I am currently testing the Performance difference between UE4 and UE5. In this case, I converted my project from UE4 to UE5. (Newest Drivers)
This scene doesn’t have moving parts, it’s only an indoor scene with a direction light, a few non-shadow casting point lights, no Skylight. Everything is fully dynamic, nothing baked.
Distance Field are activated, the Dircation Light uses DF-Shadows, no cascaded ones.
In UE5 I reduced the Distance Field Voxel Density from 0,2 to 0,05, which resembles more the UE4 density. Changing this made no Performance difference.
No asset got converted to Nanite, no asset is using WPO, no skeletal meshes, camera is fixed/steady.
I know, that UE5 automatically renders everything in a lower resolution. (I suppose 54%?)
For this Performance comparison, UE4 is rendering everything in Native 4k, and UE5 is set to render everything in native 4k.
Only one Project is opened at a time, no background tasks which may distort the results.
I repeated these Test within a few days to get the most accurate results.
These are the UE4/UE5 Project Settings:
Dynamic Global Illumination Method: None (no Lumen or SSGI)
Reflection Method: Screen Space
Shadow Map Method: Shadow Maps (no VSM)
Anti-Aliasing Method: None (no TAA or TSR for accurate comparison)
Here are the performance comparisons:
Here are a few Performance differences with different Resolutions and Project settings:
100% no AA = 13ms
100% TAA = 13ms
54% TAA no upscaling = 8ms
54% TAA with upscaling = 8ms
100% no AA = 18ms
100% TAA = 18ms
54% no AA = 11-12ms
54% TAA no upscaling = 12ms
54% TAA with upscaling = 12ms
54% TSR = 12-13ms
Why is there still this big of a performance difference?
Which settings should change to get back to UE4 Performance?
I have tried console setting: “r.nanite 0” which probably help a few in the forums with the EA Build, but in 5.0 there is no difference.
Deleting volumetric clouds made no performance difference between UE4/UE5.
What is still missing?
Fortnite seems to do well in UE5, this cant be the end?