5800X3D to 9800X3D Upgrade Benchmarks (MSFS 2020/2024, VR, 4070Ti)

Hi all, this is a follow up to:

I recently upgraded from a 5800X3D to a 9800X3D with the same graphics card, a 4070Ti, and wanted to give people a before and after to see the CPU impact on FPS and stuttering. NOTE - I did upgrade my RAM as well, 32GB to 64GB, so this is not perfect apples → apples. RAM data provided too for analysis though.

Full PC specs in the other post. Also I’m not going to post all the 5800X3D data in, just the 9800X3D and the difference, other post has 5800X3D data.

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.

All AI traffic disabled, photogrammetry disabled.

All tests unless otherwise noted were run using TAA at 100% render resolution. No DLSS.

Tl;dr: It’s complicated.

  • The CPU upgrade, across the board (*Note 1), saw improvements in FPS. However, 1% lows (stuttering) increased in MSFS 2020 Desktop - it ran faster, but a bit choppier.
  • The CPU upgrade drastically improved MSFS 2020 VR stuttering - but not enough for a smooth VR experience on it’s own.
  • 4070Ti is 100% compute-loaded in every configuration. It’s not powerful enough for MSFS 2020 or MSFS 2024 to run 60FPS stable on Desktop or 30FPS stable on VR.
  • 4070Ti is VRAM limited in every configuration except medium settings, where it’s only compute limited.
  • MSFS 2024 performs worse, is a RAM hog, and looks worse at low altitudes. The lighting is probably better, but subjectively to me, it’s not a revolutionary upgrade, it’s a marginal one, and not worth it for the worse and more inconsistent performance. I’ve uninstalled it to wait a year or two to see if Asobo gets it together. Two months after launch with terrible community outreach and low rate of progress is not leaving me with a good taste in my mouth though.

Detail:

  • 2020 vs 2024: 2024 is a memory hog. It used 2.5-3.25x as much RAM as MSFS 2020 (up to 25.58GB in Ultra VR). MSFS 2020 didn’t use any more memory between builds, MSFS 2024 increased it’s RAM use by 13% (Desktop) and 36% (VR). This may indicate my 5800X3D MSFS 2024 performance numbers were somewhat impacted by lower RAM - but I did have free RAM in BOTH configurations, so I really don’t think 64GB vs. 32GB is really improving MSFS 2024 performance.. That said, looks like 32GB RAM is totally fine for MSFS 2020.
  • 9800X3D made everything (*Note 1) run faster. 4-61% faster!
  • MSFS 2020 appeared to be CPU-limited on Medium settings with a 5800X3D + 4070Ti. Going to a 9800X3D improved FPS by a whopping 48% (I reran this test three times to confirm since I was so shocked).
  • MSFS 2020 saw worse 1% lows (stuttering) on Desktop runs with the 9800X3D, -25% across Medium and Ultra settings. You can see that the max thread was at 100% on the 9800X3D, whereas it sat at 85(Medium)/90(Ultra) on 5800X3D. So it’s getting better utilization, better FPS, but it’s still CPU-limited? Weird.
  • 1% lows improved significantly for everything except MSFS 2020 Desktop. MSFS 2020 VR saw a huge boost to 1% lows - 160 (Medium) / 152 (Ultra), and was noticeably smoother. 15.9/12.6FPS 1% low is still bad, but it’s a lot better than 6.1/5 FPS.
  • MSFS 2024 performs worse than MSFS 2020 in both configurations. Subjectively, it looks about the same in Desktop and worse in VR. Worse in VR because LOD is turned way down for the same preset level in VR preset graphics (Medium 2020 has much higher LOD than Medium in 2024 for the same framerate).
  • (*Note 1) Weirdly, the 9800X3D rig performed worse in VR. I assume there’s some configuration I did on my last install I haven’t done on this one - but it ran better in MSFS 2020 Desktop and VR and MSFS 2024 in VR. I’m going to chock that up to ‘this game is buggy’, this isn’t meant to detail how to max out MSFS 2024 performance, just to give a benchmark for the CPU. So, take that data as an outlier.

Numbers:

Format: Raw Number (% change vs. 5800X3D). So, 64.4 (48.7%) means it ran at 64.4 FPS, 48.7% faster than with the 5800X3D. Outliers are italicized

MSFS 2020 Avg FPS 1% Low CPU Max Thread Load GPU Load (max) RAM Usage (Max) VRAM Usage (max)
5120x1440
Medium 64.4 (48.7%) 22.2 (-17.2%) 57 (1.8%) 100 (28.2%) 5.26 (14.1%) 7.07 (-7.2%)
Ultra 41.5 (4.8%) 13.1 (-34.2%) 100 (11.1%) 100 (0%) 8.5 (-6.6%) 11.63 (.8%)
VR
Medium 36.5 (8.3%) 15.9 (160.7%) 88 (-12%) 99 (-1%) 5.52 (-3.5%) 9.79 (-3.5%)
Ultra 24.7 (14.4% 12.6 (152%) 99 (8.8%) 100 (0%) 10.59 (-39.8%) 11.62 (-.3%)
MSFS 2024 Avg FPS 1% Low CPU Max Thread Load GPU Load (max) RAM Usage (Max) VRAM Usage (max)
5120x1440
Medium 49.4 (4%) 27.9 (26.2%) 100 (0%) 99 (0%) 17.16 (16.3%) 9.91 (-11.8%))
Ultra 18.1 (61.6%) 7.9 (125.7%) 95 (-5%) 100 (0%) 23.59 (11.0%) 11.62 (-.5%)
VR
Medium 20.3 (-37.9%) 8 (2.6%) 77 (-23%) 99 (0%) 21.3 (36.9%) 10.81 (-7.4%)
Ultra 6.8 (-66%) 1.5 (0%) 88 (-12%) 100 (0%) 25.58 (36.6%) 11.71 (1.2%)
2 Likes

Thanks for the report.

I really didn’t see much of an advantage for those who don’t use VR and remain in 2020. Only 4.8% performance increase, from my perspective, doesn’t justify the amount invested.

Now when it comes to VR or already venturing into the still-under-development 2024, it really is a change of worlds.