating more pixel to pixel variance compared to be more saturated DX11 render.

As to which one is better.... i like the "FAKE" DX11 more saturated colour, Hopefully DICE will get them inline.

If it is such a simple issue, why hasn't DICE and AMD addressed it? Why release Mantle in this state? It's not like they had to get Mantle working for a dozen games or even 5 games or even 2 games, just ONE game. :hmm:

 

[bystander36]

AA isn't broken, it just shifts the bottleneck back to the GPU drastically cutting performance.

When switching back and forth between the two quickly on my desktop, mantle actually shows more detail in the sand up close. The washed out look can absolutely be fixed with gamma and contrast settings. I put my screenshots into Lightroom and upped the constrast 50% and put 100% saturation on the blues and the differences are largely gone(These are not the ones I changed.)

Looking at the screen shots at full size, the differences are mainly just how bright the colors are. I agree that it really isn't an IQ difference and just slightly different in color. I can't even say which really looks better, other than when it is smaller, I prefer the brighter blues of the DX11 version. In fullscreen, I don't really care one way or the other.

 

[bystander36]

If it is such a simple issue, why hasn't DICE and AMD addressed it? Why release Mantle in this state? It's not like they had to get Mantle working for a dozen games or even 5 games or even 2 games, just ONE game. :hmm:

The slight color difference isn't really a problem. It isn't something that is clearly better or worse, just slightly different. They have more important things to do than try to make the two API's look identical.

You'll notice that games run on AMD vs Nvidia have similar slight differences. Which is really better isn't clearly defined and generally a non-issue.

 

[PPB]

Next tuesday there is a new BF4 PC patch that includes the new DLC content (yay for more Metro.... /NOT). Repi said the contrast/fog bug didnt make it to this last patch, so it might be on next week.

So if I were some of you, I would start drumming that "less IQ on mantle" even louder for this last week, so when we would see totally similar SCs of the game under both renderers you would look even sillier

 

[VulgarDisplay]

Looks to me from those screenshots that Mantle does not have as much detail as DX in regards to tessellation and DX looks more crisp.

May just be those two screenshots, but if that is consistent throughout, as it stands right now DX has better IQ vs Mantle which for PC gaming is the most important thing because IQ with performance is what sets PC's apart from consoles mostly. That doesn't mean that Mantle can't match DX IQ, but from those shots it looks like it isn't atm.

Do you know what tesselation is? I see nothing in those shots that indicate a difference other than the color grading. The slight differences in foliage are due to the fact that it is swaying and I took the shots at different times during the sway.

 

[3DVagabond]

The only map where I saw any draw distance issues was the Test Range. There was also something wrong because it showed less detail in test range, but in the actual game the difference wasn't there.

I think something is just funky in the test range.

EDIT: I Figured out why the draw distance issue was in my Mantle screenshot in test range. I took the SS too fast and the tanks hadn't spawned yet on the left side.

Updated SS's

Direct X 11

Mantle (Post AA is off in this SS and HBAO is on)

The LOD on the weapon/hand is back to normal also. Sorry the settings are different in the new picture. I just wanted to show that the draw distance is actually fine.

The most telling difference between the 2 shots is the smoothness of the GPU graph.

 

[VulgarDisplay]

The most telling difference between the 2 shots is the smoothness of the GPU graph.

That smoothness translates to what you see on screen. Faster but smoother is now a thing I guess.

It should be known that seemingly at random I do get periods of about 5 seconds of really bad performance for whatever reason. I don't know if it's something my CPU is doing in the background or what. The CPU portion of the graph will just go haywire for those 5 seconds and then it will return to normal and be fine again for however long.

 

[itsmydamnation]

If it is such a simple issue, why hasn't DICE and AMD addressed it? Why release Mantle in this state? It's not like they had to get Mantle working for a dozen games or even 5 games or even 2 games, just ONE game. :hmm:

sigh....

first who said it was broken?
second who said it was easy?

The maths that goes into modern games is very complex. Its very complex because its all approximations because things like π (PI) and large large floating point numbers are expensive to computate. All you need is something in your many layers of calculations to either be rounded differently or clamped differently and the effect can be quite pronounced. All there needs to be is a difference at the hardware level between the mantle and the DX paths to end up with different results. Both are correct, just different thats why things like IEEE 754 exist for standardisation.

 

[krumme]

Played for two hours. The stuttering have gone for me, running great.

 

[MBrown]

Anyone else get the sound loop crash? I can't play for more than 1 minute before I crash with this patch...

 

[Erenhardt]

The random spikes I had under mantle are mostly gone. Now I have lesser spikes which may be caused by my old and broken HDD (smart errors).
I will try to benchmark my system:
Phenom II 965 3,4Ghz OC@3,5GHz
HD 7870 1000/1250MHz

Did test range shots:

Free performance
Looks like those spikes under mantle I was talking earlier were caused by background tasks - chrome was refreshing mining pools dashboard every 10 sec or so, which was not noticable under dx. And because mantle is taking every last bit of the available CPU, when background task kick in, it takes resources used by mantle to feed GPU. Resulting in a spike.