NVIDIA GeForce RTX Ray Tracing In Battlefield V Explored Pre And Post Patch
Test Setup, RTX Explanation, Game Settings and Hardware Tested
|
|
|
| Hardware Used: Intel Core i9-9900K (5.1GHz 8-Core) ASUS Maximus XI Extreme (Intel Z390 Chipset) GeForce RTX 2080 Ti Founders Edition GeForce RTX 2080 Founders Edition GeForce RTX 2070 Founders Edition 32GB Corsair DDR4-3200 Samsung SM961 512GB Samsung 860 EVO 4TB Integrated Audio & Network |
Relevant Software: Windows 10 Pro x64 - ver. 1809 NVIDIA Drivers: v416.94 (Pre Patch) v417.22 (Patched) Benchmarks Used: Battlefield V |
These are the cards we are testing today. These will be representative of the relative performance you can expect when playing Battlefield V on your rig only if you mirror our components exactly. Otherwise this should only be used as a rough guideline for what you can expect to see.
For each card we used the following resolutions.
- RTX 2080Ti - 2560x1440
- RTX 2080 - 1920x1080 (Pre & Post Patch) 2560x1440 (Post Patch)
- RTX 2070 - 1920x1080
These resolutions were chosen as they provided the best playable experience for the respective GPU. All other settings except resolution are shown below. Post patch testing showed that the 2080 actually leveled up a bit and got enough of a boost at 1440p to make it viable for this GPU, so we included it in both resolutions.
Note: Film grain, vignette, Chromatic aberration and Lens Distortion were disabled to ensure we could best show the lighting effects without any strange or aberrant behavior. For example, film grain would cause graininess, which could be mistaken as a Ray Tracing artifact and negatively impact the visuals during our capture.
Either Low or Ultra settings were used.
As you can see from the above screens, DXR will be tested with the following settings.
- DXR Off
- DXR Low
- DXR Ultra
The reason we test this way is that DICE and NVIDIA have confirmed that any setting above LOW is somewhat broken in the initial implementation of DXR/RTX in BFV. That meant that medium and high did not function correctly and all of them produced similar results to Ultra. Therefore, we tested the known working parts of the game both pre- and post-patch. This means that we are testing Ultra as the worst case or most stressful scenario for a set resolution.
How the bug NVIDIA/DICE described affects the game is tricky to digest, so lets take a look at some of what RTX is doing in a brief explanation to ensure you all understand some of the processing going on behind the scenes.
- Low applies ray tracing to materials with 0.9 smoothness or higher, and sets maximum ray count to 15.0 percent of the total number of screen pixels
- Medium applies ray tracing to materials with 0.9 smoothness or higher, and sets maximum ray count to 23.3 percent of the total number of screen pixels
- High applies ray tracing to materials with 0.5 smoothness or higher, and sets maximum ray count to 31.6 percent of the total number of screen pixels
- Ultra applies ray tracing to materials with 0.5 smoothness or higher, and sets maximum ray count to 40.0 percent of the total number of screen pixels
What this tells us is that every potentially reflective object built into a scene within Battlefield V has a smoothness level. The higher the smoothness level, the more reflective it will be. A mirror would be smoothness level ideally of 1.0, so what you are seeing is that there are two things DXR looks at when deciding whether or not to cull a reflection. You have the smoothness (reflectiveness) of an item and the total overall ray count it can implement, which will be limited by the DXR level selected. Ultra will obviously be the most stressful as it allows reflectivity of surface down to a .5 smoothness with up to 40 percent of a scene being ray traced. Now, that 40 percent can be pretty important. Lets say you are in a water scene on Ultra -- you will get tons of reflective items but it will likely perform some distance based culling to show objects which are closer in the reflection. This means you may see certain textures pop in on reflections based on how many items are in the scene and / or how far away they are from the reflective surface.
You will see this behavior in the video embedded on the previous page. If you carefully watch the video, you will notice that the matte, or less glossy, ice is culled at lower quality levels to only fill the most reflective small patches of the ice with reflections. Meanwhile, in Ultra mode, far more of the ice will fit the smoothness factor needed to trace the rays of explosions / fire. The area we tested in is not ideal for showing performance improvements with the patch as it is filled with ice. Ice in the game, unlike water, is not as intensely reflective or mirror like. This means that while we may see performance increases with the new patch, the increase could be more dramatic on a scene where there is substantially more to ray tracing going on. We feel this is ideal, as it puts the new patch in somewhat of a worst case scenario, because the reflections are not pegging the meter. With massive amounts of reflections there is more than can be culled to bump up performance and cause a larger pre- and post-patch delta, but we feel the area we tested in will be representative of the kind of real world performance benefits you can get in the majority of the game.
One of the things NVIDIA and DICE did to optimize to tame the initial, heavy performance impact of RTX was to cull reflections of specific smaller items such as falling leaves or foliage where they would not have a dramatic visual impact. This drastically reduces the workload as the ray tracing can now be focused on larger, more scene appropriate reflections such as vehicles, buildings, fire, and people.
Now lets get Battlefield V fired up with RTX and see what all of this fuss is about...
