Batman Arkham City Review, DX11 Explored
Graphics: DX9 vs. DX11
Arkham City's graphics are a good news / bad news situation. The good news is that the game looks great in DX9. The bad news is that even after waiting several weeks for a DX11 patch to arrive, the advantage of using that mode is surprisingly small.
We tested DX9 vs. DX11 by running the built-in benchmark and using FRAPS to capture screenshots every second. Detail levels were set to 'Very High' in both cases. Ambient occlusion was enabled, FSAA was disabled, PhysX was on and set to normal. Tessellation was also set to normal in DX11 mode.
The fog in the images below is created by PhysX. All DX9 screenshots are on the left, DX11 is to the right. We went to some trouble to compare the DX9 vs. DX11 shots to ensure we weren't missing effects, even using AMD's "The Compressonator" tool to highlight the exact differences between images.
There's virtually no difference between these two in DX9 and DX11
Above is as perfectly identical a screengrab as we could manage. The only apparent difference between DX9 and DX11 is that the DX9 shot is brighter.
PhysX is used well, but again, the difference here is minimal
Another comparative set. We're guessing that the randomized hairdos are part of how the game randomizes NPC appearances. Again, virtually the only difference between the two screenshots is the general level of brightness, although the lens flare from the light in the back-right corner is somewhat larger in the DX11 version. Normally, DX11 is used to create more realistic shadows, but the lighting change is global rather than shadow-specific.
This is an area where DX11's tessellation would be expected to make a significant difference
A giant plant. This is one area where we might expect to see tessellation make a serious difference, but there's none to be seen--at least not here. Watch what happens as the camera continues panning and we pass through the grate you can see to your left.
Here, it does -- but notice the limited view and short camera distance.
Holy major tessellation, Batman! That's a heck of a difference. So why is it all focused in just this one spot? Check the previous set of images, and there's zero evidence of anything like this. Suddenly, in one area, boom. We don't know if Rocksteady was being lazy or deliberately trying to create a false impression of what DX11 delivers in Arkham City, but there's virtually nothing this dramatic in the rest of the game.
Here's a more realistic example of how tessellation is used.
Check the curve of the lion's back and the area under his chin, around the mane
Compare the lion sculpture in DX9 vs. DX11 and you'll note that the DX11 version is more organic and less angular. The stone appears to have been shaped rather than struck off at sharp angles. Sitting at the computer flipping between the two modes, the DX11 version is clearly superior—but you'll never notice in-game.
Here's the last, and arguably most-important comparison series from the benchmark. As always, DX9 is on the left, DX11 on the right.
DX9 Mode, Left - DX11 Mode, Right
That's from a system using a pair of GTX 480's in SLI. We have no idea where the extra performance is going, because it sure as heck isn't going into rendering much on screen. While the DX11 system doesn't spend much time at the minimum frame rate, it's enough to notice.
In this case, given the choice between turning on features like PhysX and AA or running in DX11, we'd choose DX9 and take PhysX and AA all day long.
We tested DX9 vs. DX11 by running the built-in benchmark and using FRAPS to capture screenshots every second. Detail levels were set to 'Very High' in both cases. Ambient occlusion was enabled, FSAA was disabled, PhysX was on and set to normal. Tessellation was also set to normal in DX11 mode.
The fog in the images below is created by PhysX. All DX9 screenshots are on the left, DX11 is to the right. We went to some trouble to compare the DX9 vs. DX11 shots to ensure we weren't missing effects, even using AMD's "The Compressonator" tool to highlight the exact differences between images.
;)
;)
There's virtually no difference between these two in DX9 and DX11
Above is as perfectly identical a screengrab as we could manage. The only apparent difference between DX9 and DX11 is that the DX9 shot is brighter.
;)
;)
PhysX is used well, but again, the difference here is minimal
Another comparative set. We're guessing that the randomized hairdos are part of how the game randomizes NPC appearances. Again, virtually the only difference between the two screenshots is the general level of brightness, although the lens flare from the light in the back-right corner is somewhat larger in the DX11 version. Normally, DX11 is used to create more realistic shadows, but the lighting change is global rather than shadow-specific.
;)
This is an area where DX11's tessellation would be expected to make a significant difference
A giant plant. This is one area where we might expect to see tessellation make a serious difference, but there's none to be seen--at least not here. Watch what happens as the camera continues panning and we pass through the grate you can see to your left.
;)
;)
Here, it does -- but notice the limited view and short camera distance.
Holy major tessellation, Batman! That's a heck of a difference. So why is it all focused in just this one spot? Check the previous set of images, and there's zero evidence of anything like this. Suddenly, in one area, boom. We don't know if Rocksteady was being lazy or deliberately trying to create a false impression of what DX11 delivers in Arkham City, but there's virtually nothing this dramatic in the rest of the game.
Here's a more realistic example of how tessellation is used.
;)
;)
Check the curve of the lion's back and the area under his chin, around the mane
Compare the lion sculpture in DX9 vs. DX11 and you'll note that the DX11 version is more organic and less angular. The stone appears to have been shaped rather than struck off at sharp angles. Sitting at the computer flipping between the two modes, the DX11 version is clearly superior—but you'll never notice in-game.
Here's the last, and arguably most-important comparison series from the benchmark. As always, DX9 is on the left, DX11 on the right.
DX9 Mode, Left - DX11 Mode, Right
That's from a system using a pair of GTX 480's in SLI. We have no idea where the extra performance is going, because it sure as heck isn't going into rendering much on screen. While the DX11 system doesn't spend much time at the minimum frame rate, it's enough to notice.
In this case, given the choice between turning on features like PhysX and AA or running in DX11, we'd choose DX9 and take PhysX and AA all day long.
