Battlefield 4: AMD’s Mantle causes washed out, foggy graphics compared to … – ExtremeTech
In the rush to get Mantle out the door by the end of January, the team at Dice appears to have missed a spot when it comes to porting over the game. Display quality between Mantle and BF4 isn’t quite identical — and we’ve got screenshots of the effect in the same Golmud Highway map that we used in our Mantle benchmarking. The impact may be greater in other maps; we’ve stuck to comparing the same one we used for test purposes.
Unfortunately, other players took a fairly dim view of my screenshot attempts and made it difficult to get the truly representative shots I wanted. In the screenshots below, Mantle shots have no Fraps counter, while the DirectX 11 shots show the Fraps counter in the top left. The shots themselves were captured using in-game tools from BF4 in both cases, to avoid any additional post processing or Fraps-introduced issues. In the shots below, Mantle is on top, Direct3D is below. Images are full size, high-quality JPEGs (click to zoom in).
The difference between the Mantle shot and the Direct3D shot is twofold. First, the sky is washed out — less blue, more white. Alongside this, the mountains are a bit grayer; colors aren’t as clear. Mantle might be a little more foggy, too. Foreground shots, texture levels, and up-close details haven’t changed.
This set of comparisons shows more detail on ground cover as well as the lighting model differences. The Direct3D image has more “pop,” the color saturation is higher, and the game has more visual punch. But the actual detail and rendering workloads aren’t necessarily different — this is quite possibly a post-processing tweak. Dice has confirmed that there’s an issue that’ll be fixed with an upcoming patch, and that the bug is on its side of the code.
Why didn’t we catch it yesterday?
My original plan for Mantle was to do in-depth D3D image quality comparisons on both titles. Then AMD dropped the driver on Friday and the public driver on Sunday. In the scramble to get benchmark data, I wasn’t paying tremendous attention to the details of long-distance mountain rendering or sky boxes. The difference between the two areas, if you’re focused on the people trying to kill you, isn’t very large.
Still, I missed noticing the differences, and I apologize for not calling them out yesterday. Right now, there is a gap between the output of Mantle and D3D. The performance impact, however, should be minimal; the amount of in-game resources being devoted to rendering static, non-interactive backgrounds is always small. Given that AMD’s Mantle driver is already a very early revision, this doesn’t seem likely to change anyone’s opinion on the API. The performance figures stand — but the rest of the implementation is very rough around the edges.