I’ve recently been trying to get into VR Simming and have had mixed experiences with performance. I wanted to identify where the bottlenecks were since I plan to upgrade my computer soon, so I did a bunch of benchmarking and figured I’d share the results in case anyone else is looking for ‘real world’ performance data to compare against.
I have a 9800X3D on the way, ETA late February. I’ll redo the tests then to give people info on how a CPU upgrade impacts things with the same 4070ti GPU.
Also on reddit since the formatting in this forum is a bit bad… https://www.reddit.com/r/MicrosoftFlightSim/comments/1hwnch6/msfs_2020_vs_2024_vr_and_4k_benchmarks_5800x3d/
Ryzen 7 5800X3D
RTX 4070ti
32GB Corsair Vengeance LPX 2x16GB 3200
Samsung 980 Pro 2TB NVME
ASUS TUF x570 Plus Wifi
Quest 3 connected via Virtual Desktop
Main monitor (non VR) is 5120x1440 *so no actually 4k… but 89% of 4k
US East servers, 700MB/s up/down (tested and confirmed) on 1GB LAN. Testing done between 4:30 and 7:30AM Eastern so ‘server load’ should not be a significant factor.
Flight plan:
Asobo FA-18 Hornet, depart KJFK 31L in full afterburner. Maintain 1,000 feet MSL to Brooklyn, then perform unrestricted climb to 10,000 feet AGL, perform split S to 1,000 feet over midtown Manhattan and cross central park ~700 knots.
AKA, absolutely cooking it.
All AI traffic disabled, photogrammetry disabled.
All tests unless otherwise noted were run using TAA, since I find the ghosting with DLSS distracting in VR and wanted to get hard render numbers without Nvidia trickery. I didn’t even try DLSS in MSFS2024 since MSFS2024 is so much visually worse than 2020 I don’t care to bother trying to get more FPS in it.
Tl;dr:
MSFS 2024 ‘performs’ similar to MSFS2020 but looks FAR worse. Render distance for buildings is far worse, building textures are far worse, etc. Extremely noticeable, not ‘you have to look to see it’. For a low and fast flight plan, MSFS 2024 really does look like MSFS 2000. Again, PG off in all tests.
At every graphics level I see the CPU main thread reaching 100% load in spikes - which I see correlating directly with game stuttering in OBS recordings (specifics below on how I did this). Both MSFS edition’s main threads are also much spikier than their render threads - while the render threads will be fully loaded in both games, they’re consistently loaded (not spiky). So, I do think that a CPU upgrade will most directly impact stuttering performance.
Past Medium (desktop) settings, the 1% lows plummet. This might imply that while the CPU is directly correlated to stuttering at every performance level, once I upgrade that it will just more clearly show the GPU bottleneck at higher resolutions. But I’m not clear if it will be slower but smooth or slower and stuttering with a CPU upgrade. Stay tuned to find out!
| Column 1 | Avg FPS | 1% Low | CPU Load (Max) | CPU Max thread Load (% max) | GPU Load (max) | RAM Usage (max) | GPU VRAM Usage (max) |
|---|---|---|---|---|---|---|---|
| 5120x1440 | |||||||
| Medium | 43.3 | 26.8 | 56 | 85 | 78 | 4.61 | 7.59 |
| Ultra | 39.6 | 19.9 | 68 | 90 | 100 | 9.1 | 11.55 |
| Ultra low LOD | 38 | 20 | |||||
| VR | |||||||
| Medium | 33.7 | 6.1 | 70 | 99 | 99 | 13.09 | 11.57 |
| Ultra | 21.6 | 5 | 78 | 91 | 100 | 17.59 | 11.66 |
| Ultra DLSS Quality | 29 | 3.2 | 75 | 100 | 100 | 15.53 | 11.55 |
| Low End | 33.2 | 5.7 | 71 | 100 | 100 | 5.72 | 10.14 |
| Column 1 | Avg FPS | 1% Low | CPU Load (max) | CPU Max thread Load (% max) | GPU Load (max) | RAM Usage (max) | GPU VRAM Usage (max) |
|---|---|---|---|---|---|---|---|
| 5120x1440 | |||||||
| Medium | 47.5 | 22.1 | 100 | 100 | 99 | 14.76 | 11.24 |
| Ultra | 11.2 | 3.5 | 100 | 100 | 100 | 21.25 | 11.71 |
| VR | |||||||
| Medium | 32.7 | 7.8 | 99 | 100 | 99 | 15.56 | 11.68 |
| Ultra | 20 | 1.5 | 98 | 100 | 100 | 18.72 | 11.57 |
Discussion:
Starting with Desktop to give a baseline, then moving to VR.
