Bright scenes: sky blows out to white and PBR materials wash out to grey / white #1051
Labels
No labels
amd
avatars
bug
builder
cannot-reproduce
crash
epic
feature
from-viewer
gl-removal
known-limitation
monitoring
parity
perf
platform
priority
critical
priority
high
priority
low
priority
medium
provenance
rendering
rt
status
in-progress
status
testing
status
up-next
triage
ui
upstream
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
sky_line/Slipstream#1051
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
In a bright daytime scene the sky blows out to solid white (with heavy bloom over it), and PBR materials lose their colour and go pale grey / white. Examples: a blue PBR shirt renders light grey, and tan PBR hooves render white-grey. Blinn-Phong (legacy) parts of the same avatar (body, mane, tail) keep their colours. The ground also looks over-saturated yellow-green.
Seen on dev 1.5 (
b2761279e3), RTX 5090, DLSS RR. The owner found it while testing settings. It is not GI. The owner was turning reflection probes on and off to see whether they're involved.Likely area: the sky / environment radiance that PBR shading reads for image-based lighting (sky env, reflection probes, or their fallback when probes are off or not ready) is far too bright, or reaches tonemapping unexposed. PBR relies on that environment term much more than legacy materials, which fits the colour loss on PBR only. Related: #872 (dark PBR materials tinted blue, also suspected to be sky / probe light leaking into PBR ambient and specular).
To narrow down:
A screenshot is in the owner's report.
Narrowed down with probes off at the same spot:
Conclusion: the root problem is the sky radiance itself (rendered sky + the sky term fed to lighting / IBL). The PBR washout is a symptom. Next test: DLSS vs FSR / native at the same spot, since the RR chain has its own exposure / un-expose step; then the sky preset / time of day.
Parked for now (owner). The sky code is identical between v1.4.0 and dev, and this parcel's environment is probably set very bright, so this is likely an old behaviour rather than a regression: an over-bright sky preset (no upper clamp after srgb_to_linear, uncapped scene-gamma HDR scale, and no exposure / tonemap for legacy skies with probe ambiance 0) with our HDR bloom on top. To revisit later: whether over-bright legacy skies should go through exposure / tonemapping or a capped bloom input, instead of clamping to white.