Unreal Engine 4 DXR Comparison Videos
Published:
by:
Jack 'NavJack27' Mangano
Estimated reading time: ~3 minutes
Please throw some money my way either directly VIA PayPal or with my Ko-Fi maybe don’t do my Patreon until I fully set it up.
What am I comparing exactly?
The Unreal Engine finally has development builds with DXR enabled and I am figuring out ways to compare quality between them and other builds of the engine that are out there. I decided to use the Infiltrator sample scene because I am a masochist that loves torturing my computer. This is an intense scene traditionally let alone with ray tracing or with GameWorks enabled features. We aren’t looking at performance at all with what I’m doing here, just trying to find rendering oddities and differences.
My method
Using the built-in cinematic matinee feature I can capture a movie with a fixed frame rate of the scene. I do that after I convert the project to the appropriate version of the engine and spend a bit of time fixing up the lights and other things to make the feature set work correctly due to limitations of the technology. Lights in some cases need to be changed from stationary or static to movable. I also don’t sandbag any of the versions by disabling things that are traditionally needed. Examples of this are like in the VXGI scene, I used the baked lightmaps where needed. I do attempt to highlight the strengths of each build at the same time. For example, with the ray traced scenes, I made sure there were no baked lightmaps and everything concerning the lighting was from the ray traced features.
Each video was captured in 1280x720. The first part was in 30fps, but the second part was in a full 60. These recordings are encoded to a MJPG AVI file by the engine. The videos are resized to 150% to fill the frame in the way that it does. The background, although blurred, is both the videos on top of each other with a difference blend and 300% scaling. Anti-aliasing was used in every video with the built in Unreal Engine TAA method except for the GameWorks scene which used Nvidia’s TXAA method. The final render was to a NVENC h265 video with maximum quality in Adobe Premiere using the Voukoder plugin.
Here is the codec information for each video in the production step
The Unreal Engine Export
General | |
Format | AVI |
Format/Info | Audio Video Interleave |
Format profile | OpenDML |
Overall bit rate | 255 Mb/s |
Video | |
ID | 0 |
Format | JPEG |
Codec ID | MJPG |
Bit rate | 255 Mb/s |
Width | 1280 pixels |
Height | 720 pixels |
Display aspect ratio | 16:9 |
Frame rate | 59.940 or 30 |
Color space | YUV |
Chroma subsampling | 4:2:2 |
Bit depth | 8 bits |
The Export For YouTube
General | |
Format | MPEG-4 |
Format profile | Base Media |
Codec ID | isom (isom/iso2/mp41) |
Duration | 3 min 34 s |
Overall bit rate | 18.6 Mb/s |
Video | |
ID | 1 |
Format | HEVC |
Format/Info | High Efficiency Video Coding |
Format profile | Main@L6@Main |
Codec ID | hev1 |
Codec ID/Info | High Efficiency Video Coding |
Duration | 3 min 34 s |
Bit rate | 18.6 Mb/s |
Width | 3840 pixels |
Height | 1260 pixels |
Display aspect ratio | 16:9 |
Frame rate mode | Constant |
Frame rate | 59.940 or 30 |
Color space | YUV |
Chroma subsampling | 4:2:0 |
Bit depth | 8 bits |
Color range | Limited |
Color primaries | BT.709 |
The videos
Dev-Rendering vs Preview 1 (Part 1)
VXGI vs Preview 3 (Part 2)
My thoughts
There are some issues that crop up here and there. In the part one video neither ray tracing build of the engine handles the spotlights correctly nor some of the translucent effects or particles. There are missing things in each. Turning on global illumination tends to darken the scenes. I’m guessing that is due to every material being very metallic, although I’m not sure why dynamic entities seem to be lacking bounces on their mesh.