They were plentiful at my local computer shop. I did some like for like tests with the 7950x3D to compare. I'm no good at the benchmark graphs...sorry. The honest reason I decided to go with this CPU is simply that it's compatible with my current motherboard. Doesn't require any of the headache that goes along with getting a new Mobo.

 

In summary, in very heavy situations (all settings in Ultra & TLOD at 400), there was between a 5-12 FPS increase, and that's with Frame Generation enabled. So under extreme load, there's not much of an uplift at all. There was, however a noticeable uplift in fluidity (I'm assuming higher 1% and 0.1% lows)The scenarios I tested were:

FS2024
iniBuilds EGLL - All options enabled
FENIX
BATC w/ traffic at 6
 

iniBuilds KLAX (not native to FS24)
iniBuilds A350
No BATC
 

7950x3D (custom undervolt)

9950x3D (Stock, no undervolting)

 

 

In FS2020, I didn't take screenshots, but sitting at the gate in the FBW A380 at iniBuilds OMDB, with the 7950x3D I was getting ~65 FPS. With the 9950x3D I was getting around ~85-90 FPS. This was kind of a rush test (because it takes so long to load into FS2020), I only noticed at the end that the TLOD was at 280, so take those numbers with a grain of salt. I will, however say that it was quite a large uplift: ~10 real frames. Unfortunately, we're still a ways away from being able to get a solid 60 real FPS with no frame generation.

 

That being said, in scenarios that are more realistic for the average flight simmer (i.e. using AutoFPS), the CPU really started to show it's muscles. Higher TLOD on the ground with far less stuttering overall. At the time of writing this, I was doing a short flight in FS2024 from MK Scenery EIDW to iniBuilds EGLL. The whole flight (for the most part) was very smooth, however some of the same issues still remained (panning too fast causes stuttering/hitching). Also, if I was "out of the sim" for too long, the system would slow down. I have the feeling that it's something on my end (conflicting software/driver or something that needs updating).

Is it worth it? Well...I don't know yet. So far, it may not be, but this is with very limited testing and literally dropping it in, updating a few drivers, and that's it. The CPU temps which can't be seen in the above screenshots are about the same, maybe 2-3 degrees warmer (60 degrees as compared to 57 with the 7950x3D), but far lower than I originally anticipated, especially seeing how much CPU utilization there is in the flight sim. I think if your system is dialed in and you already have a 7950x3D, it may not necessarily be worth it...it is a $700 chip after all. I'll continue to test and update this post if it has any interest.

Glad you were able to get to a new proc. But for the money spend vs the gain in performance, you could have kept what you already had, unless it was failing. Just my opinion.

After doing more testing, Fs2020 has been the most fascinating. Frametimes are the most consistent I’ve ever seen. Using AutoFPS at AS EDDF with BATC traffic nearly full in the FBW A380, FPS are solid at 58 with a TLOD of 140. The most interesting part is how smooth everything is. It feels like more than that, but looking at the frametime graph, there are no stutters observed…at all. No stuttering down the runway or when changing views. I think that if this remains the case in other scenarios, a 5090 with 3 or 4x frame generation will be unbeatable in every situation. 

Do you think the same tests between a 9800X3D and 7950X3D would yield about the same deltas?

Thanks for taking the plunge and sharing your insights/results. Are you using any form of affinity masking to limit 2020 or 2024 to the X3D cores, or just letting the system run on all of them? I used process lasso in 2020 to limit my 7950X3D to just the X3D cores which resulted in a 20% or so improvement in FPS. The results in 2024 were nowhere near as significant. Not sure if the 9950X3D is going to be the same or if the new chipset drivers have made a difference. 

Edited March 13, 20251 yr by KL Oo

22 minutes ago, KL Oo said:

Thanks for taking the plunge and sharing your insights/results. Are you using any form of affinity masking to limit 2020 or 2024 to the X3D cores, or just letting the system run on all of them? I used process lasso in 2020 to limit my 7950X3D to just the X3D cores which resulted in a 20% or so improvement in FPS. The results in 2024 were nowhere near as significant. Not sure if the 9950X3D is going to be the same or if the new chipset drivers have made a difference. 

