BLUEI'S BETTER VISUALS v6.15.0 VELVET AIR - WINDOWS PERFORMANCE NOTES

FACTORY COST
The untouched configuration keeps Medium water reflection and cloud quality,
Generated PBR, warm cinematic lighting, and the complete Vanilla atmosphere.
Motion blur, depth of field, chromatic aberration, tilt shift, lens flare,
parallax, and material self-shadowing are off.

CLOUD COST
Windows cloud march counts are 10 / 16 / 24 / 34 / 52 for quality levels 0-4.
Stable spatial phase hides coherent sample planes
without temporal shimmer. Early altitude, family, distance, empty-cell, solid-
core, and transmittance exits prevent unnecessary density work.

Cloud density now uses continuous world-space coordinates instead of quantized
voxel cells. The same march budgets remain in place, and the former per-sample
voxel-tone hash has been removed from the lighting path.

The new far-terrain haze is an analytic depth blend using values already present
in the final pass. It does not add a texture lookup or ray-march step.

ATMOSPHERE COST
The warm/cool scene grade, exponential distance curve, cloud-palette retune,
softer shadows, and reflected-environment palette are arithmetic-only changes.
Bloom uses the existing tap count at wider offsets. Cloud and water ray counts
are unchanged from v6.14.1.

Default terrain cloud shadows now use one low-frequency world-space sample.
Quality 3 adds a second family and slow weather variation; quality 4 adds the
third family. This reduction matters in established worlds because terrain
shading covers most of the frame.

SHADOW COST
Quality 0 uses one terrain shadow tap. Quality 1 uses three near taps and two
far taps. Entities and particles use one, two, or four taps based on quality.
Higher tiers retain wider stable Poisson kernels.

PBR COST
Fresh install uses Generated PBR. Surface normals and micro-roughness derive
from the albedo sample already fetched by terrain. Default parallax and height
self-shadowing are disabled. Entity skin atlases bypass generated PBR entirely,
which prevents both atlas seams and unnecessary material work in mob-heavy
worlds.

REFLECTION COST
Reflection quality 1-4 uses 4 / 6 / 8 / 12 coarse depth steps on macOS plus
1 / 2 / 2 / 3 refinements and final validation. Windows quality 4 uses 14
coarse steps. Invalid, sky, behind-camera, below-water, and off-screen rays exit
early. The ray is deterministic and has no temporal history pass.

The geometry layer uses a flat plane. Animated wave normals only affect the
analytic environment, sun and moon glints, foam, and refraction. This both
stabilizes reflected scenery and keeps the wave model visually active.

Quality 2 and above also attempts one guarded screen-sky capture. It performs
one depth lookup, and only reads sky color when that location is on screen and
contains sky. Geometry ray-step counts are unchanged from v6.14.0.

UNDERWATER COST
The former interpolated 3D FBM volume is absent. Default quality performs one
refracted scene lookup, Beer-Lambert attenuation, one soft directional shaft,
one low-contrast caustic evaluation, and two sparse sediment steps. Derivatives
gate caustics by surface orientation without another texture lookup. Only the
two highest tiers add spatial refraction filtering.

FPS EXPECTATION
No fixed FPS target can be guaranteed from static shader code. Benchmark the
same location, camera angle, resolution, render distance, weather, mods, and
profile. A large established world can be CPU-bound even when shader GPU cost
falls substantially.

PERFORMANCE ORDER
1. Try the untouched configuration in the same benchmark scene.
2. Select Vanilla Low if clouds and water occupy much of the frame.
3. Lower Reflection Quality if water dominates the screen.
4. Lower Cloud Quality if open skies dominate the screen.
5. Lower render distance or resolution if every profile remains similarly slow.
