The AMD Mantle Thread

[krumme]

This should be the first independent test:
http://www.hardcoreware.net/battlefield-4-mantle-kaveri-performance/

Thanx. Its a fine test, but is it MP or SP?

What is shows is imho medium and 1080p is to long a stretch for kaveri mantle or not.

I think there is a solid improvement where it mappers at the 1% worst situations. A 15-19% improvement here. The 720 mantle graph looks really, really smooth. I think personal subjective test is nessesary to assess the improvement. A good start for a beta, i could use that on my machine. lol.

 

[sontin]

On the other hand the frametimes of the 1080p setting looks much better with DX...

 

[Gloomy]

That's disappointing. Seems like, even if there's a gain it'll be useless without a proper non-stuttering driver. So no matter what, even more waiting.

Still, 10% is probably within the margin of error for what Dice got, considering variance due to testing methodology (this isn't a canned benchmark), on the APU.

 

[DamnedLife]

That is just quirky, unreleased beta driver, what will we get is better beta driver.

 

[Abwx]

Still, 10% is probably within the margin of error for what Dice got, considering variance due to testing methodology (this isn't a canned benchmark), on the APU.

I wouldnt take as a significant result a test done with the 7850k
iGPU wich is already heavily bandwith starved even with DX11x,
the same APU with an average dGPU would me more indicative.

 

[krumme]

On the other hand the frametimes of the 1080p setting looks much better with DX...

I dont think so, we have some spikes up, but the buttom stays more consistent. At 1080 and medium kaveri is severely gpu bound, and bandwith starved to the extreme, and still we have a 6% overall improvement. This test clearly shows there is more than cpu bound optimizations and that they are noticiable. Anyway 1080 is not the important for Kaveri, its not there at least for medium.

I have no question playing bf4 MP with mantle on kaveri on 720 is a much better experience. In big 64 maps the 1% percentile could easily increase to perhaps 30% improvement.

And still is this sp or mp?

 

[krumme]

That is just quirky, unreleased beta driver, what will we get is better beta driver.

Naa. I think we will get the same beta

The 720 graph clearly show its a completely new rendering pipeline. On a first time API and driver. Ofcource there is more comming in the nexts months. I guess most of us wants the 60fps to be there all the time, and a lot also the 120. For us enthusiats mantle looks very good. And good for more pro playing.

 

[desprado]

Naa. I think we will get the same beta

The 720 graph clearly show its a completely new rendering pipeline. On a first time API and driver. Ofcource there is more comming in the nexts months. I guess most of us wants the 60fps to be there all the time, and a lot also the 120. For us enthusiats mantle looks very good. And good for more pro playing.

yup it is same beta given to first check and review.

 

[beginner99]

As it is right now, Mantle could NEVER replace DirectX, as it has no backward compatibility, nor cross architectural compatibility. But if it cannot replace DirectX, what good is it?

DirectX isn't backward compatible at all too. If you card does not support a specific version you can't use it and for each version, the game developer must create a different render path. So no, I do not really see the backward compatibility especially fi you add in that you must also have the right version of windows, namley 8.1 for all features regardless of what hardware you have. Mantle actually does better in OS compatibility as it works the same in windows 7 and 8.

 

[DamnedLife]

Battlefield 4 Medium Quality 720P 1080P
AMD A10 7850K (IGP) DX11 50 28
AMD A10 7850K (IGP) Mantle 57 31
AMD A10 7850K (R9-290X) DX11 102 100
AMD A10 7850K (R9-290X) Mantle 162 162


Battlefield 4 Ultra Quality DX11 720P 1080P
AMD A10 7850K (IGP) DX11 20 10
AMD A10 7850K (IGP) Mantle 21 10
AMD A10 7850K (R9-290X) DX11 90 78
AMD A10 7850K (R9-290X) Mantle 150 86

From Guru3D

 

[BrightCandle]

DirectX is fully backwards compatible, which effectively means if you have a game based on an old version of the API it will still run, which it will. Its not forward compatible, old games don't see all the improvements made in future versions of DX because the API keeps changing, but then that would be impossible.

On the Starswarm benchmark I just saw a 300% performance improvement of using Deferred Context, a DX11 feature that only Nvidia has implemented that improves multithreading at the API level. The AMD cards seem to be crashing (could do with that being confirmed). I would now quite like to see BF4 add deferred context so it can be compared to Mantle, as that would put the two on a more equal footing in regards to multithreading.

 

[desprado]

Battlefield 4 Medium Quality 720P 1080P
AMD A10 7850K (IGP) DX11 50 28
AMD A10 7850K (IGP) Mantle 57 31
AMD A10 7850K (R9-290X) DX11 102 100
AMD A10 7850K (R9-290X) Mantle 162 162


Battlefield 4 Ultra Quality DX11 720P 1080P
AMD A10 7850K (IGP) DX11 20 10
AMD A10 7850K (IGP) Mantle 21 10
AMD A10 7850K (R9-290X) DX11 90 78
AMD A10 7850K (R9-290X) Mantle 150 86

From Guru3D

Where is the link?

