summary forthcoming
- Crysis Shrine because Dictator
- old way: one to one per-pixel shading, but as resolution increase so did shading demands
- no point in shading pixels that look the same, that's what VRS fixes
- 2x2, 4x4, even 2x1/1x2 shading methods for vertically/horizontally long elements
- VRS utilizes color, brightness, even speed via motion vectors to determine shading rate
- a hardware supported feature, so it should be widely support
- Wolfenstein has a really good VRS implementation
- 10% more performance in Youngblood
- Gears Tactic: 13%+ in on setting, 30%+ in performance setting
- Gears Tactics has a harsher implementation
- uses Tier 1 variant of VRS; more limited in how it can perceive and reduce shading
- VRS here also affects the UI, which Wolf doesn't do
- VRS isn't filtered, so devs has to add a filter themselves, will look pixelated otherwise
- scenes in motion hide quality degredation
- (Gears Tactics) 9% better performance on a 2070S in cutscenes, 10% in performance mode
- less dark areas/less motion blur/etc, less VRS will help
- VRS performance scales with VRAM limitation
- reducing shading settings reduces VRS performance gains
- models with super dense triangle count (triangles smaller than the pixels) make VRS less effective
- that makes GPUs less efficient anyway
- works well with dynamic res scaler and even more with mesh shaders
- Gears Tactics is the easiest method of VRS
- checkerboard rendering stems from techniques like CSAA/EQAA and has lead to massive gains in IQ and performance
- taking advantage of VRS in non-standard ways could lead to major gains performance and IQ
- thinks like screen space AO/reflections/ect could benefit
- VRS-like techniques to improve lighting performance? could improve performance by 30%
Last edited:
