Skip to content

HighOmega v3.01
Compare
Choose a tag to compare
@toomuchvoltage toomuchvoltage released this 20 Feb 03:51
· 1 commit to master since this release
v3.010000
a6e8d02
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

This release is mainly a render and performance update over 3.0:

  • This release reverts back to using basic next event estimation with multiple importance sampling. As a result it is slightly darker as it has minimal bias, yet still manages to maintain its temporal stability. The de-noising procedure is almost identical to the i3D2018 version except spatial de-noising appears first. The plane-aligned component (i.e. the trapezoidal looking part) of the filter is reserved only for diffuseness to maintain contact hardening in gloss. Upsampling is done last (meaning vast majority of the filter happens at low-res) as this helps with performance. First bounce now also uses blue-noise sampling.

  • For the first bounce of gloss it attempts to do ray tracing against a fixed capacity grid similar to CryTek's Neon Noir. This has been showcased on Twitter and Reddit. More info on this later.

  • Higher resolution shadow maps everywhere. Including on lower VRAM devices previously using smaller shadow map resolutions.

  • Vast majority of assets compressed in this release using BC1. This should be reducing VRAM consumption to around 1/3rd compared to last release. Even 2GB VRAM devices (albeit at low framerates) are supported if the 2x2x2 option is selected for hybrid tri-grids and applied before running the demo.

  • Fixed an SDL bug involving Windows 10 high-DPI support.

Assets 3
Loading