 

[Gloomy]

Where is the link?

http://www.guru3d.com/news_story/amd_mantle_update_driver_released_to_press.html

 

[desprado]

Battlefield 4 Medium Quality 720P 1080P
AMD A10 7850K (IGP) DX11 50 28
AMD A10 7850K (IGP) Mantle 57 31
AMD A10 7850K (R9-290X) DX11 102 100
AMD A10 7850K (R9-290X) Mantle 162 162


Battlefield 4 Ultra Quality DX11 720P 1080P
AMD A10 7850K (IGP) DX11 20 10
AMD A10 7850K (IGP) Mantle 21 10
AMD A10 7850K (R9-290X) DX11 90 78
AMD A10 7850K (R9-290X) Mantle 150 86

From Guru3D

Score at 1080p are really disappointed to me.
What is the logic of using R9 290X with medium settings.
At ultra there there is not much difference in result on 1080p

I dont think there people that will use cheap processor like 7850k with R9 290X.Core I7 is the right path for it.

 

[DarkKnightDude]

Sounds like Guru3d kind messed up the scores, that makes no sense, without mantle results

 

[Gloomy]

http://www.hardwareluxx.de/index.ph...te-ergebnisse-im-kampf-mantle-vs-directx.html

i7-3960X + R9 290X. Seems you do only get 10% when GPU limited, give or take a bit. Not bad, not really anything amazing. I want to see frametime graphs though.

10% is definitely not in the range of noticeable, for me at least. The minimums have to be massively better/more consistent for me to case about such a small average bump.

 

[desprado]

http://www.hardwareluxx.de/index.ph...te-ergebnisse-im-kampf-mantle-vs-directx.html

i7-3960X + R9 290X. Seems you do only get 10% when GPU limited, give or take a bit. Not bad, not really anything amazing. I want to see frametime graphs though.

Dam only 4 fps difference in 1600p and at 4k only 2 fps wow that is true revolution.


now that is sure that AMD Mantle will only and mostly benefit APUs.

 

[Gloomy]

http://www.guru3d.com/articles_pages/amd_mantle_preview.html

Again no frametime. When GPU-limited, only 10%. And no testing of mainstream gamer CPUs either (FX, 4670K)

 

[desprado]

http://www.guru3d.com/articles_pages/amd_mantle_preview.html

Again no frametime. When GPU-limited, only 10%. And no testing of mainstream gamer CPUs either (FX, 4670K)

Gloomy u have to know that games like BF4 there no difference in 3960x and 3770k.

As what guru3d says
The gains might now be massive, but at enthusiast level these remain to be significant and intersting gains alright. Above is the final example of how I think you guys play the game right now, at this very moment. Obviously 1440P is 2560x1440 WQHD btw and even there you'll see a 10% performance increase.

But 10% free performance is also considered as a massive achievement.

 

[DamnedLife]

Dam only 4 fps difference in 1600p and at 4k only 2 fps wow that is true revolution.


now that is sure that AMD Mantle will only and mostly benefit APUs.

Single gpu @4K=>BOTH DX and MANTLE=>Not playable, 2 fps increase doesn't even matter, will you play it like that NO
Doble gpu @4K=>DX=>Average, not great at ultra because CPU bottleneck, barely playable at med to high
Doble gpu @4K=>MANTLE=>Gains will be much higer like %58 from DICE bench, will make it possible to play at ultra at those res. No independent bench as of yet but sure it is on the way. Increase will be huge considering Mantle is proven to be great at removing bottlenecks.

 

[Dresdenboy]

Dam only 4 fps difference in 1600p and at 4k only 2 fps wow that is true revolution.


now that is sure that AMD Mantle will only and mostly benefit APUs.

Well, it's no news that Mantle doesn't help much in a GPU limited scenario.

Computerbase saw an 18.6% increase for a i7-4770K OC with R9 290X:
http://translate.google.com/transla...efield-4-mit-mantle-erster-eigener-benchmark/

CPU is an overclocked to 4.5 GHz Core i7-4770K is used, which renders on Radeon R9 290X in Uber mode . In the game we activated the resolution of 1,920 × 1,080 with maximum details, 4xMSAA, FXAA and 16x anisotropic filtering.

 

[Gloomy]

Gloomy u have to know that games like BF4 there no difference in 3960x and 3770k.

Yeah, I'm not expecting much. Just saying it would be nice to see a proper performance review, to see how much it scales with cores and HT.

Not going to complain about 10% free performance (I have 4770K and single card) but I am pretty disappointed.

Interesting to note, but I think that 10% performance difference is likely coming from worst case cratering of frame rates in DX. Which is why it diminishes at 4k+AA as your avg frame rate approaches your worst minimums...

 

[DarkKnightDude]

Well, it's no news that Mantle doesn't help much in a GPU limited scenario.

Computerbase saw an 18.6% increase for a i7-4770K OC with R9 290X:
http://translate.google.com/transla...efield-4-mit-mantle-erster-eigener-benchmark/

That's actually pretty good all things considering.

 

[jj109]

I would now quite like to see BF4 add deferred context so it can be compared to Mantle, as that would put the two on a more equal footing in regards to multithreading.

BF4 has deferred context on both AMD+NV and command lists on NV GPUs. The really interesting benchmark here would be NV vs AMD in CPU-bound scenes.

 

[ShintaiDK]

http://www.hardwareluxx.de/index.ph...te-ergebnisse-im-kampf-mantle-vs-directx.html

i7-3960X + R9 290X. Seems you do only get 10% when GPU limited, give or take a bit. Not bad, not really anything amazing. I want to see frametime graphs though.

10% is definitely not in the range of noticeable, for me at least. The minimums have to be massively better/more consistent for me to case about such a small average bump.

Wow those numbers are far from the 50%+. It could just aswell been an optimized driver.