OFXR Bridge V439, release 36
This release adds per-game selection of NVIDIA DLSS Frame Generation
Preset A and the new performance Preset C, with measurements from actual
Galactic Racer racing in OpenXR Simulator.
Changes
- Select the FG model in OFXR. Open Quality and performance → DLSS
Frame Generation model → Choose game executable, then select
Preset A (standard), Preset C (performance), or Game default.
The selected executable survives tray restarts. Preset C requires the
updated NVIDIA app beta FG runtime and a compatible driver. - Preserve other NVIDIA settings. OFXR uses NVIDIA's public per-game
profile interface and verifies its writes. Game default removes the two
FG overrides; other profile settings are preserved. A global NVIDIA app
override can still apply, and a game profile can cover several executables. - Command-line control for repeatable comparisons. Run
OFXRBridgeTray.exe --dlssg-preset A|C|default|status "full game.exe path". - Better preset diagnostics and testing. Verbose NGX logging retains
runtime and effective-preset messages when the game allows OFXR's logging
callback. The newscripts/test-dlssg-preset.pycompares A/C using an
isolated temporary NVIDIA profile and retains its results and logs.
Restart required
Restart the game after changing A/C. NVIDIA caches the model within the
game process; changing the driver profile and recreating OFXR's FG feature
did not switch the active model in our test. Turning a controller upside
down still opens OFXR's existing panel, but does not provide a live A/C
toggle. FG resolution and the NVIDIA model preset remain separate settings.
Galactic Racer race results
Development-build measurements used Tatooine / King of the Racers with
the game's built-in autopilot, separate fresh game processes, an RTX 5090,
driver 617.42, NVIDIA FG runtime 310.9.16, and OpenXR Simulator at 90 Hz.
Both runs used 3004×3004 per eye, 100% native FG resolution, forced 2X,
shared stereo, and camera fit off. After warmup, each sample lasted 30 seconds.
| Preset | Median stereo FG GPU time | Rendered application FPS |
|---|---|---|
| A | 1.977 ms | 45.0 |
| C | 1.180 ms | 45.0 |
Preset C reduced measured FG GPU time by 40.3% (0.797 ms). This is GPU
headroom, not a measured overall FPS increase: 90 Hz / 2X pacing held rendered
FPS at 45. Both samples contained 1,355 GPU spans, continuously advancing
vehicle positions, active DLSS guides, and no reported OFXR control or guide
publication errors. These were separate race segments rather than identical
replays; the desktop mirror changed to fullscreen before the A sample.
Native correctness tests passed under both presets, and an isolated profile
test verified C → A → default → C while preserving an unrelated setting.
Image quality remains scene dependent; recorded replay tests showed a
performance/quality tradeoff. These simulator measurements do not establish
moving-head headset comfort or full-race visual quality.
See setup, verification, and detailed measurements.
Package and installation
The Windows x64 archive contains the tray, OpenXR and Vulkan layers,
benchmark, production NVIDIA DLSS-G runtime, documentation, and licenses.
The bundled SDK runtime is unchanged; the new Preset C model is supplied
by NVIDIA's app beta update. OpenXR Simulator is a separate project and is
not included. SHA256SUMS.txt lists the packaged files' checksums, and a
separate checksum accompanies the ZIP.
Disable OFXR and close games using it before replacing an existing install.
Extract the archive and run OFXRBridgeTray.exe to enable it again. For
Galactic Racer, select Griffin/Binaries/Win64/SWGR-Win64-Shipping.exe as
the game executable, choose the preset, and restart the game.