Why does Vulkan SM6 give me a 50-100% increase in frame rate vs DX12 SM6 when using a nanite landscape? I’m not using ray tracing(supported is checked, but all the other options are turned off. I’ve tested it with and without, no difference). I am using lumen GI and reflections, along with VSM shadows. Virtual textures are also enabled. Other than that, it’s a pretty default setup on the rendering end of things.
I have a landscape scene with a 4km map. It has a single landscape layer, but I use some splatmaps scaled to fit in the material(making interactive aerial concept art where I don’t need ultra high quality at foot level). There are maybe a thousand scattered boulders and large rocks and a single layer water plane.
EDIT: I also tested with the game packaged(shipping) and I have buttons to switch between high/epic scalability presets:
Vulkan: 92/63 fps
DX12: 50/42 fps
VRAM and RAM usages are fine, not OOMing on either. 4.9gb on epic with Vulkan and 4.3gb with DX12 out of the 8gb this card has.
I’ve checked all the various debug views, like making sure the VK mode was actually using nanite and it was. Visually, they look pretty much identical.
RTX 2080 with the latest 566.03 drivers. I’ve also cleared out the shader caches for both Nvidia and the Unreal as well. Windows 11 23H2 fully updated (still waiting on it to push the 24H2 update)
Is Vulkan shortcutting something vs DX12? Is this some limitation to this old GPU like maybe it doesn’t like some type of calculation precision that DX12 is using, but Vulkan isn’t? Is this some kind of bug in the code? It’s had me scratching my head for a couple days now.