Who likes data? 
My goal has always been to get MSFS working on 3 screens. My problem is that when I built my PC, I had 1 screen, and got used to High and Ultra settings – but keeping Ultra settings on 3 screens becomes … a challenge.
I wanted to understand the range of performance from Low to Ultra, on single and triple screens, with default and tuned settings. I started doing circuits but realized the number of settings combinations was going to take forever. After capturing data at various stages of flight - from gate, takeoff, cruise, approach, landing, the numbers showed that sitting at the gate / parking tends to be a fair way to measure, since TLOD and OLOD have significant impact at the airport. And I really just wanted to compare the same scenarios across different settings.
Testing Set Up
| Variable |
Test |
| BIOS |
Stock |
|
Tuned |
| Aircraft Locations |
C172 at KTEB Parking 83 |
|
A330 at KJFK Gate C31 |
| MSFS Preset |
Low |
|
Medium |
|
High |
|
Ultra |
|
Custom (High + Max 30, FGx2, VSync Half, TLOD 400) |
| Screens |
1 screen 4K |
|
3 screens 2-4-2 K |
|
3 screens 4-4-4 K |
Flight conditions:
- Few Clouds, No Traffic, Noon on May 5
- Load in cold and dark
- Start up, enter flight plan, capture data (screenshot)
Specs & Default vs Tuned
Edit May 9 - updated screenshot to include Rolling cache, and clarify VSync settings.
Results - A330-200 at KJFK
Results - C172 at KTEB
- 1 Screen = Limited by MainThread
- 3 Screens = Limited by RdrThread
I stick with DLAA because when you update DLSS using DLSS Swapper, the details in scenery, and in the cockpit, are really good, better than TAA for me. And because I sit close to large screens, I notice when those details are missing if I use DLSS Quality render scaling.
The performance numbers look worse on paper than they behave. The Dynamic setting does work (lowering LOD to preserve 30 FPS + FG), and things are mostly smooth even at a busy airport. I do have some minor jitters on 1 of 3 screens but I think that is down to the difference in cable (DisplayPort to HDMI adapter).
It becomes clear how beneficial Memory tuning is especially in demanding scenarios. Even if only minor differences show in FPS, most of the metrics and anything CPU related improve, and you can feel the difference. It’s also clear MSFS won’t use more than (I think) 36 GB of RAM even if more is available, and I wonder if that is arbitrary or Asobo might consider expanding that …
I’ve settled at a custom setting of mostly “High”, with Max Frames 30 + FSR Frame Generation x2, and still experimenting between 2-4-2 K and 4-4-4 K resolutions.