Well, I know everything has been said about this processor, and obviously I´m not going to say anything new. But since I ran some numbers to compare it with my previous Ryzen 9 3900x in MSFS, I thought maybe someone could be interested too.
DISCLAIMER:
This is not intended to be a rigorous scientific test. Just some quick checks to compare with my base line and see if I wasted my money, or it was worth it.
TEST METHOD:
I don´t know how to recreate exactly the same conditions for both processors, so what I did was this:
I opened a new flight in the same gate of the same airport with the same plane. Then I selected exactly the same time of the day (12:00 UT) so the shadows and lighting were the same. I used the “scattered clouds” weather preset. Cloud coverage is similar this way, but individual clouds are not the same, so there are some differences. Then I moved the point of view to be as similar as in the previous test as possible, but didn´t turn crazy trying to match every pixel.
I can´t control IA aircraft at the airport, but don´t want to remove them because they are a must for me, and testing without them wouldn´t be the way I use the simulator. So some details varied from test to test, but in the end had little impact in frame rates.
I repeated the test three times with every location and processor, and used the median of the three. Results were very similar, despite little differences in clouds, POV or AI aicraft at the airport, with differences down to 1-2 fps. It´s good enough for me (maybe, not for you).
TEST CONDITIONS:
FS-2020 SU9 run as admin, 4K (100%), DX-11, TLOD 200, most settings in Ultra, a few in High (volumetric clouds), Photogrammetry On, IA Aircraft density in airport 100%.
VSync on, and 60 FPS cap. Yes, my mistake, I just forgot for the base-line with the 3900X, so I did the same with the 5800X3D. Not the end of the world, both would have scored a bit higher (never went above 60 fps in the complex scenery I used for testing).
System: Asus Prime X570 Pro (latest AMD drivers and BIOS), 64 Gb DDR4-3200, nVidia RTX-3080 driver 521.59. Sim installed in Samsung EVO 970 1Tb M2 drive.
Aircraft: A32NX experimental version 27/4/2022. Volotea 8K livery.
Tested at (external view and in-cockpit):
-
Madrid Barajas LEMD from LVFR, gate 325. Madrid default photogrammetry on (plus LVFR´s improvements on it).
-
Palma de Mallorca LEPA, from JustSim, gate 22, with Maestro Erzo´s extraordinary and highly detailed photogrammetry in the background.
-
Santander LEXJ freeware, gate 1. Again, with Maestro Erzo´s amazing Santander photogrammetry in the background. Only exterior view.
(By the way, I call Erzo “Maestro” because he deserves it, it´s my compliment to him, but it´s not his official name in Flightsim.to).
TEST RESULTS:
ADITIONAL COMMETS:
I was really surprised to see how much the new processor improved in LVFRs Madrid Barajas. Main thread times whent down to less than half (so, potentially could achieve more than double the frame-rate the 3900 did), and became limited by GPU, still with a healthy and unexpected almost 50% performance boost.
Unfortunatelly, this aiport was very stutter prone before, and when landing there with the 5800X3D, still gets some light stuttering. So, it performs much better, but still not good enough to prevent stutters on final. I don´t know what is causing this nor how to prevent it.
General impresion elsewere is of a MUCH smoother experience than before. Almost no stutters at all. The 3900X also suffered from some stuttering in Palma de Mallorca, and in this airport the 5800X3D was totally, completely, absolutelly stutter free now, despite it being a complex airport, and having Erzo´s Palma photogrammetry in the background.
Frame-rate there improved only 13%, but the feeling of it was much much better than the 46% gained at Barajas. Bottom line: what kills / saves the experience is not average frame rate, but the low 1% (which I didn´t measure). Hardly something new or surprising to anyone…
In the air, now I´m almost always hitting my V-sync 60 fps cap. That was unexpected, I didn´t think the 3900x was the limit in that situation, but aparently, it was (I only use developer mode very seldom, when performing tests like theese).
I´m happy with the upgrade.
I gave up a lot of performance for image processing sofware going from 12 to 8 cores, but that is something I seldom do. I had to wait several minutes anyway, so it´s not the end of the world. I play FS-2020 maybe 5 times a week, so for me, the change to the new processor is worth it.
The problem is… now I need a faster graphics card… 
Best regards.