MSFS 2020 Desktop (non-VR)
5120x1440 (again, almost 4k) Medium runs mostly cleanly in MSFS 2020. There were some minor stutters (note the 1% lows dropping below 30FPS) which are noticeable but not immersion breaking. I’m also not sure what’s causing the drops - max thread load, gpu load and gpu vram all have some headroom.
Going from Medium to Ultra increase all of CPU load, GPU load and VRAM usage. It’s playable, with some stuttering - but it’s honestly not that much worse than medium. We can see both GPU load and VRAM usage maxxed here, with some apparent CPU headroom.
MSFS 2024 Desktop (non-VR)
First: MSFS 2024 is visually worse in every way than MSFS 2020. Medium in MSFS 2020 looked far better than ultra in 2024. So, even though the FPS numbers look similar, the quality to the eye is MUCH worse in 2024. In MSFS 2020, I could see the New York skyline in the distance from takeoff at JFK even at low-end quality settings. In 2024, it only renders a low-poly Empire State Building from that distance even at Ultra quality. Once in Manhattan, MSFS 2024 has a much lower texture quality, it’s a bit of a meme but it does look like MSFS 2000. Live traffic, as a sidenote, is an absolute mess.
Granted, MSFS is not a ‘fly low and fast’ simulator for everyone, but for this flight plan, MSFS 2024 is a straight downgrade. So, the numbers lie a bit - MSFS2024 looks FAR worse for the same FPS. Ultra is completely unplayable.
MSFS 2020 VR
- Every single visual level has significant, noticeable stuttering - even low end.
- Ultra is clearly GPU limited (the CPU main thread was not running at 100%). All other tests were both CPU and GPU limited (both are maxxed). However - I do expect a CPU upgrade to improve the 1% lows here, since I noticed that the main thread was the thing most often correlated with stutters.
- DLSS Quality does improve average FPS… kind of. I ran this test at Ultra, and it both increased the average FPS and harmed the 1% and .1% lows - in other words, it made the average better but the stuttering worse. There was minor, but noticeable, ghosting - especially along the edges of the cockpit and fuselage. All in all, it’s not really a better experience and I wouldn’t expect DLSS to save you or your FPS in VR.
MSFS 2024 VR
Same comments as in desktop. It looks much worse, buildings and terrain detail are much lower than 2020 for any given settings level, rendering VFR nigh-impossible for this flight plan. It also is much less stable and crashed much more often. It loads all CPU cores and system RAM noticeably more than MSFS 2020, as well (and for no visual upgrade, so…). VR ‘performance’ is largely the same - so, it runs mostly the same but looks much, much worse.
Conclusion
I can’t really draw any comparisons about what bottlenecks my system - CPU, GPU and VRAM and consistently maxxed on every setting in both simulators. MSFS2024 does use significantly more RAM and loads all cores more in every single scenario for worse visuals.
Finally, correlating the developer mode overlay with perceived stuttering, I think that the CPU is the main thing causing noticeable stuttering, so I’ve got some hope the 5800X3D → 9800X3D change will directly improve the stuttering (specifically the base and boost clock upgrades: 3.4->4.5(32%) and 4.5->5.2 (15%)).
Extra detail:
CPU: PBO enabled, SMT on, standard clockspeed
Mem: XMP enabled, standard profile / no custom tunings
Mobo: Resizable BAR enabled
Monitor: Samsung Odyssey OLED G9 (5120x1440 - 89% of 4K)
VR Headset: Meta Quest 3 (2064x2208 per eye)
Testing methodology:
I used CapFrameX to record metrics. Between each setting change I exit to main menu, updated the settings, load into the game (on-runway) then enable CapFrameX recording once I’m rolling. I did two passes: one running OBS to record the video with the CapFrameX and Developer Mode overlays, and one with OBS and all overlays disabled, recording CapFrameX in the background (there was not a noticeable impact to performance between overlay recording and background monitoring). I also confirmed the clockspeeds/loads for CPU I was seeing in CapFrameX with Ryzen Master (they tied out fine).