5800X to 5800X3D 'Upgrade' - User Report

[sanins192]

I just switched out my 5800X to a 5800X3D, paired with a FE 3090. I was seeing lots of good praise for the extra 3D cache and decided to give my AM4 motherboard a last hurrah with this upgrade.

Overall I am now very happy with how MSFS behaves, and the X3D gives a boost in performance in CPU demanding scenarios. I performed 2 benchmarks before and after switching out the CPU, both cases trying for a more demanding usage case vs many of the benchmarks from websites which show a chase plane view at several thousand feet. I’ve always been happy with the FPS at altitude, so I wanted to test how the new CPU behaved with more demanding scenarios with study-level planes at complex airports with AI traffic. My overall aim is to get smooth and consistent frame times.

Test Conditions:
FE 3090 FE
32GB 3600MHz (2x16GB)
M.2. SSD 1TB
Win10, HAGS Off, Game Mode Off
MSFS SU10, DX12, Ultra Preset, TAA + 100% Render Scale 3840 x 2160, Terrain LOD 200

Scenario 1:
PMDG 737-800 taxi @30 knots East to West along one of the taxiways at London Heathrow, cockpit view, heavy rain showers, default traffic on 100%.

You can see there is a decent boost of 9FPS to the average framerate, with the 1% and 0.2% also increasing. The frametime graphs look very similar in terms of consistency, but just showing an overall reduction in average time.

Interesting how there are still the 4 large frametime spikes, which are clearly a sim issue relating to the scenery loading in.

I wish I had stressed the CPU a little more by trying with AIG traffic and possibly the Fenix A320, but I would imagine the performance upgrade would show a similar trend.

Scenario 2
Beechcraft Bonanza starting on runway at Tokyo Haneda, take off and fly at 1000 feet North over Tokyo, scattered clouds.

Interestingly the 5800X nudges slightly ahead in this test, but not by very much. My feeling is that this test, which I thought would stress the CPU, is in fact not very heavy on the CPU and means that I am GPU bound in this scenario. Either way, I was quite happy with the ~50FPS performance over the very heavy Tokyo photogrammetry scenery.

My thoughts overall with the upgrade are positive. After buying the 5800X3D and selling the 5800X, this will cost me ~£200 to upgrade, and I think it is definitely worth it to boost the performance where it needed boosting the most, which was in CPU heavy scenarios on the ground in study-level airliners. I’ve no doubt the best setup in the next 6 months will be a 4090 + 7800X3D/7900X3D, but for me that is effectively a completely new PC, so I am very happy in the meantime with my setup and MSFS performance.

[nhc150]

I suspect you’re not entirely CPU limited, which is why the results were flipped for the second scenario. One way to ensure you’re CPU limited is cranking up terrain LOD to 400 instead of 200. This is almost always CPU limited, even at 4K.

[michihimself]

Thanks for testing! Exactly the 1. Scenario is the supermess for the mainthread and a cpu.
In the 2. Scenario I bet that you are in the GPU limit. The systems of the Bonanza are much lighter than those of the complex PMDG and much less burdensome for the main thread.
Tip, use the DEV mod to get more detailed information on what is limiting the performance.

[KanadeNyan]

Thanks for the test.
Of course the MSFS version and resolution are different from when I tested in Tokyo, but I thought that in your case the resolution is 4K and therefore the test is in the GPU area.
If you can get it, try to look at the total power of the GPU when benchmarking. You are probably using power to almost the limit as defined by TGP.
I would say that it is a very balanced configuration to be able to use the GPU to its limits with MSFS.

[Michail71]

Is your hyperthreading on or off? There was benchmark on here a while back that showed better results with SMT off.

Thinking of doing the same swap.

[sanins192]

SMT On for all tests above.

I did a bit more benchmarking with the new 5800X3D but didn’t share the results. Below summaries are using the test conditions as described above for Scenario 1, only changing the following variables for each test:

• SMT Off - No difference in FPS
• Game Mode On - No difference in FPS
• HAGS On - No difference in FPS
• DX11 - Lower average FPS (do not know the numbers off hand, but DX12 is much more fluid so this test was mainly out of curiosity)

‘No difference’ means the FPS results and frame time variances were within the test methodology repeatability (+/-5%), which I determined by performing 3 repeat tests.

So for my PC I have stayed with SMT On as there was no discernible difference and kept Game Mode and HAGS both off, as I have read reports of these causing performance issues. As I said, DX12 introduced in SU10 is the real winner as it evens out frame time variances vs DX11.

[Michail71]

For me DX12 performs much worse. But it might be from my VR configuration. It’s smooth, but periodically chokes up.

That other thread showed a significant difference with SMT off. Interesting how results can differ, but there are so many variables at play.

Thanks for the detailed testing!