I’m frustrated that RX 9000-series owners have largely been ignored on this issue for over 7 months, but I want to focus on constructive feedback.
Based on extensive testing, the blocky/square artifacts observed around lighting sources are not significantly impacted by FSR preset selection (Balanced vs Quality, etc.).
Artifacts are reduced but not eliminated when switching from FSR to TAA.
2 Distinct Artifact Types Observed
- Blocky/ square artifacts.
- Seen in cockpits and ground textures when aircraft-induced lighting is present. Fenix cockpit flood lights are a good example. PMDG 737-800 landing lights are a great example of how these artifacts appear on the ground. These artifacts are characteristic of tile- or block-based operations in the post-processing or temporal lighting chain
- Triangular/ Jagged Edge Artifacts
- Seen along wing edges, tail surfaces, cockpit window edges, and curved fuselage geometry. These manifest as crawling, triangular, stair-stepped edges and are consistent with rasterization / coverage / edge reconstruction failures, not classic spatial aliasing
Artifacts get worse over each session: Temporal accumulation
Both types of artifacts worsen over the course of longer flights. Visual instability increases gradually rather than appearing fully formed at scene load, which strongly indicates temporal history feedback or accumulation, rather than a static shader or lighting error.
This time-dependent degradation shows that corrupted data is being persisted and reintroduced frame-to-frame through the temporal pipeline.
Shared Root Cause:
Both artifact types can be symptoms of the same thing: Temporal history contamination at depth and normal discontinuities.
If Sim Update 2 altered the order in which emissives or bloom feed into the temporal history, precision changes (FP16 vs FP32) in lighting or intermediate buffers, motion vector generation or consumption, or history clamping/rejection thresholds,
THEN the observed behavior is to be expected. Block artifacts where low-res temporal or lighting buffers become corrupted, and jagged silhouettes where reprojection and clamping progressively fail at geometry edges.
Why FSR Presets don’t matter much:
FSR uses additional logic to mitigate ghosting and instability. However, if it’s getting incorrect inputs to start, then changing FSR presets from quality to balanced will not materially alter the artifacts. Switching to TAA will reduce the artifacts because it uses different history weighting and rejection heuristics.
All of this points to temporal issues further upstream than FSR.
The Smoking Gun
Artifacts worsen significantly when panning the camera, artifacts are present in daylight (look at the edges of backlit switches in the daytime), artifacts get dramatically amplified at night, and artifacts worsen over the course of longer flights.
This confirms a temporal reprojection failure.
Daytime presence rules out “night lighting only” explanations. Instead, SU2 likely introduced a regression in one or more always-on DX12 temporal inputs like motion vector accuracy or encoding, depth/ normal discontinuity handling, or history acceptance/rejection thresholds.
Night time conditions amplify the issue because of extreme HDR contrast, rapidly changing emissive lighting, bloom/glare behavior, and exposure/tone mapping instability.
Why DX12 Matters
Unlike DX11, DX12 places far more responsibility on the engine implementation for motion vectors, synchronization, precision handling, and temporal history management.
Since SU2, a DX12-specific temporal reprojection regression causes crawling/shimmering silhouette edges, blocky square artifacts around lighting, and progressive degradation over time (especially at night).
Unlike MSFS 2020, users cannot fall back to DX11, making this a blocking visual issue rather than a minor regression.
TL:DR
Since Sim Update 2, MSFS 2024 (DX12-only) shows a temporal rendering regression causing blocky square artifacts around lighting and crawling/shimmering jagged edges on geometry silhouettes. The issue worsens with camera movement, accumulates over longer flights, is present in daylight but becomes severe at night, and is largely unaffected by FSR preset choice (though switching to TAA reduces it). This behavior strongly points to a DX12 temporal reprojection/history issue introduced in SU2, not a driver-only or hardware problem.