An isolated Big Lake test now renders directional sunlight and ambient fill on terrain, hills, buildings and replacement mechs. Matched tests measured a 2.7 percent average frame-rate cost for the new lighting. Campaign investigation and model previews continued alongside it, with unresolved gameplay defects and offline visual work kept separate from verified native rendering progress.
What changed
- Created a separate Big Lake sun-test launcher with its own game instance, pilot and registry. Existing Remaster and alpha profiles were preserved during the lighting experiment.
- Added configurable sun direction, elevation, strength and ambient fill. The light stays fixed in world space as the camera moves. Sky and HUD retain their existing rendering.
- Retained GPU geometry supplies model and world normals. A limited shader bridge reconstructs visible terrain positions and face normals from original projected geometry, extending sunlight to Big Lake's ground and hills without claiming a complete retained terrain scene.
- Native comparisons covered opposite sun directions and normal diagnostics. The final sun run matched all 1,000 original frame, draw-count and hash records. A sun-off repeat reproduced the control records and both captured images exactly. An earlier sun-off divergence remains recorded and unexplained.
Performance measurements
Four matched Big Lake runs used the same executable and route at 3440 by 1440, FXAA, 16x anisotropic filtering, a 1,200-unit draw distance and a 1,100-unit fog endpoint. Run order was reversed on repetition, with startup and final capture excluded from measurement.
- Sun off: 124.85 average FPS, 8.010 milliseconds per frame and 37.47 FPS for the 1 percent low.
- Sun on: 121.50 average FPS, 8.231 milliseconds per frame and 36.75 FPS for the 1 percent low.
- Sunlight reduced average FPS by 2.7 percent and added about 0.22 milliseconds per frame. Scene and geometry counts matched across all four runs; repeated captures matched within each lighting mode.
These are actual host-throughput measurements on a scripted route with a fixed simulation clock, not live-clock whole-game results. The average leaves room for further experiments, but slower frames still need investigation. Normal mapping, shadows and additional lights are not included in this cost.
A separate earlier Remaster-versus-alpha comparison found no reliable net GPU-geometry speedup: 87.2 versus 88.0 average FPS at the same distance settings, within run variation. Those separate-session results must not be used to infer a gain from the new sun-test numbers. CPU projection removal remains disabled in playable profiles because its matching-scene checks have not passed.
Campaign and display investigation
- Both regular and alpha playtesting profiles were updated to the requested 1,200-unit drawing distance and 1,100-unit fog endpoint, including authenticated main and zoom camera handling. Replacement-model fog remains a separate known gap.
- Campaign Mission 2's Bravo missile launcher remains an unresolved blocker. The player observed it disappear while approaching before firing, while it remained selectable and could not be hit. Both platform actors exist in private load tests; that does not reproduce or resolve the reported transition.
- A separate short-frame defect in original turret aiming was reproduced under a controlled high-frame-rate test. The optional correction prevented the observed collapsed aiming transforms and was selected for the regular playtesting build. This does not establish that Bravo's disappearance or mission progression is fixed.
- Mech Lab layout and high-frame-rate interactions still need user playtesting. Resolution-adaptive menus and settings for implemented graphics features remain planned.
Models and visual development
- Refined worn colors for Black Hawk, Champion and Vulture; retained Thor's current worn treatment. The Vulture texture update was installed locally, while new chassis candidates remain separate from runtime integration.
- Prepared offline donor-based visual candidates for Puma, Annihilator, Cauldron-Born, Sunder and Supernova. Walking, material and assembly checks support review; native retargeting, weapon placement, damage, distance transitions and gameplay validation remain open.
- Compared original-model Avatar, Firefly and Strider normal-map previews, reviewed an Owens donor, and prepared a more detailed Strider geometry prototype. These are offline visual experiments, not installed native features.
- Reaffirmed the stronger 12x-equivalent MW5 normal-map preview as the appearance target after reviewing higher-quality 12x and 24x comparisons. Native material strength will need its own calibration; preview settings do not enable normal mapping in the game.
What comes next
Development is paused at the user's request, with a rendering handoff saved. The recommended next feature is optional normal mapping on one replacement mech using the existing sun, followed by a bounded local-light test and one shadow caster and receiver pair. General outdoor shadows need real caster geometry collected before camera culling; visible terrain reconstruction cannot supply off-camera casters.
The current sun is direct light plus constant ambient fill. It does not sample normal maps, cast shadows or calculate bounced global illumination. Broader zoom, damage, distance-transition, reload and sustained-combat checks remain open. Vulkan validation layers were unavailable for the sun tests.
The accepted startup-root count remains 131 of 144. No whole-game fidelity, formal milestone or campaign-completion credit is added. Original game media, replacement assets, generated game code and native executables remain local; this update is not a downloadable release.
The source repository is currently private; GitHub links require collaborator access.