There is no need for that any more. One of the bigger features of new 9950x3d is its new CCD management. You can see in screenshots that FPC counter only recognizes 16 logical cores. It sees only one CCD. There is a new windows service that lies to games about processor and reports only one CCD wit 3DVcache.

Biggest take away is low 1% improvement to handle stutters. Tnx for report....

 

 

1 hour ago, Dankovic said:

There is no need for that any more. One of the bigger features of new 9950x3d is its new CCD management. You can see in screenshots that FPC counter only recognizes 16 logical cores. It sees only one CCD. There is a new windows service that lies to games about processor and reports only one CCD wit 3DVcache.

Biggest take away is low 1% improvement to handle stutters. Tnx for report....

 

 

You're referring to the OP's image? I just assumed he had turned SMT off and was showing the cores without SMT. @V1ROTA7E can you confirm?

I'm not entirely sure about "windows lieing to games about processes and only reports on the V-Cache CCD." My understanding from what I've read was that the chipset drivers were just better at ensuring processes where assigned to the most appropriate cores...be they the V-Cahce cores or the non-VCache cores.  I think core parking (which appears to be what you're referring to) isn't required anymore. 

Edited March 13, 20251 yr by KL Oo

You need the Windows GameBar being installed for this "feature" that Windows puts the game only on the CCD with Cache. No GameBar, no automatic assignement. Big show stopper for me. 

I currently have a 13900k with 8 cores (HT off) plus 12e-cores (4 disabled). With Process Lasso I have assigned all the side-apps (FSHUD, LNM, AutoFPS, Volanta, etc) on the e-cores leaving MSFS20/24 on the main cores.

I can say I am really happy with my performance, It is actually better in MSFS2024 than in 2020 sicne the multi-threading is helping a lot.

Back in december I built a new machine with 9800x3d and didn't find any improvement in the performance but most important, I didn't have the additional cores to run the side-apps so everything was in the same 8 cores (SMT off) and the stutters were noticeable. I sold that PC (I made some money on it since 9800x3d was scarce) and went back to my 13900k.

Now, the 9950X3D seems to have more cores, will I be able to use process Lasso to run the side apps on those cores and keep the Sim in the main cores? or how does it work?

1 hour ago, AnkH said:

No GameBar, no automatic assignement. Big show stopper for me. 

Why? it's just a game bar...

In the OP’s screenshots, I’m noticing that with both CPUs, all cores are in use. Whereas with my 7950X3D, the sim uses CCD0 with the 3D V-cache exclusively. Normally a few cores on CCD1 are used by other apps. 

2 hours ago, Ianrivaldosmith said:

Why? it's just a game bar...

I prefer as little such unnecessary tools as possible. Tidy systems pose the smallest risk of running into issues, so I only have things installed that are a must. Well, anyway not going to get a 9950X3D, so 😉

8 hours ago, KL Oo said:

I just assumed he had turned SMT off

Yep, SMT is off. 
 

regarding the affinity mask setting; not necessarily. After the changes AMD made to the chipset driver, I am using the “recommended” method. Balanced power mode, etc. if the game is in focus, you’ll see CCD0 at near full activity and CCD1 relatively idle. 
 

I do, however put background processes like BATC on CCD1

7 hours ago, ttbq1 said:

Now, the 9950X3D seems to have more cores, will I be able to use process Lasso to run the side apps on those cores and keep the Sim in the main cores? or how does it work?

That’s more or less it.

5 hours ago, Cpt_Piett said:

In the OP’s screenshots, I’m noticing that with both CPUs, all cores are in use. Whereas with my 7950X3D, the sim uses CCD0 with the 3D V-cache exclusively. Normally a few cores on CCD1 are used by other apps. 

This is the case with FS2020, FS24 I’m noticing far more overall cpu usage. I’m going to run some tests because even with the increase in usage, performance in FS24 appears to be quite a bit worse compared to FS20 and it was the opposite before the switch.