In our January 6 report, we highlighted NVIDIA’s CES announcement of DLSS 4.5. While the headline-grabbing “6x Multi-Frame Generation” remains locked behind the paywall of the RTX 50 series, we noted that the most critical improvement for flight simulation, the 2nd Generation Transformer Model (specifically “Preset M”), was coming to all RTX cards.
For the vast majority of Microsoft Flight Simulator 2024 pilots, the “ghosting” issue, where avionics smear into unreadable blurs during turns, has been the single biggest drawback of using DLSS. The simulator currently defaults to an older render profile (“Preset F”) that struggles with the high-contrast motion of digital flight decks.
Starting this week, however, you can force the simulator to use the new linear-space training model we detailed in our previous analysis. The result is arguably one of the single largest visual upgrades for glass cockpits. Here is how to unlock it using the official Nvidia App!
The New Standard: “Recommended” vs. “Latest”
In previous guides, you may be familiar with a “Latest” setting for DLSS Override in the NVIDIA App. Well, Nvidia has now replaced this setting with a nuanced option called “Recommended.” This setting automatically assigns the correct DLSS model based on your in-game quality preference:
- Performance Mode: Automatically uses Model M (The “Ghost Killer” optimized for high motion).
- Ultra Performance Mode: Automatically uses Model L (Optimized for 4K scaling).
- Quality/Balanced/DLAA: Uses the best fit, defaulting to Model M or K depending on the scenario.
The Official Setup
- Open the Nvidia App and ensure it is updated to the latest public release.
- Go to the Graphics tab and select Microsoft Flight Simulator 2024.
- Scroll down to “Driver Settings.”
- Locate “DLSS Override – Model Preset.”
- Select “Recommended”.
For RTX 40 & 50 Series Users: This is all you need to do. Your card supports native FP8 processing, meaning Model M (Preset M) will run efficiently, killing the ghosting without a performance hit.
The Critical Warning for RTX 20 & 30 Users
This is an important update from Nvidia. The new DLSS 4.5 models (M and L) are built for FP8 (8-bit floating point) compute, which is native to the newer Ada Lovelace (RTX 40) and Blackwell (RTX 50) architectures.
The RTX 20 (Turing) and RTX 30 (Ampere) series lack native FP8 support.
- The Risk: If you force “Recommended” (Model M) on an RTX 3080 or 2080 Ti, you may see a heavier performance impact because the GPU has to work harder to run the model.
- The Solution: If you switch to “Recommended” and notice a drop in FPS, Nvidia advises manually selecting “Model K” (DLSS 4) from the dropdown list. Model K offers the optimal balance of performance and quality for older architectures, though it may not handle ghosting quite as aggressively as Model M.
Final Thoughts
For the past year, choosing between DLSS (for performance) and TAA (for clarity) has been a frustrating trade-off in the MSFS 2024 graphics menu. With this update, Nvidia has given us a powerful new tool to bridge that gap, but it’s still important to manage expectations.
For pilots on RTX 40 and 50 series cards, this is likely the upgrade you have been waiting for. The linear-space training of Model M handles glass cockpits significantly better than previous iterations, making “Performance” modes viable for serious IFR flying.
However, as with any PC tweak, your mileage may vary. This update isn’t a guaranteed cure-all for every system, but it is a promising step forward for clarity in the age of AI upscaling. Test it, tweak it, and fly safe!
